Valley Recap

🏗️ Imasons: AI runs on software, but it scales on infrastructure ⏱️Two Weeks To Go fo AI INFRA 5! 💵 Bay Area Startups Collectively Secured $3B+ in April MTD

Sat, 18 Apr 2026 15:00:00 +0000

🏗️ AI runs on software, but it scales on infrastructure

For the past few years, most of the focus sat at the model layer. Bigger models, better outputs, faster iteration cycles. That’s where the attention went, and for a while, that’s where the bottlenecks lived.

That pressure has moved. What determines progress now sits lower in the stack, inside the physical systems that support everything above it. The limiting factor is no longer what can be built in software, it’s what can be sustained in the real world once it’s deployed.

The pace of buildout makes that clear. It’s industrial scale, happening fast enough to strain every dependency at once. Timelines are compressed, supply chains are tight, and multiple layers that used to operate independently now have to line up at the same moment for anything to come online cleanly.

Insight coming from Infrastructure Masons State of the Digital Infrastructure Industry 2026 report tracks how much of this is already in motion across construction, power, and compute, and how tightly those layers are starting to bind together.

Releasing on AI INFRA 5 // MAY1

That level of activity doesn’t move cleanly through a fragmented system.

Demand is increasingly being committed early. Capital is deployed ahead of delivery, capacity is often pre-allocated, and projects are moving forward with strong expectations that usage will be there when systems go live.

What determines success is whether that capacity can actually be delivered. Power sits at the center of that equation. Access to it, timing of it, and the ability to scale it reliably over time. In major markets, timelines to secure meaningful power can extend for years. That forces demand to move. Infrastructure is shifting toward regions where power can be secured with certainty, even if those regions weren’t part of the original footprint.

At the same time, the scale of what’s being built keeps increasing. Individual facilities have turned into multi-site campuses. Megawatt targets continue to climb. Delivery timelines are being pushed down even as project scope expands. More has to be built, and it has to come online faster, under tighter constraints.

That combination forces everything into the same frame. Energy, policy, capital, and local support aren’t separate tracks anymore. They show up together, and they have to be solved together. Projects are negotiated earlier, scrutinized more closely, and shaped by factors that extend well beyond the data center itself.

You can see the pressure building. Communities are more engaged, regulators are adapting in real time, and developers are being brought into conversations earlier than before. With less room to operate quietly in the background. The ability to secure approval and maintain support carries as much weight as technical execution.

At the same time, the underlying systems are being rebuilt to handle densities and workloads that didn’t exist even a few years ago. Power delivery, cooling strategies, and facility design are all under pressure to evolve at the same pace as the workloads they support.

Geography follows those constraints. The next wave expands beyond traditional hubs into regions where the full set of requirements can be met at once, power access, permitting alignment, capital flow, and local support.

That’s the shift that tends to get underestimated. AI runs on software, but it scales on infrastructure. Right now, the question isn’t what can be built at the model layer. It’s what the system can actually carry.

Featured Panel: AC to DC: Powering the AI Era

Because this shift isn’t theoretical anymore, it’s operational. The transition from 15 kW racks to 600 kW+ systems isn’t just a hardware upgrade, it’s a complete re-architecture of how power is generated, delivered, and consumed across the stack.

2 Weeks left! Sign up today

This conversation brings together the operators sitting at the center of that transition—from hyperscale data center buildouts to next-generation power systems engineering—to unpack what it actually takes to deliver capacity at this scale. Not in models or projections, but in live deployments happening right now.

The panel will focus on the real constraints shaping the next wave of AI infrastructure:

How power availability is dictating geography and timelines
What it takes to move from legacy AC systems to high-voltage DC architectures
How operators are redesigning facilities for extreme density and liquid-cooled environments
Where efficiency gains are real—and where physics still wins

This is where the abstract conversations about “AI scale” get grounded in execution. Because the question is no longer how fast models can advance. It’s how fast infrastructure can keep up.

Grab Your Ticket Today

Spotlight: Infrastructure Masons

https://imasons.org/

iMasons is a global nonprofit association of digital infrastructure industry professionals. Digital infrastructure is the information technology equipment that transmits, processes and stores data, the facilities that house that equipment and the networks that connect them all.

Industry veteran Dean Nelson founded iMasons in April 2016 to unite the builders of the digital age on a mission to aggregate, amplify, and apply their ideas, insights and initiatives to develop a greater digital future. Today, iMasons has more than 6,000 members, in more than 130 countries, representing more than $1.5 trillion in digital infrastructure deployments.

What iMasons Does (PIE: People, Innovation, Efficiency)

Creates a neutral forum to define priorities, standards, and direction for the industry
Builds a trusted global community where individuals come together to collaborate
Advances workforce development to close the talent gap at scale
Drives innovation through initiatives that turn conversation into action
Champions efficiency and sustainability across power, infrastructure, and operations

Why It Matters

Digital infrastructure is essential to how society functions. Demand is accelerating faster than coordination, trust, and shared understanding.
iMasons exists to bring clarity, alignment, and accountability to an industry shaping the future. Not just to connect people, but to turn intent into action at the right time

Who It Serves

The current and next generation of leaders, operators, and builders across the digital infrastructure industry.

Become a Member

Bay Area Startups Collectively Secured $3B in April MTD

Q2 started off quietly but picked up as business uncertainty around global events decreased. The April total hit $3B as funding activity increased this week. The eight megadeals so far were evenly spread between transportation, biotech, cleantech, fintech, computer networking and (of course) AI.

Exits, IPOs: SV Tech IPOs have been quiet since February, but this week Fremont-based Alamar Biosciences raised $191M in its IPO, pricing at the top of its $15 - $17 range. The company reacted to strong demand by upsizing their offering from 9.4 to 11.3 million shares, and their stock surged 40% in the first day of trading, taking their market valuation from $1.1B to $1.6B. Alamar raised over $300M from Illumina Ventures, Morningside Ventures, Sherpa Healthcare Partners and others. Their pre-IPO high valuation was $240M in 2023, making this an “up” IPO.

Webinar Alert & Date Change: The Fundraising with LinkSV series has been moved back one week to April 29. We'll cover when to start preparing and introduce a new tool for finding investors, customers and networking. Two attendees will receive 90-day Entrepreneur memberships, plus an individual session/deep dive on finding prospective investors with LinkSV. Interested? Register here for the link.

Early Stage:

Mariana Minerals closed a $73.5M Series A, software-first, vertically integrated minerals company supplying the minerals critical to modern energy, AI, and defense technologies.
Resolve AI closed a $40M Series A, AI for running and operating software in production, which underpins how businesses operate.
Parasail AI closed a $32M Series A, delivering a programmable deployment network that gives AI builders production-ready endpoints in minutes.
HexemBio closed a $10.4M Seed, developing regenerative therapies to rejuvenate aging blood stem cells.
Ultralight closed a $9.3M Seed, unifies EHR, practice management, and patient engagement with powerful AI natively built in.

Growth Stage:

Factory closed a $150M Series C, pioneering agent-native development.
Slash Financial closed a $100M Series C, a single platform purpose-built for the financial workflows of modern businesses
Zum closed a $100M Private Equity, managing every vehicle, driver, student passenger, parent, and ride on the Zum CMX™ platform.
Loop Payments closed a $95M Series C, a full-stack, verticalized AI platform for logistics and supply chains.
HockeyStack closed a $50M Series B, the enterprise revenue intelligence platform built on an event-based data architecture.

Your Feedback Matters!

Your feedback is crucial in helping us refine our content and maintain the newsletter's value for you and your fellow readers. We welcome your suggestions on how we can improve our offering. engage@ignitegtm.com

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

The Biggest Names in Food Are Testing This New Tech

What do the world’s three biggest foodservice providers, the biggest food manufacturer, and some of the industry’s most beloved brands have in common?

They’re all partnering with Automated Retail Technologies (ART) to serve food without kitchens or delivery.

With ART’s Just Baked™ automated robotic kiosks, serving peoples’ favorite meals 24/7 is affordable and easy. That’s why the foodservice giants Aramark, Sodexo, Sysco, and Compass Group joined Nestlé and White Castle as ART partners.

And ART’s 800 units deployed currently only scratches the surface of their growth plans. They have 400 more units ready for immediate deployment, 340,000+ additional targeted locations, and a leadership team that knows exactly what it takes to scale. In other words, ART is perfectly positioned.

Earn guaranteed bonus stock as an early-stage ART shareholder through April 25.

^{This is a paid advertisement for Automated Retail Technologies Regulation CF offering. Please read the offering circular at}^{https://invest.automatedrt.com/}

🦞NemoClaw: The Missing Layer Under Agents? 📈Largest Quarter of VC Investments Ever: $216B💰

Sat, 11 Apr 2026 15:00:00 +0000

🦞NemoClaw: The Missing Layer Under Agents?

Most people look at NVIDIA NemoClaw and see another agent framework. That misses the point.

What NemoClaw actually introduces is a layer that has been missing from the agent conversation entirely, something closer to an operating environment for autonomous systems than a model toolkit. Agents were already capable of planning, executing, and calling tools, but they were unreliable in production because nothing underneath them enforced how they behaved once they started acting on their own.

That gap shows up fast when agents move beyond demos. They read files, make network calls, access credentials, trigger workflows, and sometimes generate their own follow-on tasks. Without control at that level, the system becomes difficult to trust, especially in enterprise environments where data boundaries and auditability matter. NemoClaw is designed to sit directly in that gap, inserting policy, isolation, and control into the execution layer itself rather than trying to bolt it on afterward.

Jensen Huang CEO NVIDIA at GTC

That changes how agents can be used

Instead of being treated as tools that run occasionally, agents start to look like persistent processes that operate continuously across systems. NemoClaw is built to support always-on agents that can run locally, in the cloud, or across hybrid environments, with the expectation that they are active over long periods of time rather than spun up per request. That alone shifts the infrastructure profile. Compute demand stops behaving like bursts and starts behaving like background load that never fully turns off.

Security becomes part of the execution model, not a wrapper around it. NemoClaw uses a sandboxed runtime that controls what an agent can access, how it interacts with systems, and where data flows, effectively treating every agent action as something that must be governed in real time. This is less about protecting against failure and more about making autonomous behavior predictable enough to deploy at scale.

That predictability is what enterprise adoption has been waiting on.

Conceptual diagram of secure AI agent architecture (inspired by NVIDIA NeMo / NemoClaw)

Agent frameworks proved that autonomous systems could exist. NemoClaw is part of the push to make them usable inside organizations that care about compliance, data boundaries, and operational risk. Once those constraints are handled at the infrastructure layer, the conversation shifts from “can we trust this” to “how much of our workflow can this take over.”

Agents start chaining work together, calling other systems, and generating their own demand for compute. When those systems run continuously, the load they create compounds quickly. A single agent isn’t the problem. Thousands of them, each running multi-step workflows around the clock, start to look like a new class of infrastructure demand.

This is where NemoClaw connects back to the broader shift. It is making agents stable enough to scale, and scaling is what breaks assumptions across the stack. Power, cooling, scheduling, and cost models all begin to feel the impact once agents move from experimentation into production.

Join us on MAY1 at Plug and Play

This is Where AI gets Real.

This isn’t built for passive attendance. It’s where real estate, power, compute, and model teams are in the same room, working through these constraints together. It’s where the people deploying that $527B in CapEx meet the ones responsible for turning it into actual capacity.

If you’re building, buying, or connecting in this market, proximity matters. The conversations shaping what gets built aren’t happening in isolation anymore. They’re happening in rooms where the full stack shows up.

Join the waitlist today.

Highlights from Enterprise AI & HPC Summit NYC

On Thursday, the Enterprise AI & HPC Summit at Supermicro NYC made a few things clear: AI success is about operating at scale, securing deployments, and managing cost predictably.

Key takeaways:

Agentic AI is taking hold: Enterprises are shifting from single-user prompts to coordinated, multi-agent systems
Built for the AI factory: Supermicro and AMD highlighted H14 systems designed for EPYC “Turin” and Instinct accelerators, pushing memory bandwidth and efficiency where it actually matters
Cooling is now a gating factor: Direct-to-chip liquid cooling can reduce power consumption by up to 40%, and is quickly becoming standard at higher rack densities
From pilots to production: Teams across finance and enterprise are deploying AI and HPC closer to real workflows, fraud detection, trading, and core operations

Looking ahead: The “Open AI Factory” model is gaining ground, modular systems and open standards that let enterprises scale without getting locked in.

Live from New York!

Join us for AI Infrastructure Week

The South Bay is ground zero for AI infrastructure—and we’re turning it into a full week of high-signal events leading into AI INFRA SUMMIT.

In collaboration with Plug and Play, we’re bringing together the ecosystem across three days: kicking off Wednesday in downtown San Jose with a deep dive on data centers, followed by a full-day Physical AI program on Thursday with Telus Digital, and culminating in the AI INFRA SUMMIT as the grand finale.

Capacity is limited—lock in your spot early for AI Infrastructure Week See you there.

Join the waitlist today

WED APR 29th 3p-6p

Limited Space Available

THURS APR 30th

AI INFRA SUMMIT 5

FRI MAY 1st

APPLY FOR BUILDERS PASS

Builders Pass — Powered by Micas Networks

Scaling AI isn’t just about GPUs — it’s about how everything connects.

From high-speed switching to cluster-level interconnects, networking is now a core constraint in building AI systems at scale.

Together with Micas Networks, we’re offering a limited allocation of Builders Passes for engineers building across compute, networking, and infrastructure layers.

Application-only. Curated for builders.

→ Apply for access

Look Ma! We’re on the 101! (San Carlos Northbound)

CLAWCAMP // APR16

We're excited to partner with ClawCamp.us on the community’s full-day workshop event, Thursday, April 16th. Attendees experience a full-day of upskilling (beginner—advanced) that includes Partner Perks (incl. free OpenClaw agents), Onboarding (for beginners), Workshops (all levels), Skills Building, Skills Competition, Networking, and more!

You can even get your own M4 Mac mini, fully configured, if you purchase that package. RSVP and learn more about the event here

Bay Area Startups Collectively Secured $213B+ In Q1 2026

Q1 2026 was the largest quarter of VC investment ever, with more than $300B invested globally. $213B (70%+) of that went into Silicon Valley startups, fueled by OpenAI's $122B round. That and forty-five other megadeals made up more than 80% of the $213B total, with a majority of the money going to AI and AI infrastructure. But megadeals also closed in robotics, energy, biotech, fintech, communication, applied & vertical AI, as more capital continues to concentrate in fewer companies.

A similar dynamic is playing out in VC, as funds with $1B or more dominate fundraising and emerging managers continue to have an uphill battle. In Q1, Andreessen Horowitz raised $15B, Thrive $10B, Founders Fund $6B, Kleiner Perkins $3.5B, Battery Ventures $3.3B and Lux Ventures $1.5B. Those funds alone total more than half of the $66B 2025 full year total.

Webinar Alert: LinkSV is restarting our Fundraising with LinkSV series on April 22 at 4PM. We'll cover when to start preparing and introduce a new tool for finding investors, customers and networking. Interested? Register here for the link.

Webinar Alert: LinkSV's Tech, Talent & Investment Trends is next Friday, April 10 @ noon PT. Join us for a look at Q1 2026, the unprecedented funding levels and where Q2 is likely to take us. How long can AI Infra funding continue at this level, and where do the non-AI startups go for investment? We'll have updates on funding to women and the increasing obstacles to closing Series A after your seed round. Register here for the link.

Early Stage:

Aria Networks closed a $125M Series A, delivering the next frontier in AI networking made possible by advances in fine-grain telemetry, AI, and Cloud.
Noon closed a $44M Series A, the first design tool rebuilt from the ground up by product designers, for product designers.
AfterQuery closed a $30M Series A, an applied research lab investigating the boundaries of AI capabilities and curating high-quality datasets to power the world’s leading AI models.
Rork closed a $15M Seed, an AI app builder made specifically for mobile apps.
Foundry Robotics closed a $13.5M Seed, rebuilding American manufacturing as an AI-first, assembly-focused, dual-use Everything Factory.

Growth Stage:

SiFive closed a $400M Series G, transforming the future of computing by bringing the power and flexibility of RISC-V to the world.
EnerVenue closed a $300M Series B, commercializing energy storage systems designed to behave like infrastructure to meet the demands of AI-driven energy systems.
Endovascular Engineering closed a $80M Series C, a commercial-stage medical technology company focused on advancing the treatment of venous thromboembolism (VTE).
Route 92 Medical closed a $50M Series F, improving outcomes for patients undergoing neurovascular intervention.
Luminai closed a $38M Series B, enterprise AI automation platform built for healthcare operations.

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

Stack BTC while you sleep

Tired of trying to time the market? YieldClub puts your money on autopilot. Deposit from your bank account, and your balance starts earning automatically, routing yield into Bitcoin around the clock. No charts, no timing the market, no crypto expertise required.

Start Stacking BTC

🔥 SIGN UP FOR AI INFRASTRUCTURE WEEK. 💰Bay Area Startups Collectively Secured $20B in March MTD

Sat, 04 Apr 2026 15:00:00 +0000

AI infrastructure in 2026 doesn’t respect boundaries anymore. Power decisions hit model performance. Cooling limits shape rack density. Model behavior drives how hardware is actually used. Everything is tied together, and when one layer slips, the rest follow.

That’s where things start to break for teams still operating in silos.

Model builders can’t assume infrastructure will stretch to meet whatever they design. Power isn’t flexible on that timeline. Cooling isn’t optional. At the same time, infrastructure teams can’t keep building generic capacity and expect workloads to behave. What gets deployed now has to match what actually runs, down to how models consume tokens and how long they stay active.

✅ FollowGuy Massey, for tech and infrastructure truths made simple

So the work is shifting. Co-design is becoming the default way things get done. Model teams are sitting with facility engineers earlier. Operators are designing around specific workloads instead of broad categories. Decisions that used to be sequential are now happening side by side, because waiting introduces cost that compounds fast.

This is what connecting the stack looks like in practice. Land, energy, compute, and models all need to line up at the same time, under the same assumptions. Capacity isn’t something you just install anymore. It has to come together cleanly. Real estate without power stalls. Power without deployed compute sits idle. Compute without tuned workloads burns money faster than expected. The output that matters is usable intelligence, something that behaves more like inventory than infrastructure.

OpenClaw goes Enterprise

You can already see the pressure building at the enterprise layer. Systems like NVIDIA NemoClaw point toward a world where agents run continuously, planning, calling tools, and interacting with other systems without waiting for a human prompt. That kind of workload doesn’t spike and settle. It stays active, expands, and generates its own demand. Small inefficiencies don’t stay small when the system is always on.

That changes how scale works. It’s no longer about pushing more requests through a fixed pipeline. It’s about whether the entire system can support long-running, multi-step execution without costs drifting out of control or performance becoming unstable.

The advantage is starting to show up in how well teams coordinate across layers. Optimizing one part of the stack in isolation doesn’t carry the same weight anymore. Lining everything up does. When hundreds of billions in capital are moving each year, execution becomes the constraint, and execution depends on how tightly these pieces fit together.

Join us on MAY1 at Plug and Play

That’s the context behind AI INFRA SUMMIT 5. Join the waitlist today.

Visit SambaNova.ai

SambaNova will be joining us at AI INFRA SUMMIT, bringing its full-stack AI platform built to run large models efficiently in production. By combining purpose-built hardware with an integrated software layer, SambaNova enables a seamless path from model development to real-world deployment.

What SambaNova Will Showcase

Purpose-built AI hardware optimized for large-scale model workloads
Integrated software stack for both training and inference
Pre-configured systems that simplify generative AI deployment
Performance tuned for throughput, latency, and cost efficiency

Why It Matters
Running large models is expensive and complex. Most teams are stitching together infrastructure, software, and optimization themselves. SambaNova packages that into a single system, reducing time to deployment and improving efficiency at scale.

Who It Serves
Enterprises, government organizations, and AI teams deploying large language models and generative AI systems in production environments. Learn more at sambanova.ai.

Join us for AI Infrastructure Week

The South Bay is ground zero for AI infrastructure—and we’re turning it into a full week of high-signal events leading into AI INFRA SUMMIT.

Capacity is limited—lock in your spot early for AI Infrastructure Week See you there.

Join the waitlist today

WED APR 29th 3p-6p

Limited Space Available

THURS APR 30th

AI INFRA SUMMIT 5

FRI MAY 1st

CLAWCAMP // APR16

You can even get your own M4 Mac mini, fully configured, if you purchase that package. RSVP and learn more about the event here

Bay Area Startups Collectively Secured $7.2B in March MTD

The last day of March almost tripled total funding, with a $12B round for OpenAI plus other megadeals that raised the total to $20B for the full month. That brings the first quarter of 2026 to an unprecedented high of more than $200B, more than any previous quarter.

Exits, IPO dollar highs and number lows: 2026 is likely to top 2021 in IPO proceeds if Anthropic, OpenAI and SpaceX complete their offerings. All three are expected to be among the largest IPOs ever, with predictions of $75B for SpaceX and $60B each for Anthropic and OpenAI. Those three alone would exceed the 2021's $142B in proceeds and could discourage other IPOs, resulting in lower numbers than 2022's low of 181. But if the previous six-year cycle plays out again, 2027 could be a return to highs, both in numbers and dollars.

Early Stage:

Sycamore Labs closed a $65M Seed, the trusted agent operating system for the enterprise.
Cognichip closed a $60M Series A, developing the world’s first Artificial Chip Intelligence to re-imagine chip design.
NomadicML closed a $8.4M Seed, is the Visual AI platform for physical autonomy.
Miravoice closed a $6.3M Seed, an AI-powered platform for precision voice-based data collection.
Anvil Robotics closed a $5.5M Seed, builds low volume customized robots for teams at Nvidia, Google, Co.bot, Path Robotics, Ultra Tech, and more.

Growth Stage:

OpenAI closed a $12B Series unknown, an AI research and deployment company.
Also. closed a $200M Series C, an electric mobility company
depthfirst closed a $80M Series B, an applied AI lab developing novel security solutions for businesses facing modern, AI-era threats.
Alcatraz AI closed a $50M Series B, our mission is to make authentication secure and friction-less.
Stedi closed a $50M Series C, the only programmable healthcare clearinghouse.

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

Market Volatility Exposes Weak Delegation

When markets get shaky, advisors don’t just manage portfolios. They manage fear, questions, follow-up and a flood of client communication.

That’s where weak delegation gets expensive.

If meeting prep, paperwork, CRM updates and account admin still run through you, response times slip and the client experience takes the hit.

BELAY created the free Financial Advisor’s Delegation Guide to help you identify what to hand off, what to keep and how to stay client-facing without losing control.

Inside, you’ll learn how to reduce bottlenecks, protect responsiveness and free up more time for the work only you should be doing.

Download BELAY’s Guide!

🔥 From GTC: The Infra Era Has Arrived 💰Bay Area Startups Collectively Secured $7.2B in March MTD

Sat, 28 Mar 2026 15:00:00 +0000

🔥 From GTC: The Conversation Has Moved Below the Model

We just came out of three days at GTC—seven events, nonstop conversations, and a front row seat to how fast this market is shifting. What’s clear is that the center of gravity has moved.

This is no longer just about models.
AI has hit the physical layer.

Across builders, operators, and capital allocators, there’s a shared recognition that we’ve entered a new phase. The industry isn’t experimenting anymore—it’s industrializing. The concept of the “AI factory” is becoming real: systems designed to generate intelligence continuously, not just run workloads.

But that shift is breaking everything underneath it.

Most existing infrastructure wasn’t built for this. Power constraints, thermal limits, and network bottlenecks are no longer edge cases—they’re the defining challenges. Rack densities are pushing into territory that forces liquid cooling as a baseline, while energy strategy is moving closer to the source, with behind-the-meter approaches and localized generation becoming part of the new standard.

At the same time, the compute stack itself is evolving. New architectures are beginning to collapse the boundary between hardware and reasoning, embedding more intelligence directly into the system to support agentic workloads. The problem is that most of the world’s infrastructure can’t support this yet. We’re not just scaling—we’re running into physics.

That tension is reshaping the economics of the sector. The bottleneck is no longer software alone—it’s power, supply chain access, and the ability to deploy high-density systems at scale. Control over infrastructure is quickly becoming the defining advantage.

And just as that shift is happening, workloads are evolving again. We’re moving toward agentic systems that reason and execute over longer horizons, dramatically increasing demand on inference. These systems generate far more internal computation per interaction, shifting the burden from training to continuous, high-throughput execution.

What emerges is a different kind of challenge. It’s not just about compute—it’s about coordination. Latency, reliability, and the ability to maintain performance across increasingly complex systems are becoming critical. We’re effectively building infrastructure for a persistent, always-on layer of intelligence operating at global scale.

AI INFRA SUMMIT 5 · Luma

Friday, May 1, 2026 — Silicon Valley, California
Hosted by IgniteGTM

luma.com/aiinfra5

All of this is happening at once, forcing a level of alignment the industry hasn’t needed before. The constraints are physical, the timelines are compressed, and the capital requirements are massive.

That’s why we’re bringing this ecosystem together on May 1.

AI INFRA SUMMIT 5 in Silicon Valley is designed to cut through surface-level narratives and focus on what’s actually happening on the ground—where builders, operators, and capital come together to compare notes in real time.

If you’re involved in scaling AI infrastructure in any capacity, this is the room where those conversations are happening.Get your tickets for AI INFRA SUMMIT 5 here

https://www.plugandplaytechcenter.com/

Plugged in with PLUG AND PLAY

Plug and Play is a global innovation platform and active early-stage investor, with investments in 35+ unicorns, connecting startups, corporations, and investors across 20+ industries. With 65+ offices worldwide and partnerships with over 500 leading organizations, the platform helps bring new technologies to market and scale innovation globally. At the center of this effort are its Silicon Valley headquarters in Sunnyvale and its growing AI Center of Excellence in San Jose, serving as a hub for next-generation AI collaboration.

AI infrastructure sits at the foundation of every industry Plug and Play operates in, and is the focus of a dedicated week in Silicon Valley. Across its ecosystem, energy partners are addressing the rising power demands of AI, semiconductor companies are advancing the compute layer, and construction tech leaders are building the physical backbone at scale. At the same time, foundation and world model innovators are turning compute into intelligence, while enterprise and physical AI companies are deploying it into real-world systems and applications.

AI Infrastructure Week, April 29 through May 1 in Silicon Valley, brings together the full stack: three focused events, one week, and a room full of the people actually building what comes next.

Join us for AI INFRA SUMMIT 5

You can even get your own M4 Mac mini, fully configured, if you purchase that package.

RSVP and learn more about the event here

Bay Area Startups Collectively Secured $7.2B in March MTD

After a $1.4B funding total for the week, March's startup funding now stands at $7.18B. After last week's slowdown for GTC, funding rebounded – and that's before OpenAI closes the $10B addition to their February $110B round (coming soon).

This week is the start of the "demo day" spring season, with the Y Combinator Winter 2026 cohort having theirs this week, Berkeley Skydeck's Batch 21 cohort next week and Speedrun in mid-April. These accelerators are one of the key elements that drives the dynamic character of the startup ecosystem in the Bay Area, creating something close to a “network effect” for founders. These three high-profile accelerators alone – and there are many others - will graduate more than 800 startups in 2026, with a high percentage of those companies coming from or choosing to remain in the Bay Area.

Early Stage:

Gimlet Labs closed a $80M Series A, an applied research lab dedicated to building next-generation AI infrastructure.
Highlight closed a $40M Series A, the shared intelligence layer that turns your team's work into collective knowledge.
Deccan AI closed a $25M Series A, helps frontier AI labs and enterprises improve model performance through research-grade evals, benchmarks, and novel datasets at scale.
EPIC Microsystems closed a $21M Series A, a leading semiconductor company which possesses cutting-edge know-how and experience in the area of power semiconductor products and solutions targeted at Artificial Intelligence and Data Center infrastructure.
Memories.ai closed a $16M Seed, building Large Visual Memory Models, giving machines the ability to see, understand, and recall visual experiences across unlimited time-frames.

Growth Stage:

Harvey closed a $200M Series F, a modern, AI-native platform that helps physical product businesses manage the flow of goods, dollars, and data across procurement, inventory, orders, fulfillment, and finance in real time.
Zipline closed a $200M Series H, the world's largest and most experienced autonomous delivery service.
Xona Space Systems closed a $170M Series C, the navigational infrastructure company bringing centimeter-level certainty to any device, anywhere on Earth.
Doss closed a $55M Series B, a modern, AI-native platform that helps physical product businesses manage the flow of goods, dollars, and data across procurement, inventory, orders, fulfillment, and finance. in real time.
Shepherd closed a $42M Series B, provides insurance for the builders and operators shaping our physical world.

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

Your $4T Wealth Multiplier: Secure an 18.8% return

Why settle for 2% dividends when Wall Street dominates a $4 trillion market yielding 3x that? You can access the Goldman Strategy thanks to mogul.

They let you invest in elite rentals for less than the cost of a new Rolex. The Ex-Goldman veterans who handled $10 billion of institutional capital even pick the properties and manage them for you. That means all you need to do is watch the rent checks hit your bank account and relax.

Here’s The Breakdown:

7-12% Annual Cash Yields
18.8% Average Annual IRR for superior long-term growth.
Full Tax Benefits of direct real estate ownership.

Stop watching from the sidelines while institutional giants harvest the yield in your neighborhood. Secure your share of the most resilient asset class on earth today.

Unlock New Deals Now

Past performance isn't predictive; illustrative only. Investing risks principal; no securities offer. See important Disclaimers

🔥 GTC’26: Where AI Factors Meet Human Chemistry ⚛️ 🤖 Are Bots Your New Users? 💰Bay Area Startups Collectively Secured $5.8B in March MTD

Sat, 21 Mar 2026 15:00:00 +0000

AI infrastructure has been built around a simple assumption: a human initiates a request, a model returns a response, and the interaction ends. That model begins to break down as agents enter the system. What replaces it is not a faster version of the same pattern, but a different kind of workload altogether, one where machines generate demand on behalf of other machines.

Agentic systems do not operate in single turns. They plan, decompose tasks, call tools, evaluate outputs, and iterate before producing a final result. Much of that activity never surfaces to the end user, yet it still consumes compute. A single request can expand into dozens or hundreds of internal steps, each requiring inference, memory access, or external calls. The visible output may look similar to a chatbot response, but the underlying workload is significantly heavier and far less predictable.

This shift changes how compute demand accumulates. Traditional inference models assume relatively stable cost per request, which allows operators to forecast capacity and optimize around latency and throughput. Agentic workloads behave differently. The number of tokens generated per task increases, the duration of each interaction stretches, and resource usage depends on how the agent chooses to solve the problem in real time. That variability introduces a level of non-determinism that existing infrastructure was not designed to handle.

Traffic patterns are changing alongside the workload itself. Instead of humans driving the majority of requests through interfaces, agents increasingly communicate directly with other services, models, and agents. APIs become the primary surface area, and machine-to-machine interaction begins to dominate over human-driven sessions. The web starts to look less like a collection of pages and more like a network of continuously interacting processes.

That transition carries implications beyond raw compute. Systems designed for human interaction rely on clear boundaries, authentication models tied to individual users, and relatively simple billing structures. Agent-driven environments require a different foundation. Agents must be able to discover each other, establish trust, negotiate access, and exchange value without human intervention at each step. Existing API frameworks were not built for that level of autonomy or scale.

Emerging standards aim to address these gaps by defining how agents identify themselves, verify intent, and transact with one another. Protocols for discovery, reputation, and payment are becoming necessary components of the stack, particularly as agents begin to operate across organizational boundaries. Without those standards, the system fragments into isolated silos, limiting the effectiveness of agent-based workflows.

The result is an infrastructure environment that must accommodate higher compute intensity, less predictable demand, and a growing volume of machine-originated traffic. Planning for that environment requires a shift in how capacity is modeled, how services are exposed, and how interactions are governed.

As agents take on a larger share of digital activity, they reshape the assumptions that underpinned the previous generation of internet architecture. The next phase of AI infrastructure will be defined by how well it supports systems that no longer wait for humans to act.

Watch Jensen’s full keynote (2:20)

🔥 GTC’26: Where AI Factors Meet Human Chemistry ⚛️

We just wrapped an intense week at GTC, hosting seven back‑to‑back gatherings that brought AI leaders, builders, and visionaries into the same rooms while 30,000+ attendees turned downtown San Jose into an “AI campus.” While the main conference spotlighted AI factories, physical AI, and the next wave of infrastructure, our focus was simpler and more human: curated spaces where founders, operators, and investors could trade notes on what’s actually working, compare playbooks, and build relationships that will outlast any single product announcement.

Supermicro NVIDIA Networking Party kicked off the week with energy

Across intimate executive dinners, late‑night parties, and builder‑led sessions, a few themes kept coming up: everyone is feeling the acceleration in demand for AI infrastructure, nobody believes they can win alone, and the most valuable conversations are still happening off the main stage—in the corners where teams compare architectures, hiring bets, and go‑to‑market experiments. GTC made it clear that AI is now core infrastructure; our events made it equally clear that this new era will be shaped by tight, trusted networks that share real metrics, failures, and wins in the open.

Ignite The Edge Party was 🔥🔥🔥 - Best pics #WEKAignitesAI wins Meta Ray-Ban glasses!

If you felt that energy at GTC—or wish you had—AI INFRA SUMMIT on May 1 is the next stop. Hosted by IgniteGTM, it’s a full‑day summit in Silicon Valley designed to reunite this community and go deeper on the infrastructure race that now defines AI, with visionary keynotes, technical deep dives, and networking built for serious builders. Join us to keep the momentum going: AI keeps rollin’, infra keeps turnin’—and the people you met at GTC will be there, ready to write the next chapter together

Join us for AI INFRA SUMMIT 5

Bay Area Startups Collectively Secured $5.8B in March MTD

This was a quiet week, with just $1.3B in startup funding, for a March MTD of $5.8B. All AI-related eyes were on NVIDIA's GTC, with most companies preferring to delay their funding announcements. Frore Systems was the sole AI infrastructure-related funding in the Top Five. M&A was slow and the IPO market has been quiet - no, dead - as Kraken chose this week to pause their IPO plans due to the downturn in the crypto market.

After last month's record-breaking funding total, this month is settling into a less-frothy pattern that is closer to last year's monthly numbers. The March 2025 spike included OpenAI's $40B round; with a little over a week to go, the March 2026 total is well below that.

Webinar alert: Founders, next week don't miss: Legal Foundations, from Seed to Sale – Part 12 (the final edition), and LinkSV Fundraising – Finding Angels (plus partners, customers and more)

Follow Link Silicon Valley on LinkedIn to stay on top of what's happening in and around startup funding, venture capital and key players in the startup ecosystem.

Early Stage:

R1 Therapeutics closed a $77.5M Series A, focusing on the development of first-in-class therapies for patients with kidney disease.
Andromeda Cluster closed a $60M Series A, building the systems, network, and orchestration layer that makes the world’s AI infrastructure more accessible.
Halycon closed a $21M Series A, the AI platform for energy.
RAVEN.IO closed a $20M Seed, a cybersecurity company that protects applications at runtime and prevents both known and unknown attacks in real time.
Fort closed a $.5M Pre-Seed, building the most comprehensive personal health tracker on the market.

Growth Stage:

Frore Systems closed a $143M Series D, a pioneer in advanced thermal technologies that unleash performance across data centers and edge devices.
Rox closed a $138.5M Series B, provides AI Agent Swarms to help top sales teams grow their books.
RoboForce closed a $52M Series Unknown, building the future of Physical AI — scalable, deployable Robo-Labor designed for demanding industrial environments.
Beautiful.AI closed a $45M Series B, presentation software that makes it easy to turn your ideas into beautiful visual stories.
Paraform closed a $40M Series B, the hiring marketplace for elite recruiters helping companies fill their most critical roles faster.

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

Here’s how I use Attio to run my day.

Attio is the AI CRM with conversational AI built directly into your workspace. Every morning, Ask Attio handles my prep:

Surfaces insights from calls and conversations across my entire CRM
Update records and create tasks without manual entry
Answers questions about deals, accounts, and customer signals that used to take hours to find

All in seconds. No searching, no switching tabs, no manual updates.

Ready to scale faster?

Start now →

🔊Announcing First Wave of AI Infra Summit 5 Speakers 💪🏼 Who's Ready for GTC?! 💰Bay Area Startups Collectively Secured $4.5B in March MTD

Sat, 14 Mar 2026 15:00:00 +0000

Join Us There

The first featured speakers for AI INFRA Summit 5 are ready to be announced. We are excited to welcome Jeremiah Owyang (Blitzscaling Ventures), Gaurav Saxena (Ford Motor Company), and David Campbell (Scale AI) to the stage. These leaders are working at the front edge of the systems that power modern AI, from platform engineering inside large enterprises to the security and control models required for autonomous agents. This is only the first wave, with more speakers from across the AI infrastructure ecosystem to be announced soon.

The scale of the current data center expansion is difficult to grasp without stepping back from individual projects and looking at the numbers in aggregate. By 2030, global data center investment is projected to approach $7 trillion. Nearly two trillion dollars of that will go into construction alone, making the physical act of building these facilities the second-largest cost category behind the servers themselves.

That amount of capital compresses decision timelines across the entire value chain. Hyperscalers, colocation operators, energy providers, and private infrastructure investors are all deploying capital simultaneously, often across multiple campuses and regions. Finance teams are now responsible for allocating billions of dollars under conditions where delays cascade quickly, supply chains remain tight, and demand for compute continues to accelerate.

Under those conditions, traditional project planning methods begin to break down. Static construction schedules were built for projects where sequencing rarely changed and disruption could be managed manually. The current environment looks very different. Equipment lead times fluctuate, permitting timelines shift, and design specifications evolve as AI workloads push density higher. When dozens of construction tasks interact with hundreds of supply constraints, a fixed schedule becomes fragile.

Generative scheduling has emerged as one response to that complexity. Instead of relying on a single predefined plan, the approach evaluates thousands or millions of potential construction sequences while accounting for resource availability, site conditions, and project constraints. Planners can then test how different decisions affect delivery timelines or cost structures, allowing teams to identify faster execution paths or mitigate the impact of disruptions before they cascade through the project.

The benefit is not simply faster scheduling; it is better visibility into trade-offs. Finance teams can translate schedule changes into capital implications, evaluating how acceleration affects cash flow, procurement timing, and milestone payments. A delayed transformer delivery might extend construction by several weeks, but it may also delay revenue recognition or increase financing requirements. Scenario modeling makes those consequences visible early enough for executives to act.

Speed, however, is only half the challenge. The intensity of the current buildout forces developers to maintain tight discipline over costs, particularly when construction spending alone approaches two trillion dollars globally. That pressure has revived interest in should-cost modeling, a method that builds project budgets from first principles rather than relying on historical estimates or supplier quotes.

By breaking facilities down into materials, labor inputs, equipment costs, and overhead, finance teams gain a clearer view of what a project should cost under realistic conditions. The approach often reveals inefficiencies in design assumptions or procurement strategies that would otherwise remain hidden. Even modest improvements can compound quickly across portfolios measured in gigawatts.

The companies that master both speed and financial discipline will define the next phase of infrastructure development. In an environment where trillions of dollars are moving into the physical backbone of AI, execution quality has become just as important as access to capital.

Join us for AI INFRA SUMMIT 5

READY FOR GTC WEEK?

Bay Area Startups Collectively Secured $4.5B in March MTD

This week funding activity bounced back from last week's slowdown, bringing the month to $4.5B.

Mega-size VC funds continue to dominate the VC fundraising landscape. After Andreessen Horowitz' $15B raise in January, Lightspeed Venture Partners announced $9B across multiple funds, and Thrive Capital closed their $10B Thrive X fund. Now we hear that General Catalyst is raising a $10B fund, Founders Fund is close to closing a $6B fund and Spark Capital is working on a $3B raise. Per Pitchbook, the year started with ~$300B in dry powder and has seen a steady run of $100M+ and some $1B+ megadeals. These latest rounds indicate that the dry powder is being steadily restocked, underline the continuing consolidation of venture capital into fewer firms and larger funds.

For startups raising capital: Stay on top of who's raising, who's closing and who's investing with the Pulse of the Valley weekday newsletter. Founders get the newsletter, database and alerts for just $7/month ($50 value). Check it out and sign up here.

Follow Link Silicon Valley on LinkedIn to stay on top of what's happening in and around startup funding, venture capital and key players in the startup ecosystem..

Early Stage:

Mind Robotics closed a $500M Series A, building the world’s leading industrial robotics platform, capable of performing dexterous, variable, and reasoning-intensive tasks.
Rhoda AI closed a $450M Series A, building a new class of robot foundation model designed to bring general intelligence into the physical world.
Eridu closed a $200M Series A, advancing the frontier in AI networking.
Standard Kernel closed a $20M Seed, building AI infrastructure with AI.
Peltier closed a $5.6M Seed, powers active, connected, and precise cold chain solutions.

Growth Stage:

Nexthop AI closed a $500M Series B, building the most efficient Al infrastructure for hyperscalers and the world's largest cloud operators.
Quince closed a $500M Series E, an affordable luxury brand that sells high-quality fashion and home goods at radically low prices— direct from the factory floor.
Replit closed a $400M Series D, the agentic software creation platform that enables anyone to build applications using natural language.
Sunday closed a $165M Series B, building robots for busy households, delivering in 2026.
ORO Labs closed a $100M Series C, a procurement orchestration company on a mission to humanize the procurement experience.

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

Your Tax Data, Finally in One Place

Are you tired of hunting down data, fixing errors, and manually updating disconnected spreadsheets?

Tax reporting isn’t a simple as it used to be. You need real-time, flexible reporting so you can confidently make decisions backed by accurate, centralized data.

Learn how bringing all your tax information into one central system automates repetitive tasks, improves scenario planning, and frees your team to focus on strategy instead of data entry.

Whether you operate in one country or dozens, Longview Tax scales with you—reducing risk, speeding up your close process, and helping you optimize tax policies across all jurisdictions.

Download the paper

✍️AI Workloads Are Being Rewritten💰Bay Area Startups Collectively Secured $193B

Sat, 07 Mar 2026 16:00:00 +0000

Join us for AI INFRA SUMMIT 5

For the past several years, AI infrastructure planning has revolved around training. The largest models demanded enormous clusters, enormous power draw, and enormous capital outlays, and the industry treated those training runs as the central event around which the rest of the ecosystem revolved. That emphasis is now shifting as the economics of AI move from model creation toward model usage.

By the middle of this decade, the majority of compute demand is expected to come from inference rather than training. Estimates across the industry point to a crossover point around 2027, when the cumulative workload of answering queries, generating content, and powering software agents exceeds the compute devoted to creating new frontier models. The infrastructure priorities that follow from that shift look very different from those that defined the early model race.

Training rewards raw scale. Large clusters maximize floating-point throughput, and the objective is to compress months of compute into weeks. Inference rewards efficiency and responsiveness instead. Each user interaction triggers a sequence of token generation events that must occur quickly and at extremely low cost, since those tokens accumulate across millions or billions of requests. The question operators increasingly ask is not how many FLOPS a system can deliver, but how much it costs to produce each token delivered to an end user.

That economic pressure exposes a major inefficiency in the current deployment model. Many production services rely on models originally designed for training runs with hundreds of billions of parameters, which makes them expensive to operate when used repeatedly at scale. Running those models for everyday inference can quickly produce negative margins once energy, hardware depreciation, and infrastructure costs are accounted for. In response, companies are shifting toward model distillation and quantization strategies that compress capabilities into far smaller architectures.

The emerging consensus places many production deployments in the range of seven to eight billion parameters, where models retain strong performance while dramatically reducing memory footprint and compute requirements. Teams at large platforms have reported that these smaller models often occupy a practical “sweet spot,” delivering usable intelligence without the operating cost profile associated with frontier-scale networks.

This change in workload also exposes a different architectural constraint. Inference performance depends less on pure compute throughput and more on memory bandwidth and latency, since generating tokens requires repeated access to model weights and cached context. As a result, the economics of inference increasingly favor systems designed around fast memory access and efficient caching rather than simply adding more accelerators.

Some operators have begun describing this model as a kind of token warehouse. Context is loaded and prepared once, and the system focuses on decoding tokens as efficiently as possible across many requests. The objective is to reduce repeated computation and spread the cost of preparation across a large number of responses.

As inference demand grows, infrastructure strategy follows the economics of tokens rather than the spectacle of training runs. The next phase of AI deployment will be defined less by the largest models ever trained and more by the systems capable of delivering useful tokens quickly, reliably, and at a price that makes the business work.

GTC Week is almost here and San Jose is about to become the epicenter of the AI world.

Thousands of builders, founders, engineers, and investors will be in town, and the best conversations won’t happen on the expo floor — they’ll happen after hours.

That’s why we’re hosting a series of curated evening events during GTC week designed for high-signal networking with the people actually building the future of AI infrastructure.

⚡ Founders building the next generation of AI platforms
⚡ Infra leaders scaling trillion-parameter compute
⚡ Operators solving the real constraints of power, cooling, and clusters
⚡ Investors backing the AI infrastructure race

Whether you're a builder, operator, or investor — these gatherings are where the real conversations happen. Below are a few of the featured events happening this week. Spots are limited, so secure your place early.

The Party of Parties

Limited Availability

Reserve Your Spot

Bay Area Startups Collectively Secured $193B in 2026 YTD

March has started out slow with just two megadeals – Ayar Labs, $500M and Science Corporation, $230M - in contrast to February's massive fundings. Out of February's $163B total, 98.7% of the funds - $161B - went out via twenty-six megadeals. And out of that $161B, just three companies received 95% of the investment - OpenAI with $116B, Anthropic with $30B and Waymo with $16B. Two months into the year, 2026 YTD now stands at $193B, over 90% of the total funding raised in the full year of 2025.

Follow Link Silicon Valley on LinkedIn to stay on top of what's happening in and around startup funding, venture capital and key players in the startup ecosystem.

Early Stage:

Cylake closed a $45M Seed, building a complete, AI-native, data-driven cybersecurity platform for the world’s largest and most regulated institutions.
Guild.ai closed a $44M Series A, provides all-in-one AI that helps builders do their AEC office work, from report writing to 2D to 3D renders.
JetStream Security closed a $34M Seed, an enterprise AI governance and AI security platform built to control, secure, and scale agentic AI in production.
DiligenceSquared closed a $5.4M Seed, enables leading investors and corporates to make critical investment decisions with confidence through AI-powered market research and commercial due diligence.
Vor Systems closed a $3M Pre-Seed, an AI-enabled transaction platform built for complex energy deals.

Growth Stage:

Ayar Labs closed a $500M Series E, transforming AI infrastructure with the industry’s first proven co-packaged optics (CPO) solution.
Science Corporation closed a $230M Series C, focused on solving some of neuroscience's hardest questions and most serious medical unmet needs.
Portside closed a $65M Private Equity round, a leading provider of advanced software solutions for the global aviation industry.
Polares Medical closed a $50M Series C, a structural heart company developing a novel transcatheter therapy designed to treat patients suffering from mitral regurgitation.
ArmorCode closed a $16M Series B, helps enterprises manage security risk and governance across today's heterogeneous technology environments.

Brandon Omoregie, Co-Founder of IgniteGTM, will be leading a new weekly segment in the newsletter focused on open roles across the AI Infrastructure landscape.

In addition to his work with Ignite, Brandon is the founder of The Offr Group, where he helps high growth technology companies hire exceptional talent and connect experienced operators with the right opportunities.

What Brandon Will Cover
Each week Brandon will share a curated set of open positions across companies in our ecosystem, including startups, infrastructure providers, and growth stage AI companies.

What Readers Can Expect
• High signal job openings across AI, infrastructure, and startups
• Roles sourced directly from founders and hiring managers
• A simple way for companies to share open positions with the community

Why This Matters
The AI infrastructure ecosystem is scaling quickly. New companies are launching, teams are expanding, and the demand for experienced operators continues to grow. This segment highlights where that growth is happening and where talent is needed most.

Want to Share an Open Role?
Companies in the ecosystem can submit roles to be featured in upcoming editions.

Connect with Brandon:
https://www.linkedin.com/in/brandonomoregie/

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

Want to get the most out of ChatGPT?

ChatGPT is a superpower if you know how to use it correctly.

Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.

Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.

Download the free guide

🪫Data Center Capacity Is Disappearing💰Bay Area Startups Collectively Secured $136B MTD in Feb

Sat, 28 Feb 2026 17:06:04 +0000

We’re almost through February, and from our estimates at InfraCore, most of 2026 data center capacity will be leased by the end of this quarter. That gives the market less than 45 days to close on what’s left. If you’re an off-taker still looking for significant contiguous capacity, think 5 MW or more, the window is real, and it’s closing faster than most people think.

Once those leases are signed, the next phase begins: a mad dash to get hardware, infrastructure, and construction crews mobilized in time to hit Ready-for-Service dates. Q2 through Q4 of this year will see an incredible amount of execution pressure as everyone races to make 2026 commitments real.

The Problem on the Buyer Side: Site Overload

So why aren’t buyers moving faster? In many cases, they’re drowning. A customer needs 10, 20, maybe 100 MW. They reach out to a few brokers, those brokers tap their networks, and suddenly three brokers become 15 and five sites become 20. Each site arrives with different details, in different formats, through different channels like email, spreadsheets, text messages. The customer ends up building their own master spreadsheet, chasing follow-ups through a game of telephone, and discovering that the same site appears multiple times with conflicting information.

We’ve had customers sign an NDA, receive a detailed site deck, and show up two weeks later not even realizing this was done and they had the information. This is what causes the deer-in-headlights delay. The best customers I work with have figured it out: work with a small number of trusted partners who are ruthlessly selective about what they present. My job isn’t to hand you 12 sites and hope something sticks. It’s to understand your requirements so I can say: here are two or three that actually fit.

The Problem on the Provider Side: Two Trains in the Night

When buyer and provider finally meet, they each want completely different things. The customer wants technical validation. Is the power real? What does the floor plan look like? Can I talk to the construction team? The provider wants financial validation. Can you pay for this? Who’s the end user? Show me audited financials. Both are reasonable, and neither wants to go first. Without someone coordinating, this standoff quietly kills deals.

The solution is running both workstreams in parallel. Financial review and technical review happen simultaneously, each side taking a step toward the other, keeping the deal’s momentum alive.

What Winning Looks Like

The savviest customers we work with know their technical requirements cold before they start looking. When they see a site that fits, they move. We’ve seen companies go from NDA to site visit to redlining term sheets within a week. In that first meeting, they’re confirming power is real, reviewing floor plans, and asking for the MSA before the meeting ends. Then they kick off financial review, technical review, and legal redlining in parallel.

Providers are also shifting. Many no longer want traditional LOIs because they’re non-binding and take sites off the market without commitment. Instead, they push for term sheets that flow directly into the MSA. The message is simple: first one to sign wins.

At InfraCore, we’re laser-focused on moving our remaining 2026 capacity and building our 2027–2028 inventory. This AI boom rolled right from ’25 into ’26, and it’s not letting up. Have your checklist ready. Move fast. This stuff won’t wait.

AI INFRA SUMMIT x PLUG AND PLAY

We are thrilled to announce that the AI INFRA SUMMIT (AIS5) is officially coming to Plug and Play in Silicon Valley on May 1, 2026. This beautiful, dynamic venue is the perfect setting to host the most critical conversations of the year. As the AI supercycle reaches a massive inflection point, Plug and Play will serve as the ultimate convening ground for the builders creating the models, the buyers deploying capital, and the connectors building the physical stack.

As we transition from the speculative "Gold Rush" of GPU allocation to a true "Industrial Revolution" focused on building sustainable AI factories, the stakes have never been higher. This summit is designed to tackle what comes next, moving beyond hand-waving hype to focus on the ground-truth deployments and hard engineering required to break through the physical walls of power, liquid cooling, and interconnects. Plug and Play offers the ideal collaborative space to house these frontline intelligence discussions and tactical deep dives.

Whether you are an enterprise leader, an AI developer, or an investor shaping the physical and digital future, this is your front-row seat to influence the ecosystem. Join us to ignite strategic partnerships, share real-world playbooks, and command the conversation driving the next $2 trillion in AI infrastructure deployment

Join us for AI INFRA 5 on May 01. Grab your Super Early Bird tickets!

Use code: IGNITESGC1 for $100 off

Join our friends on March 3rd

Bay Area Startups Collectively Secured $163B MTD In February

February funding activity continued to set new records, closing the week at $111B and taking the month to $163.7B. That takes our 2026 Q1-to-date figure over $193B, already more than the 2025 full YEAR total. The largest funding this week - also the largest venture capital round ever – was the $110B round at a $840B valuation announced today by OpenAI. Just three investors made up the round, with Amazon putting in $50B, while Softbank and NVIDIA invested $30B each. $35B of Amazon's investment is based on milestones, including OpenAI attaining AGI and completing an IPO.

Follow LinkSV on LinkedIn to stay on top of SV funding intelligence and key players in the startup ecosystem.

Early Stage:

ZaiNar closed a $100M Series A, achieving sub-nanosecond time synchronization and distribution across wireless networks, transforming existing 5G and WiFi infrastructure into a comprehensive spatial sensing system.
Salma Health closed a $80M Series A, an end-to-end integrated brain health company specializing in mental health, neurological and neuropsychiatric conditions.
ChipAgents closed a $50M Series A, an Agentic AI platform that transforms specifications and code into production-ready RTL, verification assets, and automated root-cause analysis.
Rowspace closed a $50M Series A, connects structured and unstructured data across a firm's entire history, models how that firm operates and thinks, and delivers that intelligence.
Evoke Security closed a $4M Pre-Seed, the cybersecurity platform purpose-built to secure the agentic workforce.

Growth Stage:

OpenAI closed a $110B Series unknown, an AI research and deployment company dedicated to ensuring that general-purpose artificial intelligence benefits all of humanity.
World Labs closed a $1B Series C, building frontier world models that can perceive, generate, reason, and interact with the 3D world.
MatX closed a $500M Series B, designs hardware tailored for the world’s best AI models, dedicating every transistor to maximizing performance for large models.
VITURE closed a $100M Series B,develops AR technology and is the home of the #1 XR glasses brand in the US.
Ainekko closed a $14.7M Series B, building an open-source composable stack from inference to hardware.

AMD x Meta Expand AI Partnership

AMD and Meta just expanded their strategic partnership to support Meta’s next wave of AI infrastructure buildout.

This is not a one-off supply agreement. The companies are aligning roadmaps around AMD Instinct GPUs, rack-scale systems, and long-term AI deployment strategy.

For anyone tracking the AI infrastructure race, this is another signal that hyperscalers are actively diversifying GPU supply and locking in multi-year partnerships.

Read the full announcement here:
https://www.amd.com/en/newsroom/press-releases/2026-2-24-amd-and-meta-announce-expanded-strategic-partnersh.html

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

Stop Drowning In AI Information Overload

Your inbox is flooded with newsletters. Your feed is chaos. Somewhere in that noise are the insights that could transform your work—but who has time to find them?

The Deep View solves this. We read everything, analyze what matters, and deliver only the intelligence you need. No duplicate stories, no filler content, no wasted time. Just the essential AI developments that impact your industry, explained clearly and concisely.

Replace hours of scattered reading with five focused minutes. While others scramble to keep up, you'll stay ahead of developments that matter. 600,000+ professionals at top companies have already made this switch.

Join them today, for free.

🔥Heat Is Killing AI 💰Bay Area Startups Collectively Secured $52.5B MTD

Sat, 21 Feb 2026 16:00:00 +0000

AI infrastructure has hit a hard limit no software can evade: heat. Air cooling served the 5 to 10 kW rack era well enough. Those constraints no longer apply. Rack power demands have exploded, forcing thermal design to become a fundamental go/no-go factor for AI clusters in 2026.

The Density Explosion

Cloud workloads settled into 5 to 10 kW racks, where 15 to 20 kW marked high-end performance. AI workloads operate at entirely different scales. Production racks now average 40 to 50 kW, while leading GPU clusters routinely exceed 100 to 120 kW per rack. Roadmaps project even higher peaks.

A single training pod generates more heat across a few square meters than entire legacy enterprise halls once managed. Traditional hot aisle/cold aisle systems and CRAH units simply lack the capacity to shift that volume of energy. Operators can apply workarounds like rear-door heat exchangers or aggressive fan curves, but those measures cap out around 30 to 40 kW per rack. Beyond that threshold, diminishing returns turn into outright failure.

The Retrofit Crisis

The global data center inventory reflects yesterday's needs. While hyperscalers reserve high-density zones in newer facilities, most existing sites adhere to 5 to 15 kW design limits. Cooling tonnage represents only one barrier among several structural realities.

Floor loading presents the first insurmountable challenge. Modern racks, containment structures, and liquid distribution gear combine to exceed 3,000 pounds per unit. Legacy raised floors rate for 150 pounds per square foot and buckle under sustained high-density deployments.

Piping infrastructure compounds the issue. Liquid systems demand dedicated CDUs, manifolds, and runs to chillers or dry coolers, none of which exist in facilities plumbed for air-only operations. Retrofitting live environments triggers extended downtime, skyrocketing costs, and partial compromises at best.

Power delivery and redundancy complete the picture. Even perfect cooling proves useless against 15 to 20 kW feeds. Operators can isolate small high-density pods within older shells, but wholesale conversion from 2012-era enterprise data centers to 2026 AI campuses requires full demolition and rebuild. The internet's physical backbone must evolve or obsolesce.

Liquid Cooling as Operational Mandate

Liquid cooling once confined itself to HPC labs and experimental clusters, promising superior thermals but carrying a reputation for complexity. That perception has evaporated.

Direct-to-chip systems deliver coolant precisely to heat sources, extracting tens of kilowatts per server with far greater efficiency than air. Immersion setups submerge entire systems in dielectric fluid, enabling extreme densities through simplified mechanics and reduced airflow dependency. Liquid's heat capacity supports hotter components in compact footprints, while slashing fan power and stabilizing performance.

By 2026, liquid cooling defines the baseline for dense AI deployments. Roughly 80 percent of new facilities integrate it from initial design rather than as an afterthought. Air persists for lighter ancillary loads, but GPU-heavy zones demand liquid capacity upfront.

The Revised Facility Stack

This shift ripples through every layer of campus design. Shells now accommodate elevated floor loads, streamlined airflow paths, and embedded liquid corridors from the outset. Mechanical plants scale for elevated return temperatures, with loops optimized for CDU feeds rather than broad air distribution.

Server and rack architectures evolve in tandem, as OEMs standardize liquid-ready configurations for flagship AI hardware. Business-park air-cooled facilities fade from relevance. Market value concentrates in purpose-built campuses where thermal, power, and compute systems integrate as a unified whole.

Thermal Implications for Operators

Operators targeting meaningful AI scale face stark choices. New campuses must treat liquid cooling as foundational infrastructure. Partial retrofits consign most existing square footage to legacy status. Opting out forfeits access to the era's most lucrative workloads.

Liquid cooling stands as thermodynamics' verdict on the AI age. The industry confronts a thermal wall that demands passage on physics' terms alone.

Upcoming Events

Join us for AI INFRA 5 on May 01. Grab your Super Early Bird tickets!

Bay Area Startups Collectively Secured $52.5B MTD In February

February funding activity continued to set new records, closing the week at $1.02B and taking the month to $52.5B. That takes our 2026 Q1-to-date figure over $82B, already more than the full quarter total for Q1 2025. And the level of funding activity - particularly in megadeals - shows no sign of abating. OpenAI is said to be nearing their first close on a $100B round (which would set a new record) with Amazon, Softbank and NVIDIA in talks to invest up to $50B, $30B and $20B respectively. Microsoft is also in the mix, along with VC firms and sovereign wealth funds, making it possible the round will exceed OpenAI's target.

Follow us on LinkedIn to stay on top of SV funding intelligence and key players in the startup ecosystem.

Early Stage:

Cogent Security closed a $42M Series A, building autonomous agents that close the execution gap in cybersecurity.
Kana closed a $15M Seed, an agentic marketing platform powering brands to win in the agent economy.
ZaiNar closed a $10M Series A, foundation layer for Physical AI, transforming 5G and WiFi infrastructure into a comprehensive spatial sensing system.
Skygen.AI closed a $7M Seed, specializing in the creation of autonomous software agents.
Sift Biosciences closed a $3.7M Pre-Seed, developing a new class of T-cell modulators to treat cancers and autoimmune diseases by leveraging AI.

Growth Stage:

World Labs closed a $1B Series C, building frontier world models that can perceive, generate, reason, and interact with the 3D world.
Heron Power closed a $140M Series B, enables renewable energy, storage, and datacenter developers to directly connect to medium voltage transmission without using a transformer.
Render closed a $100M Series C, build and scale applications and websites with a global CDN, DDoS protection, preview environments, private networking, and Git auto deploys.
Braintrust AI closed a $80M Series B, the AI observability platform helping teams measure, evaluate, and improve AI in production.
ProSomnus Sleep Technologies closed a $38M Series Unknown, the leading non-CPAP OSA therapy® for the treatment of Obstructive Sleep Apnea.

Vultr is a global cloud infrastructure provider focused on simplicity, performance, and predictable pricing. The platform delivers fast compute, storage, and networking services designed to support startups, developers, and enterprises with flexible cloud resources across the world.

What Vultr Delivers
• Bare metal, virtual machines, and dedicated cloud compute options
• High performance block storage and object storage solutions
• Global data center footprint that enables low-latency deployment
• Consistent pricing without hidden fees for predictable infrastructure cost

Why It Matters
As applications scale and workloads become more demanding, teams need cloud infrastructure that is reliable, performant, and easy to deploy. Vultr helps companies of all sizes move faster with cloud instances that can be provisioned in minutes and scaled with demand.

Who It Serves
Developers, startups, SaaS businesses, and enterprises looking for foundational cloud infrastructure they can trust for production workloads, edge use cases, and global distribution.

Vultr brings simple, powerful cloud infrastructure to teams building the next generation of applications.

See more at https://www.vultr.com/

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

Ship the message as fast as you think

Founders spend too much time drafting the same kinds of messages. Wispr Flow turns spoken thinking into final-draft writing so you can record investor updates, product briefs, and run-of-the-mill status notes by voice. Use saved snippets for recurring intros, insert calendar links by voice, and keep comms consistent across the team. It preserves your tone, fixes punctuation, and formats lists so you send confident messages fast. Works on Mac, Windows, and iPhone. Try Wispr Flow for founders.

Start flowing free

🪴Structural Scarcity Is Taking Root💰Bay Area Startups Collectively Secured $50B MTD in Feb

Sat, 14 Feb 2026 16:00:18 +0000

The AI supply chain is in a phase of structural scarcity. This is not a temporary dip in supply; it is a multi‑year shortage in the components required to build AI factories, especially the advanced packaging that binds GPUs to high‑bandwidth memory and the HBM itself.

The constraint is simple: you cannot build an H100, MI300, or any comparable AI accelerator without advanced packaging and high bandwidth memory. Advanced packaging capacity at TSMC and high‑bandwidth memory output from vendors like SK Hynix are effectively sold out through 2026. The limiting factor is no longer demand for GPUs; it is access to the packaging and memory capacity that makes those GPUs real.

Fabrication is no longer the main bottleneck. Even if you secure leading‑edge wafers, those chips still need to move through scarce advanced packaging and be matched with high‑bandwidth memory that is already fully allocated. TSMC and the major memory houses are running flat out, yet demand for AI systems sits structurally above what the current capacity curve can support. That gap is persistent, not cyclical.

This is what structural scarcity looks like:

Capacity expansions are slow, capital‑intensive, and themselves constrained by upstream tools and processes.
The largest buyers have already pre‑booked the majority of incremental output via multi‑year agreements.
New entrants and slower movers are left to compete over whatever residual capacity remains at the margin.

GPUs Use More and More Memory

In this environment, the old playbook fails. Buyers cannot rely on spot markets, end‑of‑quarter deals, or opportunistic secondary channels. When output is pre‑sold, there is nothing to “spot” in the first place. Price becomes a weak lever because the real constraint is allocation, not willingness to pay.

The mandate for 2026 is capacity, not discount. Procurement teams need to behave like strategic financiers of their own AI infrastructure, not transactional buyers of commodity hardware. That means moving from short‑term purchasing to explicit capacity locking.

In practice, that shift looks like:

Multi‑year commitments for GPUs, memory, and packaging capacity with clear volume and roadmap visibility.
Direct relationships with foundries, memory suppliers, and integrators, backed by credible long‑term demand signals.
Willingness to trade some pricing flexibility for guaranteed allocation and delivery priority.

The 2026 environment is a pressure cooker for anyone building or scaling AI infrastructure. Organizations that approach this as a standard hardware procurement cycle will find themselves stuck in waitlists while competitors with locked‑in capacity continue shipping products, models, and features. The supply chain is now a competitive weapon.

This is not a story about “once supply catches up, things go back to normal.” AI demand, packaging complexity, and HBM intensity are moving faster than capacity can be added. The companies that win this phase are the ones that treat advanced packaging and high‑bandwidth memory as strategic inputs to their business and lock in access ahead of time.

In a structurally constrained world, the correct question is no longer “How do we get the best price on GPUs?” The correct question is “What are we willing to commit so that we are still building when everyone else is waiting?”

Upcoming Events

Join us for AI INFRA 5 on May 01. Grab your Super Early Bird tickets!

Bay Area Startups Collectively Secured $50.8B MTD In February

February funding activity continued to set new records in week two, closing the week at $31.6B and taking the month to $50.8B. Six megadeals – including Anthropic 's $30B Series G – provided 98% of the total. The other five megadeals went to Inertia, SambaNova, Solace , Loyal and Simile.

US Market valuations surged in 2025, with median pre-money valuations the highest in a decade, exceeding even 2021 levels, per Pitchbook. But not all valuations are equal, some are stale: Pitchbook estimates that almost 25% of startups valued as unicorns in earlier years would have a current market value of less than $1B. And we have a new handle -"undercorns" - for those companies coined by Axios.

Follow us on LinkedIn to stay on top of SV funding intelligence and key players in the startup ecosystem.

Early Stage:

Inertia Enterprises closed a $450M Series A, the commercial fusion energy company, using laser-based fusion.
VillageSQL closed a $35M Series A, a drop-in replacement for MySQL with extensions for the agentic AI era.
Trener Robotics closed a $32M Series A, redefining robotics by combining advanced AI with pre-trained skill models that are expert at specific tasks.
Simple AI closed a $14M Seed, builds voice AI agents for inbound and outbound B2C calls.
Smart Bricks closed a $5M Pre-Seed, a frontier AI lab building agentic AI infrastructure for global real-estate investing.

Growth Stage:

SambaNova Systems closed a $350M Series E, a complete solution purpose-built for AI and deep learning that overcomes the limitations of legacy technology.
Solace closed a $130M Series C, a digital platform that connects patients with expert healthcare advocates who navigate the healthcare system on their behalf.
Loyal closed a $100M Series C, an animal health company developing the first drugs intended to help dogs live longer, healthier lives.
Bretton AI closed a $75M Series B, the leading AI platform for financial crime operations.
Anthropic closed a $30B Series G, an AI safety and research company working to build reliable, interpretable, and steerable AI systems.

Micas Networks is a San Jose-based open networking company that builds high-performance networking solutions for hyperscale, cloud, and data center environments. The company develops a full portfolio of open network switches and infrastructure designed to support the demands of AI, cloud, and high-throughput applications.

What Micas Networks Delivers
• High-performance open networking switches from 1G up to 800G tailored for modern data centers.
• Solutions that support flexible network operating systems including SONiC, enabling customization and scale.
• Energy-efficient designs and options such as co-packaged optics to accelerate connectivity and reduce bottlenecks in AI clusters.
• A manufacturing base and R&D capability that supports rapid delivery and customization for enterprise requirements.

Why It Matters
As data centers scale to support AI training and inference, networking performance and efficiency become key determinants of overall system throughput. Micas Networks’ open networking approach gives builders flexibility and performance by combining robust hardware with adaptable software stacks.

Who It Serves
Cloud operators, hyperscalers, enterprises, and OEMs that require low-latency, high-bandwidth networking infrastructure to keep up with modern workloads and next-generation compute demands.

Learn more at micasnetworks.com.

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

Better prompts. Better AI output.

AI gets smarter when your input is complete. Wispr Flow helps you think out loud and capture full context by voice, then turns that speech into a clean, structured prompt you can paste into ChatGPT, Claude, or any assistant. No more chopping up thoughts into typed paragraphs. Preserve constraints, examples, edge cases, and tone by speaking them once. The result is faster iteration, more precise outputs, and less time re-prompting. Try Wispr Flow for AI or see a 30-second demo.

Start flowing free

⚔️The AI Land War is brewing💰Bay Area Startups Collectively Secured $19B

Sat, 07 Feb 2026 18:34:44 +0000

AI infrastructure is running into a hard physical limit that software, capital, and silicon cannot overcome: large‑scale, reliable electricity. In 2026, advantage in AI is increasingly defined in megawatts, not model parameters.

From land wars to power bottlenecks

For a decade, developers competed for prime sites in hubs like Northern Virginia. Today, those markets are constrained not by acreage but by power availability. Dominion Energy expects electricity demand in Virginia driven largely by data centers to nearly double from 17 GW in 2023 to 33 GW by 2048. A JLARC study warns that grid and transmission upgrades are already struggling to keep pace with planned data center capacity.

Interconnection delays have become the gating factor for scale. Across several U.S. regions, 100+ MW projects face four‑ to seven‑year timelines for needed transmission, substations, and approvals. AI hardware cycles, by contrast, turn over every 12–24 months. GPUs depreciate quickly, shells can be built in a year, but utility infrastructure moves at public‑sector speed creating unprecedented stranded capacity risk: powered‑off compute waiting on electrons.

“Dirt to Token” as the new stack

AI compute now operates as an industrial system built on a physical stack: land, generation, storage, transmission, and only then compute. Loudoun County “Data Center Alley” is the leading indicator: Dominion began constraining power allocations in 2021, and county strategy documents explicitly flag power as the central risk to future growth.

Developers are responding with vertical integration “dirt to token” strategies that secure:

Land adjacent to high‑capacity transmission and fuel infrastructure.
Long‑term power, including on‑site or directly connected generation.
Storage and microgrids to stabilize 24/7 AI workloads.

In this model, power is no longer a pass‑through utility. It is the primary input that sets the marginal cost of intelligence.

Nuclear as baseload for AI

The duty cycle of AI workloads is pushing hyperscalers toward nuclear as a foundational energy source. Global data center electricity use is projected to grow from roughly 460 TWh in 2024 to more than 1,000 TWh by 2030, and to around 1,300 TWh by 2035 nearly tripling in a decade. Training clusters and high‑volume inference fleets require round‑the‑clock, firm power that solar and wind alone cannot guarantee.

In response, large technology companies are signing nuclear‑linked deals at gigawatt scale. AWS entered a 10‑year agreement with Talen Energy for a 960 MW campus adjacent to the Susquehanna nuclear plant in Pennsylvania. Microsoft has structured contracts with Constellation Energy tied to the restart and continued operation of nuclear assets to support 24/7 cloud workloads. Industry estimates suggest hyperscalers have now contracted more than 10 GW of nuclear capacity, including early‑stage small modular reactor pathways.

These 15–20‑year contracts effectively align nuclear baseload with the depreciation cycle of AI infrastructure, turning fission into the anchor of the AI load curve.

Behind‑the‑meter power and microgrids

Nuclear is a long‑cycle solution. To move faster, developers are increasingly building behind‑the‑meter generation for 500 MW‑plus campuses. Roughly one‑third of proposed U.S. data center capacity around 48 GW now plans to use on‑site or directly connected plants to partially “skip the grid,” up from under 2 GW in late 2024.

New campuses are being designed around dedicated gas‑fired plants approaching 1 GW, large‑scale batteries, and hybrid microgrids that blend gas, solar, wind, and storage. This collapses the distinction between data center developer and independent power producer: capital plans now include substations, grid studies, fuel logistics, and multi‑gigawatt generation alongside racks and cooling.

Power as competitive moat

Site selection criteria are shifting from “fiber and tax incentives” to energy system fundamentals. In Virginia alone, meeting projected data‑center‑driven demand could require 57–60 GW of new capacity by 2040 roughly three times Dominion’s 2023 generation fleet. Leading operators now prioritize:

Proximity to robust transmission and substations.
Access to gas pipelines or nuclear capacity for firm supply.
Permissive regimes for on‑site generation and innovative tariffs.

For large‑scale AI, the cost curve is migrating from silicon to electrons. Developers that control their own generation timelines and pricing gain structural advantage over peers locked into congested grids and utility queues. The land war persists, but in 2026, “dirt to token” energy strategy has become the bottleneck, the cost curve, and the enduring competitive moat.

Upcoming Events

Join us for AI INFRA 5 on May 01. Grab your Super Early Bird tickets!

Bay Area Startups Collectively Secured $19B in Feb Week1

February opened to a blazing fast start in funding activity, with deals that totaled $19B, already setting a new record for the month. Ten megadeals represented 95% of the total, with Waymo's $16B Series D at the top of the list, followed by Cerebras Systems $1B Series H, Bedrock Robotics, $270M Series B and seven more.

Exits: The eight acquisitions this week were all for undisclosed amounts, while two IPOs raised a total of $579M. A third IPO, Liftoff Mobile, was postponed due to market volatility. Of the two IPOs, Once Upon a Farm opened and rose slightly; Eikon Therapeutics opened below its offer price and fell further. Eikon's IPO was a “down round” IPO, with a valuation less than a third of the $3.5B valuation from the company's Series D funding in early 2025.

Follow us on LinkedIn to stay on top of SV funding intelligence and key players in the startup ecosystem.

Early Stage:

Fundamental Technologies closed a $225M Series A, developing Large Tabular Models purpose-built for the structured records that contain trillions of dollars in business value.
Resolve AI closed a $125M Series A, AI for running and operating software in production.
Gruve closed a $50M Series A, helps enterprises, neoclouds and AI startups build and operate infrastructure-to-agent AI solutions.
Powerline inc. closed a $7M Seed, develops AI systems for monitoring and operation of energy assets.
Airrived closed a $6.1M Seed, the Agentic OS, redefining how enterprises run cybersecurity, IT, and business operations in the agentic era.

Growth Stage:

Cerebras Systems closed a $1B Series H, building a new class of AI supercomputer.
Bedrock Robotics closed a $270M Series B, building the future of autonomous construction.
Goodfire closed a $150M Series B, building the next generation of safe and powerful AI—not by scaling alone, but by understanding the intelligence we're building.
Lunar Energy closed a $102M Series D, develops intelligent software and advanced hardware to electrify homes and connect communities to form clean, resilient virtual power plants.
Midi Health closed a $100M Series D, provides accessible, insurance-covered services in all 50 states – designed by world-class medical experts.

AMD

AMD is expanding its footprint in AI and high-performance computing with a broad portfolio of CPUs, accelerators, and system platforms that support workload scale from cloud and data centers to edge and client devices.

Recent announcements introduced new rack-scale platforms built on EPYC server CPUs and Instinct GPU accelerators, along with next-generation tools aimed at enterprise and hyperscale AI deployments. Management has also shared plans for continued investment across CPUs, GPUs, NPUs, and networking infrastructure to support what it calls the next phase of global compute demand. AMD’s data center business delivered strong growth, driven by broader adoption of EPYC and Instinct products as AI workloads scale, and executives have projected robust future increases in data center revenue as new silicon and systems reach production.

Why It Matters
Companies building AI infrastructure need options beyond traditional architectures. AMD’s strategy positions it as a diversified compute provider across servers, accelerators, and client-side AI hardware while maintaining momentum in data center sales and partnerships that support large-scale deployments.

Who It Serves
Cloud providers, enterprise data centers, AI developers, OEM partners, and organizations seeking a range of performance-optimized compute solutions.

Learn more: amd.com

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

👷‍♂️The AI Gold Rush Is Over, The Build Phase Begins💰Bay Area Startups Collectively Secured $2.9B

Sat, 31 Jan 2026 20:08:26 +0000

The Gold Rush is Over. The Industrial Revolution Has Begun.

If 2025 was defined by the frantic race for GPU allocation, a digital "Gold Rush" then 2026 marks the beginning of something far more capital-intensive and physically demanding: the Industrial Revolution of Intelligence.

While headlines debate the existence of an AI bubble, the underlying numbers tell a different story. We are witnessing the start of a $7 trillion infrastructure supercycle, driven not by speculation, but by the hard engineering reality of building the "AI Factories" required to power the next decade of economic growth.

Here is the state of the union for AI infrastructure as we enter 2026.

The $527 Billion Reality Check

The defining metric for 2026 is no longer just chip supply; it is Capital Expenditure (CapEx). Hyperscalers are projected to pour more than $527 billion into infrastructure this year alone.

Skeptics call this a bubble. Operators call it "vendor financing". We are in the deployment phase of a new general-purpose technology, comparable to the build-out of the electrical grid or the railway system. The massive upfront investment is a rational play to lower the cost of producing intelligence to the point where it becomes ubiquitous. The question is no longer if the money will be spent, but who can deploy it effectively before the window closes.

The Shift to "AI Factories"

We may need to stop calling them data centers. The facilities being capitalized in 2026 are not designed for passive storage; they are "AI Factories" designed to generate intelligence as a tangible economic output.

This shift changes the physics of the building. We have moved from the cloud era of 10kW racks to the AI era of 45kW to 100kW+ densities. This is a rebuild. The industry is effectively constructing a new physical shell for the internet, optimized for massive parallel compute, liquid cooling, and 1.6T networking fabrics.

The "Inventory" Crisis: Capital is Cheap, Capacity is Scarce

The most critical bottleneck in 2026 is execution. The primary constraint facing the market is no longer demand, but the ability to convert planned capacity into live infrastructure.

We are facing an "inventory crisis." While trillions of dollars are ready to be deployed, there is a structural shortage of finished capacity, power, cooling, and integrated systems. Paper pipelines are meaningless if you cannot plug them in. The "land war" has become a power war, and the winners of this cycle will be those who can solve the physical constraints of the grid and supply chain to deliver usable tokens to the market.

The Mandate to Convene

This transition from allocation to execution requires a new level of collaboration between the builders (infrastructure), the buyers (capital), and the connectors (supply chain).

This is why we are convening at the AI INFRA SUMMIT on May 1. We are moving past the hype to the "hard engineering" required to build the capacity for intelligence.

Upcoming Events

Join us for AI INFRA 5 on May 01. Grab your Super Early Bird tickets!

Bay Area Startups Collectively Secured $2.90B+ in January Week 4

This week's deals collected $2.90B for Silicon Valley startups, taking the January total to $29.95B and setting another record for January, more than $9B over the previous mark. This week there were ten megadeals – 80% of the total – with Genspark AI's $300M Series B at the top of the list. Seven of the ten megadeals were pureplay (AI infrastructure) or applied AI.

Exits: Check out January Week 4 Acquisitions. Seven SV companies acquired other companies, four SV companies were acquired. Sources gave Apple's acquisition of Q.ai a $2B pricetag. On the IPO front, Ethos Technologies completed theirs, raising $200M at a fully diluted valuation of $1.3B. It was a “down” IPO, their previous high valuation was $2.7B in their 2022 Series D-1 funding.

Webinar Alert: The February edition of our LinkSV-WITI Tech, Talent & Investment Trends webinar is next Friday, February 6. AI still rules, while valuations for frontier model startups are in the stratosphere - and one of them is talking up a 2026 trillion dollar IPO. We'll look at 2026 exits, liquidity and fundraising and how AI and non-AI startups are likely to be affected. Plus, a look at the AI bubbles - yes, there are multiple - and when they may burst - deflate - come down to earth. Join us at noon PT on February 6 to get the full picture. You can register here.

Follow us on LinkedIn to stay on top of SV funding intelligence and key players in the startup ecosystem.

Early Stage:

Ricursive Intelligence closed a $300M Series A, a frontier AI lab building the compute foundation for the next generation of AI.
Flapping Airplanes closed a $180M Seed, a foundational AI research lab solving the data efficiency problem.
Grid Aero closed a $20M Series A, building the future of autonomous cargo logistics.
Emobi closed a $3.4M Seed, enables seamless and secure EV charging across networks.

Growth Stage:

Genspark AI closed a $300M Series B, building agentic AI for more than one billion global knowledge workers.
Decagon closed a $250M Series C, conversational AI platform empowering every brand to deliver an AI concierge for every customer.
Span.io closed a $188M Series C, enabling electrification for all.
Prime Intellect closed a $49.9M Series B, makes frontier AI training accessible to every company.

AI factories run on GPUs. GPUs run on data. WEKA sits in the middle of that equation.

WEKA builds high performance data infrastructure designed to keep GPUs fed, saturated, and productive. As model sizes grow and clusters scale into the thousands of accelerators, storage performance becomes a gating factor for training speed, inference cost, and cluster ROI.

The WEKA Data Platform is built for AI workloads from day one. It delivers low latency, high throughput, and scalable performance across on-prem, cloud, and hybrid environments. Teams use WEKA to remove storage bottlenecks, reduce idle GPU time, and accelerate data pipelines across training and inference.

WEKA is deployed by leading AI labs, enterprises, and research institutions, and works closely with NVIDIA to support next-generation GPU architectures and large scale AI clusters.

Why It Matters
Most AI infrastructure discussions focus on chips and power. Storage and data movement quietly decide whether clusters run at 90 percent utilization or 30 percent. WEKA targets that invisible bottleneck and turns storage into a performance lever rather than a constraint.

Learn more at weka.io.

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

What Will Your Retirement Look Like?

Planning for retirement raises many questions. Have you considered how much it will cost, and how you’ll generate the income you’ll need to pay for it? For many, these questions can feel overwhelming, but answering them is a crucial step forward for a comfortable future.

Start by understanding your goals, estimating your expenses and identifying potential income streams. The Definitive Guide to Retirement Income can help you navigate these essential questions. If you have $1,000,000 or more saved for retirement, download your free guide today to learn how to build a clear and effective retirement income plan. Discover ways to align your portfolio with your long-term goals, so you can reach the future you deserve.

Get The Guide

🤖 AI Workloads are Being Rewritten 💰Bay Area Startups Collectively Secured $2.64B 🏝️ PTC'26 Recap

Sun, 25 Jan 2026 20:22:17 +0000

AI Workloads Are Rewriting Hyperscale Strategy

Three Critical Shifts Defining AI Infrastructure in 2026

The first Monday of 2026 made something clear: several major deals landed, and industry conversations pointed to an AI and data center boom that is accelerating. As 2026 shapes up to be a year defined by capacity constraints, three shifts are reshaping how AI infrastructure is built and delivered. Companies that recognize and adapt to these trends will be positioned to thrive.

The Rise of Inference Data Centers

The training era established a clear playbook: massive clusters in large facilities with abundant power and straightforward connectivity. Location didn’t matter much. What mattered was scale and uptime. If a power outage interrupted weeks or months of model training, you’d start from scratch.

Inference changes the equation. The currency is no longer just power capacity; it’s location, connectivity, and low latency. Inference workloads need to be near customers, near major metros, near the users actually running these AI applications. Clusters can be smaller. Uptime requirements are less critical than network performance.

This shift means we’ll see a proliferation of smaller, strategically located data centers serving inference workloads. The real estate strategy shifts. Infrastructure requirements follow, and the competitive advantage moves to those who can deliver capacity in the right locations with the right connectivity profiles.

Behind-the-Meter Power Goes Mainstream

Last year, behind-the-meter power generation was a stumbling block for many customers. Questions, hesitation, and uncertainty dominated the conversation. This year, that resistance is evaporating.

The outcome is straightforward: behind-the-meter solutions deliver capacity faster. For natural gas infrastructure in the United States specifically, reliability rivals or exceeds grid power. As the grid does its best to handle unprecedented AI power demands, customers are recognizing that behind-the-meter isn’t a compromise; it’s often the superior path to market.

We’re starting to see deals where behind-the-meter generation doesn’t even raise an eyebrow. The technology has proven itself, the infrastructure is reliable, and, most critically, it solves the timing problem that’s killing so many 2026 deployment plans.

The Evolved Broker Model

The traditional broker model of making introductions, passing requirements back and forth, and letting the operator handle fit-out and delivery is dead in 2026.

Brokers must get into the weeds with operators to help guarantee ready-for-service dates. This means understanding the entire supply chain: switchgear, generators, busways, PDUs, cooling infrastructure, rack and stack logistics. Any missing component can slip an RFS date, and in today’s market, those delays are expensive.

Even when acting primarily as a connector, brokers need relationships with equipment suppliers. They need to understand lead times, maintain backup options, and be prepared to help source whatever’s required to ensure delivery. The broker who can only make introductions will lose to the broker who can help guarantee outcomes.

At Infracore, we’ve been preparing for these shifts since last year. We’ve built short-RFS solutions for AI inference deployments, embraced behind-the-meter power generation, and developed an extensive partner network that can supplement any aspect of data center development to ensure RFS commitments are met.

The capacity crisis of 2026 will separate those who adapt from those who don’t. Understanding these three shifts isn’t optional; it’s foundational to closing business this year.

PTC 2026 // Honolulu, HI

Three Key Takeaways from PTC 2026:

Why This Conference Has Become Essential for Data Center Professionals: I just returned from my first Pacific Telecommunications Conference (PTC) in Honolulu, and I have to say this event has earned its reputation as one of the premier gatherings for the data center industry. While the name might suggest a focus purely on telecommunications, PTC has evolved into the go-to conference for everything data center-related, driven largely by the AI boom and explosive growth in our sector. This evolution makes complete sense when you understand the history.

Data centers emerged in locations where telecommunications infrastructure was first established, primarily in major financial centers. Cities like Chicago, New York, Tokyo, Hong Kong, London, and Northern Virginia became tier one data center markets because they already had the telecom interchanges necessary for financial transactions and global connectivity. A conference designed to bridge Asia-Pacific with the Americas, held at the beginning of each year in Honolulu, naturally became a critical meeting point for the global data center community.

Here are my three major takeaways:

The Location is Unbeatable Honolulu in mid-January offers perfect weather (t-shirt, shorts, and flip-flops throughout). The resort setting provided countless open-air venues for meetings and casual networking. Unlike many conferences where you're scrambling for meeting space or fighting over table reservations, PTC offered abundant places to connect with people organically. I noticed some companies reserved hotel suites or bar tables for the duration of the conference. While this provides a dedicated meeting space, it can also create isolation. Some of the best networking happened through spontaneous encounters when I was walking around, running into familiar faces, and meeting new contacts naturally. That organic networking is harder to replicate when you're stationed in a hotel room waiting for people to navigate elevators and find you.
Everyone You Need to Meet is There The breadth of attendees at PTC is remarkable. I connected with investors, developers, power equipment providers (critical given current supply chain challenges), cooling innovators, construction professionals, and off-takers ranging from kilowatt to gigawatt scale operations. I discovered a niche data center provider in New York City I hadn't known about and met an up-and-coming cooling company doing innovative work in liquid cooling. The conference serves as a genuine one-stop shop for the entire data center ecosystem with site selectors, hardware providers, operators, and everything in between. What I particularly valued were the deepening relationships with contacts I've been building over the past couple years. Seeing people for the third or fourth time in person, or finally meeting face-to-face after a year of phone conversations, creates momentum. These established relationships led to valuable introductions at parties and happy hours, the kind of organic business development that only happens when everyone's in the same place.
The Timing is Perfect Holding this conference in mid-January provides the ideal kickoff for the year. It sets the tone and gets deals moving early. For next year, I'm planning to extend my trip in both directions. I'll arrive a couple days early to schedule strategic meetings with more time allocated, avoiding the back-to-back rush that inevitably leads to domino-effect delays. I'm also planning to bring my family and stay several days after the conference to enjoy Honolulu together. PTC has definitely become an annual commitment for me. If you're serious about the data center industry, I highly recommend adding it to your calendar.

Upcoming Events

Join us for AI INFRA 5 on May 01. Grab your Super Early Bird tickets!

Bay Area Startups Collectively Secured $2.34B+ in January Week

This week's deals collected $2.34B for Silicon Valley startups, taking the January total to $27.05B and setting another record for the month of January, more than $7B over the previous record. There were seven megadeals – 82% of the total – across all funding stages, with a late stage $600M Series G for Zipline and at the other end, an early stage $480M seed round for Humans&. Six of the seven megadeals were pureplay (AI infrastructure) or applied AI.

Exits: The eight acquisitions today brought the week's total to thirteen. Ten SV companies acquired out-of-the-area companies for undisclosed amounts, while three SV companies were acquired, two for disclosed prices: GSK acquired Rapt Therapeutics (Nasdaq:RAPT) for $2.2B and Capital One acquired Brex for $5.15B. The Brex acquisition is a great return for their early investors, but not so much for the 2022 Series D-2 investors, when the company was valued at $12.3B.

For startups raising capital: Stay on top of who's raising, who's closing and who's investing with the Pulse of the Valley weekday newsletter. Click through to get more detail on investors and executives, including email addresses for both. Founders get the newsletter, database and alerts for just $7/month ($50 value). Check it out and sign up here.

Follow us on LinkedIn to stay on top of SV funding intelligence and key players in the startup ecosystem.

Early Stage:

Upscale AI closed a $200M Series A, a high-performance AI networking company.
Inferact closed a $150M Seed, an open source solution for lowering the cost of inference.
Haiqu closed a $11M Seed, a quantum software company developing a new application execution stack for all modalities of near-term quantum computers.
AiStrike closed a $7M Seed, the intelligence fabric that unifies detection, investigation, and response into a continuously self-improving SOC.
Blockit closed a $5M Seed, the AI scheduler for people who are back-to-back.

Growth Stage:

Zipline closed a $600M Series G, the world's largest and most experienced autonomous delivery service.
BaseTen closed a $300M Series D, providing the industry’s most advanced inference stack.
Railway closed a $100M Series B, a cloud platform that empowers developers to ship faster and more efficiently.
LiveKit closed a $100M Series C, helps you build realtime AI, and instantly transport audio and video between LLMs and your users.

We first covered Articul8 in this newsletter back in 2024, nearly two years to the day, when the company spun out of Intel to focus on enterprise generative AI.

Valley Recap - Jan 6th 2024 edition

That story has progressed.

Articul8 is currently raising a $70M Series B and has already secured more than half of the round at a reported $500M valuation. The raise comes as enterprises move from pilots to production deployments of generative AI, particularly in regulated environments.

What Stands Out
Articul8 has stayed focused on building generative AI systems that run within customer-controlled environments, with an emphasis on data governance, traceability, and operational control. That positioning has translated into growing enterprise adoption and larger contract sizes since the spin-off.

Why This Matters Now
Enterprise AI buyers are becoming more selective. Platforms built for internal deployment and oversight are seeing increased demand as organizations formalize AI usage beyond experimentation.

This round marks a clear step forward from the company’s initial spin-out phase into a scale phase.

Learn more at Articul8.

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

Introducing the first AI-native CRM

Connect your email, and you’ll instantly get a CRM with enriched customer insights and a platform that grows with your business.

With AI at the core, Attio lets you:

Prospect and route leads with research agents
Get real-time insights during customer calls
Build powerful automations for your complex workflows

Join industry leaders like Granola, Taskrabbit, Flatfile and more.

👉 Try Attio Pro for free

✍🏼 AI Workloads Are Rewriting Hyperscale Strategy 🔭 Bay Area Startups Collectively Secured $2.64B

Sat, 17 Jan 2026 16:11:38 +0000

AI Workloads Are Rewriting Hyperscale Strategy

The past two years of AI infrastructure build-out have been driven by urgency. Hyperscalers rushed to secure GPUs, expand training clusters, and stand up new capacity wherever power and real estate were available. McKinsey’s latest analysis on AI workloads and hyperscaler strategy shows that this first phase is ending.

What comes next is more deliberate. AI is no longer being treated as a special project inside cloud organizations. It is becoming the workload that defines how infrastructure is planned, financed, and deployed.

This shift has clear implications for where data centers are built, how they are designed, and what constraints now matter most.

AI is now the main driver of data-center demand

For most of the last decade, data-center growth followed broad cloud adoption. SaaS, streaming, and enterprise IT migrations set the pace. That model is changing.

McKinsey estimates that AI workloads will push total data-center power demand close to three times today’s level by the end of the decade. Hyperscalers are expected to control most of that capacity, either through ownership or long-term leases that function like owned infrastructure.

This concentrates both risk and opportunity. Capital planning, site selection, and supply-chain strategy are now being shaped by a narrower set of use cases with far higher technical requirements than traditional cloud workloads.

Inference overtakes training

Training built the first generation of AI infrastructure. Inference will account for most of the growth going forward.

Inference workloads will surpass training workloads before 2030. This reflects how AI is being deployed across products and services. Models are no longer used primarily in scheduled training cycles. They are embedded in search, support, recommendations, and internal tools that run continuously.

The infrastructure consequences are straightforward. Training favors large, centralized campuses with high power density and specialized cooling. Inference favors proximity. Latency, network access, and reliability matter more than maximum rack density.

Hyperscalers are responding by building two different physical systems at the same time. Large remote campuses optimized for training and smaller, distributed sites designed for inference. The cloud may still look centralized from a software perspective, but physically it is becoming more distributed again.

Reliability becomes a business issue

As AI moves into revenue-generating workflows, the cost of downtime changes.

Outages that once affected internal tools or background services now affect customer-facing systems. McKinsey highlights a growing shift toward higher availability standards for inference environments, including full redundancy across power, cooling, and network paths. When AI systems handle customer interactions, transactions, or operations, infrastructure resilience becomes a commercial requirement, not just a technical one.

Two infrastructure networks, not one

A central theme in the report is the emergence of a dual approach to hyperscale infrastructure. On one side are very large training campuses built in regions with access to power and land. On the other are inference-focused facilities closer to users, data, and network hubs.

This challenges the assumption that hyperscale always means centralization. The next phase of hyperscale is split by function. Centralized where scale matters, distributed where response time matters. For colocation providers, interconnection platforms, and metro data-center operators, this is not a marginal change. It expands the role of regional infrastructure rather than reducing it.

Power becomes the binding constraint

If there is one factor that now shapes hyperscaler strategy more than any other, it is access to electricity. McKinsey is explicit that grid congestion, long interconnection queues, and permitting delays are now among the biggest limits on AI expansion. In established markets, these constraints are often harder to solve than real-estate or capital challenges.

Hyperscalers are responding in two ways. First, by moving into secondary markets where power can be delivered faster. Second, by becoming more involved upstream in energy strategy, including long-term power agreements, on-site generation, and storage.

Energy is no longer treated as a utility input. It is being treated as a strategic resource that shapes competitive positioning.

Faster builds, shorter planning cycles

To meet demand, hyperscalers are changing how they deploy infrastructure.

McKinsey points to a shift toward leasing, modular construction, prefabrication, and phased campus development. The goal is speed, not architectural perfection. Facilities are increasingly designed to be upgraded over time rather than built once to a final state.

Existing data centers are also being pushed harder. Liquid cooling retrofits, higher rack densities, and upgraded power systems are extending the life of sites that were never designed for AI workloads.

Infrastructure planning is starting to resemble product development. Launch quickly, improve continuously.

Hyperscalers move closer to utilities

One of the longer-term implications in the report is that hyperscalers are taking on a role that looks increasingly like infrastructure providers rather than just cloud platforms.

They influence where power is generated, how grids are expanded, and how large-scale infrastructure is financed. In some regions, hyperscalers now act as anchor customers around which entire energy and data-center ecosystems are built.

This does not make them utilities in a regulatory sense. But it does place them in a position of structural influence that extends well beyond software and services.

What this means for the Valley

AI is no longer an overlay on cloud infrastructure. It is reshaping the foundation. Inference is becoming the dominant workload. Reliability is becoming a business requirement. Energy access is becoming the main constraint. Hyperscalers are redesigning how they build, where they build, and what they control.

The next phase of AI will not be decided only by model performance or chip roadmaps. It will be decided by who can secure power, deploy infrastructure quickly, and operate it with the reliability that AI-driven services now demand.

That is the real shift behind the charts.

NRF Networking Party // New York City // Jan11

Upcoming Events

Join us for AI INFRA 5 on May 01. Grab your Super Early Bird tickets!

Early Bird Pricing

Bay Area Startups Collectively Secured $2.64B+ in January Week 2

This week's deals collected $2.64B for Silicon Valley startups, taking the total for the month to $24.71B. That's a new record for total funding in the month of January, more than $5B over January 2022. Eight of those fundings were megadeals that made up 74% of the total funding, including $500M Series B to Etched , $400M Series to Clickhouse (just a few months after their Series C) and $252M seed round to Merge Labs.

Founder Alert: Part 10 of Legal Foundations, from Seed to Sale, covering legal essentials for startups, is next Wednesday, Jan. 21 at noon PST with Roger Royse and Bob Karr. Part 10 covers mergers and acquisitions for startups, including deal process, structure, negotiation, compliance, tax and more. You can register here.

Follow us on LinkedIn to stay on top of SV funding intelligence and key players in the startup ecosystem.

Early Stage:

Merge Labs closed a $252M Seed, developing new approaches to brain-computer interfaces that interact with the brain at high bandwidth and integrate with advanced AI.
Higgsfield AI closed a $80M Series A, an AI-native generative video platform built for creators, brands, agencies, and marketing teams producing high-fidelity videos at scale.
depthfirst closed a $40M Series A, an applied AI lab developing novel security solutions for businesses facing modern, AI-era threats to their systems.
Transition Metal Solutions closed a $6M Seed, develops custom chemical additives that enhance wild microbial communities for metal extraction.
ModeInspect closed a $3.4M Seed, empowers large design teams and their core workflows from prototyping, design QA to designing full fledged features with AI.

Growth Stage:

ClickHouse closed a $400M Series D, a fast, open-source columnar database management system built for real-time data processing and analytics at scale.
Alpaca closed a $150M Series D, a self-clearing broker-dealer and a global leader in brokerage infrastructure APIs providing access to stocks, ETFs, options, and fixed income.
DeepGram closed a $130M Series C, the real-time API platform underpinning the Voice AI economy.
Harmonic Math closed a $120M Series C, developing Mathematical Superintelligence (MSI), which is rooted in mathematics and which guarantees accuracy and eliminates hallucinations.
WitnessAI closed a $58M Series B, the unified AI security and governance platform enterprises trust to govern and protect all AI activity.

HirePilot is an AI-powered recruiting CRM and ATS built for recruiters, consultants, and hiring partners who need a better way to manage sourcing, outreach, and hiring pipelines. The platform brings structure and automation to every stage of the recruiting and consulting workflow from prospecting and engagement to pipeline management and placement.

What HirePilot Delivers

AI-assisted sourcing and outreach to identify, contact, and engage candidates and prospects Recruiting CRM + ATS to manage leads, candidates, roles, and client pipelines in one system Automated workflows and follow-ups that reduce manual work and keep deals and searches moving Collaborative tools for teams and consultants to track conversations, stages, and decisions Custom dashboards and analytics to monitor activity, performance, and pipeline health in real time

Why It Matters

Recruiting and consulting workflows are often spread across inboxes, spreadsheets, LinkedIn, and disconnected tools. HirePilot replaces that fragmentation with a centralized, AI-driven system that helps recruiters and consultants operate faster, stay organized, and scale their impact without added complexity.

Who It Serves

Recruiters, staffing agencies, independent consultants, and in-house hiring teams who want a smarter, more efficient way to manage sourcing, outreach, and hiring pipelines all from a single platform.

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

Hiring in 8 countries shouldn't require 8 different processes

This guide from Deel breaks down how to build one global hiring system. You’ll learn about assessment frameworks that scale, how to do headcount planning across regions, and even intake processes that work everywhere. As HR pros know, hiring in one country is hard enough. So let this free global hiring guide give you the tools you need to avoid global hiring headaches.

Download the free guide today

Why proximity, hierarchy, and caching define the next era of AI infrastructure🔭 Bay Area Startups Collectively Secured $25B+ 💰

Sat, 10 Jan 2026 16:00:17 +0000

Serving Intelligence: From Training Clusters to Inference Grids

Why proximity, hierarchy, and caching define the next era of AI infrastructure
By Jay Chiang
Nov 16, 2025

Jay Chiang, Series: AI Infrastructure Research №3 (2025)

Image created with AI — Inference as a Grid

The Architecture Was Built for Training

AI infrastructure today reflects the training era’s requirements. Massive centralized clusters, synchronized execution across thousands of GPUs, and relentless optimization for utilization and throughput. This architecture succeeded brilliantly. Training frontier models (newest, largest, most advanced models) demands coordinating 10,000+ GPUs in lockstep, maintaining high-bandwidth GPU-to-GPU communication, and maximizing every hour of expensive compute time. So we build where power is cheap. We keep GPUs busy. And we batch everything.

Inference operates under fundamentally different constraints.

Millions of users arrive asynchronously, each demanding sub-200ms responses. Data sovereignty regulations prohibit cross-border queries. Latency requirements make geographic positioning non-negotiable. Responsiveness trumps utilization. A GPU sitting idle but ready for burst traffic serves users better than a fully utilized GPU queuing requests.

This isn’t just a workload difference. It’s the same distinction the electrical industry faced a century ago.

Power generation happens in massive centralized plants optimized for efficiency and scale. But power distribution requires a hierarchical network of transmission lines, regional substations, and local transformers, all positioned near demand, not near cheap fuel.

Our previous article showed that training data centers cannot simply be repurposed for inference deployment. This article explains why that architectural mismatch is fundamental — and what infrastructure profile actually succeeds.

The Hidden Layer: What the Data Center Metrics Missed

Common data center metrics coined at training era focuse on visible ones, such as GPU utilization, PUE (Power Utilization Efficiency), tokens per second. According to Uptime Institute’s Global Data Center Survey 2024, median hyperscale PUE has plateaued between 1.15 and 1.25, meaning even aggressive facility optimization only yields 5–20% additional efficiency headroom. However, the bigger saving opportunities come from reducing data movement and redundant computation, a layer almost no one measures. A few milliseconds of network latency may feel irrelevant in one run. When they compounds across millions of queries, the unnecessary network hops or remote data accesses multiply waste thousands of times daily.

Take a modest enterprise workload of one million inference requests per day. The model computation might consume roughly 3,000–6,000 kWh, but the data plane behind each request (embeddings, KV-cache reads, retrieval calls, and cross-region hops) often burns just as much energy. Move that workload from a distant cloud region to a local edge site and the physics change immediately. Round-trip distance drops, retries disappear, cache hit rates rise, and cross-region traffic collapses. In practice, this cuts data movement by about half and trims total energy by 30–50%. Add avoided egress fees, easily $150–$650 per month at this scale, and the economics compound. Proximity doesn’t just reduce latency. It removes the single biggest hidden cost of inference.

The Grid Model: Learning from Electricity Distribution

If you want to understand why inference architecture must change, look not to computing but to the power grid.

Electricity generation is centralized. Massive plants are built where energy is cheapest: Hydro valleys, desert solar farms, gas fields. But electricity consumption is distributed: homes, hospitals, factories, intersections, and devices scattered across the map. The only way to bridge those two realities is a hierarchy — high-voltage transmission, regional substations, and neighborhood transformers — each placed progressively closer to demand.

That separation isn’t aesthetic. It’s physics. Move electricity too far and you incur line losses. Deliver it too slowly and voltage sags. Push too much through the wrong tier and the grid collapses. The grid works because generation scales with efficiency, but distribution scales with proximity.

Inference follows the same laws. Training can be centralized. Inference cannot.

We don’t run high-voltage transmission directly into every home and rebuild the full electrical load at each outlet. It would be wasteful, fragile, and physically unnecessary. In the same way, sending every inference request back to a central model and recomputing retrieval, reason, and context from scratch burns energy and blows through latency budgets. Inference needs tiers, not round-trips.

Just as electrical grids aren’t optional, inference distribution won’t be, either. In many industries, regulation mandates local infrastructure. Healthcare (HIPAA), financial services (SOC 2), manufacturing (IP protection), and government workloads (FedRAMP) cannot route queries across borders or tolerate the latencies of distant cloud regions. These sectors together account for more than 40% of U.S. GDP, a reminder that the majority of economic value sits in places where centralizing inference simply isn’t allowed. The grid is distributed because society needs it. Inference will be distributed for exactly the same reason.

Three Principles of Grid-Based Inference

Just as electrical grids follow fundamental principles of physics and economics, inference grids depend on three core architectural principles:

Principle 1: Proximity Reduces Waste

The speed of light in fiber is 124 miles per millisecond. A 1,000-mile round trip imposes an unavoidable ~16 ms delay. For sub-50 ms workloads, such as fraud detection, industrial control, real-time user interaction, geography outweighs almost every other performance factor.

Moving data is even more expensive. Network transfer consumes orders of magnitude more energy and money than computation. Sending even 50 KB of context across regions often costs more than running the model itself.

The grid principle: just as electrical substations sit close to demand, inference must be deployed where users and data actually reside.

Principle 2: Hierarchy Enables Scale

Most inference doesn’t need a large-scale model. Across studies from Microsoft Azure (2024), Alibaba’s Qwen team (2024), Mistral (2023), and Databricks (2024), 60–80% of real-world queries can be handled by small-to-mid models (7B–30B) with little or no loss in quality at a fraction of the energy cost¹. Treating every request as a 70B-class problem is like powering a coffee maker from a transmission line.

The economics are stark. 7B–30B models (quantized) run on one or two mid-tier GPUs, using 5–20% of the energy and 10–30% of the cost of a 70B-class model. 70B+ models require multiple high-end GPUs and consume several hundred watts per instance. This is why production architectures increasingly converge on a two-tier hierarchy:

Tier 1: Small-to-mid models (7B–30B) at edge or metro nodes handle most routine queries, retrieval steps, and domain-specific reasoning.
Tier 2: Large-scale models (70B+) in core regional sites reserved for complex reasoning, multi-step tasks, and high-stakes decisions.

The grid principle: Keep everyday intelligence local, and use large-scale capacity only when the task truly requires it.

¹ Kernreiter et al., “Energy Efficiency of Small Language Models,” TU Wien, 2025

Principle 3: Caching Prevents Redundant Work

A customer-service chatbot handling 10,000 sessions per day typically sees 60–70% semantic overlap in user questions. But today’s RAG (Retrieval Augmented Generation) pipelines treat every session as unique, repeatedly fetching the same documents, reranking the same candidates, and reconstructing nearly identical context windows. Meta’s Retrieval-Augmented Generation at Scale (2023) study showed that 60–80% of retrieval sets in enterprise queries overlap, meaning most of this work is duplicated thousands of times.

Across a 10-node inference tier, this redundancy can easily consume 600–700 kWh/day purely on retrieval, embedding, and context assembly. Introduce even a modest 70% semantic-cache hit rate, consistent with overlap observed by Meta, Google’s enterprise QA evaluations, and Zendesk support flows, and the load drops to roughly 200 kWh/day. That’s a 60–70% energy reduction without touching the model, GPUs, or data-center design.

Why Training Data Centers Cannot Be Repurposed

The question seems logical: if an organization already owns a training data center, why not simply run inference workloads on it? The answer reveals why inference requires fundamentally different infrastructure.

The Physical Mismatch

Training data centers are engineered for 80–120kW rack density with direct-to-chip liquid cooling. Inference workloads on quantized models require 15–30kW racks with hybrid air/liquid cooling.

Retrofitting training infrastructure for inference densities means:

Removing liquid cooling infrastructure (sunk capital cost)
Replacing with lower-density cooling (new capital expenditure)
Underutilizing electrical capacity (designed for synchronized 100% TDP across all GPUs)

The net result is similar cost to greenfield construction, for infrastructure optimized for the wrong workload. Meta paused planned data center projects in 2024 specifically to rescope them for AI inference rather than attempt retrofits.² The economics don’t work.

² “AI Datacenter Energy Dilemma,” SemiAnalysis, March 2024

The Geographic Constraint

Training clusters locate where power is abundant and cheap, such as rural Oregon, West Texas, Iceland. These locations optimize for cost per kWh and megawatt availability. Inference must locate near users and within regulatory boundaries.

For example, healthcare inference must stay inside HIPAA-controlled environments, either on-premise or in HIPAA-compliant cloud regions, because the model’s output and intermediate states count as PHI (Protected Health Information). Financial services must be in-country for regulatory compliance. Manufacturing must be on-site for IP protection and control-loop latency.

A training cluster in rural Oregon cannot serve EU customers (GDPR prohibits cross-border data transfer), healthcare providers beyond HIPAA compliant regional boundaries, manufacturing plants beyond cannot 100ms+ latency for real-time control requirements.

What an Inference-Ready Infrastructure Actually Looks Like

If training era data centers can’t simply be repurposed for inference, what does work? The shape of the answer emerges directly from the three grid principles, proximity, hierarchy, and reuse. Here’s the profile that wins in the inference era, and why it looks almost nothing like the hyperscale campuses we spent the last decade building.

Geography: Close Enough to Matter, Cheap Enough to Build¹

Inference lives on a clock the cloud cannot escape. The winning sites sit within 50–100 miles of major metros, close enough to stay under ~20ms latency budgets. It is critical to consider economic access real estate, substation capacity, and permitting timelines that won’t collapse the business plan. This geography satisfies two constraints at once: Latency (get the compute near people and data) and Sovereignty (keep inference inside the required region).

Scale: Not Mega-Campuses, but Metro-Scale Nodes

Training rewarded 100MW+ multi-billion-dollar campuses where GPUs converge like iron filings around a magnet. Inference punishes that model. The sites that work best are 2–20MW. Large enough to operate economically. Small enough to fit near metros where demand actually lives. This allows a fast build up within 12 months inside a single compliance boundary. A 10MW site often can serve the inference traffic of an metropolitan region.

Deployment: A Network, Not a Planet

A training campus costs $800M to $2.5B and takes 24 to 36 months to build. A metro-adjacent 2 to 20 MW facility typically costs $20M to $120M and deploys in 6 to 12 months. This opens the market to dozens of operators.

Training thrives in one place. Inference thrives in many. The future is a tiered grid of sites, each doing what they’re best at:

Edge nodes for small models and cached semantic memory
Regional hubs for heavier models and multi-step reasoning
Specialized sovereign facilities for healthcare, finance, defense

Ownership: The Market Opens Where Hyperscalers Can’t Go

Training consolidated power in the hands of a few because the workload rewarded scale. Inference breaks that logic, not because of economics alone, but because of jurisdiction. Inference runs in hospitals, banks, factories, public agencies, and sovereign borders. Unlike training, the decision itself is the regulated object, not just the data. This redraws the ownership map. A new class of operators will emerge precisely where hyperscalers are structurally constrained. We will see:

Regional AI infrastructures aligned to local compliance zones
Vertically specialized facilities (healthcare-only, finance-only, defense-only)
Sovereign and quasi-sovereign operators in emerging digital economies
Enterprise-controlled nodes on-premise or in metro-adjacent edge sites
Distributed franchise or partnership models hyperscale economics can’t justify

This isn’t fragmentation. It’s localization as a feature.

Capital Profile: A Different Asset Class Entirely

A training campus costs $800M–$2.5B, takes 24–36 months to build, and requires a hyperscaler level balance sheet to finance the GPUs, substations, and transmission upgrades. A metro-adjacent 2–20MW facility typically costs $20M–$120M, deploys in 6–12 months, and scales by gradual growth: open one site, fill it, replicate it in the next metropolitan ring. It is lower risk. Payback periods will be shorter. The developers can build where usage actually appears. This will opens the market to dozens of operators, including regional players, vertical specialists, and sovereign providers.

Sources: Structure Research 2024 Edge DC Model, C&W Cost Index 2024, JLL 2024 Metro Edge Builds.

The Democratization Opportunity

The 2–20MW scale, metro-adjacent positioning, and distributed deployment model create opportunities for:

Regional infrastructure providers who already own real estate near population centers and can deploy inference nodes in existing facilities or adjacent greenfield sites.
Specialized operators serving regulated industries (healthcare, finance, government) that require dedicated, compliant infrastructure rather than shared hyperscaler resources.
Franchise and partnership models where domain expertise (healthcare IT, financial services technology) combines with infrastructure operation, impossible at hyperscale economics.

The grid structure democratizes both deployment and access. Small hospitals can access on-premise inference. Regional banks can serve customers with locally compliant models. Manufacturing plants can deploy air-gapped quality control.

The Path Forward

Organizations don’t need to abandon training infrastructure. But they do need to recognize that inference requires different infrastructure with different economics. Both training clusters and inference grids are necessary. Training develops the models. Inference delivers them to users. The mistake is assuming the same infrastructure serves both purposes.

The market opportunity is in infrastructure positioned for the inference profile: distributed, proximate, compliance-ready, and capital-efficient.

The three principles of grid-based inference:

Proximity reduces waste. Data movement costs more energy than computation. Deploy inference where users and data concentrate.
Hierarchy enables scale. Tier compute by demand — small models at edge, large models at regional hubs.
Caching prevents redundant work. 60–80% of queries repeat semantic patterns. Cache computation pipelines across sessions.

This isn’t about retrofitting old infrastructure. It’s about building the right infrastructure for a new era. The companies and investors who recognize this distinction early will define the next decade of AI infrastructure.

Author: Jay Chiang, Pascaline Systems Inc.
Jay Chiang is a co-founder of Pascaline Systems Inc., developing cognitive-storage and AI-infrastructure technologies that connect compute to trusted knowledge. He writes about AI systems, data centers, and the economics of intelligence.

_{This essay is Part 3 of the Pascaline AI Infrastructure Research Series. Part 1 introduced the AI Index. Part 2 highlighted inference challenges.}

Upcoming Events

Join us for AI INFRA 5 on May 01. Grab your Super Early Bird tickets!

Grab Your Discount Today

Early Bird Pricing

Bay Area Startups Collectively Secured $22B+ in January Week 1

January started out where December left off, with SV startups collecting more than $22B in funding in the first week. In a week with only 21 fundings, almost half were megadeals. In dollars, the megadeals made up 98% of the week's total. xAI was at the top of the list, with a $20B Series E, and followed by eight more megadeals, half of them in biotech.

Andreessen Horowitz's $18B raise in nine funds today comes after all US VC firms combined raised just $66B in 2025, the lowest since 2019. It was clear at the end of the year that the VCs who successfully raised new funds were the older, established firms; Andreessen's announcement today just reinforces that. The prolonged IPO drought and lack of liquidity has made fundraising by emerging managers - a challenging job in any environment - much, much tougher.

Follow us on LinkedIn to stay on top of SV funding intelligence and key players in the startup ecosystem.

Early Stage:

AirNexis Therapeutics closed a $200M Series A, a clinical-stage biotech company developing novel therapeutic drugs for the treatment of COPD patients.
LMArena closed a $150M Series A, an open platform where everyone can access and interact with the world's leading AI models and contribute to AI progress through real-world feedback.
Corgi Insurance closed a $108M Series A, a full-stack insurance carrier building better, faster insurance products for startups.
Array Labs closed a $20M Series A, building the next generation of space-based radar systems.
Topos Bio closed a $10.5M Seed, focused on AI-powered drug discovery targeting intrinsically disordered proteins.

Growth Stage:

xAI closed a $20B Series E, building artificial intelligence to accelerate human scientific discovery.
Soley Therapeutics closed a $200M Series C, , tech-enabled drug discovery and development company using cells as the world’s most powerful sensors to uncover first-in-class medicines.
EpiBiologics closed a $107M Series B, advancing a next-generation protein degradation pipeline and platform that targets extracellular membrane and soluble proteins.
Apella Technology closed a $80M Series B, an OR management platform providing hospitals with real-time operational intelligence to eliminate delays, distractions, and inefficiencies.
Canopy Works closed a $22M Series B, delivers comprehensive safety coverage through a seamlessly integrated system, delivered as a complete service.

CredPR is a strategic communications and growth firm that helps startups and technology companies scale their brand presence, attract investors, and secure meaningful media coverage. The firm combines public relations, content strategy, and community engagement to elevate innovators in competitive markets.

What CredPR Delivers
• Strategic media positioning tailored for technology brands
• Narrative development and content amplification
• Investor communications and visibility campaigns
• Integrated public relations support across channels

Why It Matters
In today’s crowded startup landscape, strong visibility and messaging are essential for growth. CredPR helps companies refine their voice, articulate their value, and reach target audiences with clarity and impact.

Who It Serves
Early-stage and growth-stage technology companies, founders seeking media traction, and teams looking to strengthen brand recognition and investor awareness.

CredPR builds momentum for companies that want to be seen and remembered. For more information, please reach out to: hello@credpr.com

Asia Ca

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

3 Tricks Billionaires Use to Help Protect Wealth Through Shaky Markets

“If I hear bad news about the stock market one more time, I’m gonna be sick.”

We get it. Investors are rattled, costs keep rising, and the world keeps getting weirder.

So, who’s better at handling their money than the uber-rich?

Have 3 long-term investing tips UBS (Swiss bank) shared for shaky times:

Hold extra cash for expenses and buying cheap if markets fall.
Diversify outside stocks (Gold, real estate, etc.).
Hold a slice of wealth in alternatives that tend not to move with equities.

The catch? Most alternatives aren’t open to everyday investors

That’s why Masterworks exists: 70,000+ members invest in shares of something that’s appreciated more overall than the S&P 500 over 30 years without moving in lockstep with it.*

Contemporary and post war art by legends like Banksy, Basquiat, and more.

Sounds crazy, but it’s real. One way to help reclaim control this week:

Skip waitlist.

_{*Past performance is not indicative of future returns. Investing involves risk. Reg A disclosures:}_{masterworks.com/cd}

Happy 2026!!! 🥂 From Firefighting To Foresight: Why AI Is Becoming Humanity’s First Window Into Tomorrow 🔭 Bay Area Startups Collectively Secured $25B+ 💰

Sat, 03 Jan 2026 18:11:03 +0000

From Firefighting To Foresight: Why AI Is Becoming Humanity’s First Window Into Tomorrow

By Doug Shannon
Forbes Technology Council
Nov 13, 2025

Doug Shannon is a global leader in digital transformation, specializing in AI, GenAI and intelligent automation.

For most of human history, we have lived in a reactive world.

A fire breaks out, and we send people running. A virus mutates, and we scramble to update vaccines. A supply chain fails, and we call it a black swan and play catch-up.

Entire systems, from governments to enterprises, have been built on the idea that risk cannot truly be seen until it has already landed. The best we managed was probability: forecasts, models and dashboards. They told us what might happen, but rarely gave us a chance to act before the storm hit.

That era is ending.

We are entering what I call the Age of Pre-Cognition. AI models no longer simply process history; they are beginning to perceive what is about to come. Civilization is stepping into a world where we can anticipate events before they manifest and take meaningful action in advance.

What Pre-Cognition Really Means

Pre-cognition may sound like science fiction, yet in practical terms it means this:

Reactive: Responding after the event. A hurricane hits, then we evacuate.
Proactive: Preparing for general categories of risk. We know hurricane season is coming, so we stock supplies.
Pre-Cognitive: Perceiving specific events before they land and allocating resources precisely. We know where and when the hurricane will make landfall, so we move people and supplies ahead of time.

This is not abstract theory. It is already happening.

The Pattern: Models That Learn the World, Not Just Words

AI began by learning language. Now it is learning reality itself, the physics, chemistry and biology of the systems we depend on.

Health and Disease: MIT’s VaxSeer forecasts dominant flu strains months before they emerge, giving vaccine developers lead time. EVEscape predicts viral escape mutations, helping us prepare countermeasures before pathogens adapt.
Weather and Climate: DeepMind’s GraphCast generates 10-day forecasts in seconds, often outperforming traditional models. Google’s Flood Hub provides flood predictions up to seven days ahead across 80 countries, protecting lives in regions that lacked early warning.
Wildfires: ALERTCalifornia and WIFIRE combine cameras, satellites and AI to detect fires faster and forecast spread, turning minutes of warning into hours.
Solar Threats: NASA and IBM’s Surya predicts solar flares up to two hours before impact. Released openly on Hugging Face, it allows enterprises, researchers and governments to protect satellites, grids, aviation and even astronauts.

Why This Is Different From Past Hype

Skeptics will say every technology has its hype cycle.

Although there is truth to this, I cannot help but think that this time the difference is more existential. Past bubbles promised productivity, connectivity or entertainment. The potential of pre-cognitive AI is about resilience, survival and foresight.

When we can see tomorrow’s threats today, it changes how everything works:

Public Health: Instead of mass infections, vaccines and treatments can be pre-allocated.
Energy and Infrastructure: Instead of cascading blackouts, grids and satellites can be protected before storms.
Enterprises: Instead of supply chain collapse, shipments can be rerouted before bottlenecks choke global trade.
Insurance and Finance: Instead of catastrophic losses, portfolios can be hedged with foresight built in.

The Human And Existential Dimension

At its core, this shift is not only about industries or enterprises. It is about humanity’s relationship with time itself.

For centuries, we lived with lag, the gap between when an event occurred and when we could respond. That lag defined wars, pandemics, economic crashes and natural disasters. Entire histories were written in the ashes of our inability to see early enough.

Now that lag is closing.

AI, trained on immense datasets of physics, biology and human behavior, gives us glimpses of what comes next. It does not make the future certain, but it makes it visible enough that choices can change. Lives can be saved. Economies can be shielded. Societies can be steadied.

What Enterprises Must Do

The organizations that thrive in the Age of Pre-Cognition will be those that stop treating AI as a tool for efficiency and start treating it as an early warning system for resilience.

This requires:

Resilience Intelligence Stacks: Observatories for data ingestion, domain models like Surya or GraphCast, decision rails linking forecasts to playbooks and connectors that turn alerts into enterprise actions.
Governance That Builds Trust: Using alignment, clarity and transparency (ACT). Document data sources through a chain of custody, clarify reasoning through a chain of thought and explain actions through a chain of reasoning.
Pre-Activation Drills: Practicing the handoff from model output to human decision to enterprise action. The question is not only whether the model is accurate but also whether the organization can act fast enough.

The Bigger Picture

This is not about AI replacing humans. It is about AI extending human foresight. Civilization is moving from reactive firefighting to pre-cognitive resilience. From scrambling after the fact to preparing ahead of time. From living in lag to living closer to tomorrow. Enterprises that embrace this shift will look as though they can see the future. Those who do not will look permanently surprised, stuck in firefighting mode, while others quietly move pieces into place.

The question is not whether pre-cognitive AI will arrive. It already has. The question is whether we will be ready to act when tomorrow becomes visible today.

THAT’S A WRAP!

..comes to an end.” A look back at an eventful year

Upcoming Events

Join us for AI INFRA 5 on May 01. Grab your Super Early Bird tickets!

Grab Your Discount Today

Early Bird Pricing

Bay Area Startups Collectively Secured $25B+ in December

December was a month of extremes in funding, with Bay Area startups collecting more than $25B. There were eighteen megadeals, with seven of them more than $200M and three billion-dollars plus. Databricks' $4.08B Series L closed on the last day of the year, demonstrating again that there is no lack of ready investment for late stage AI infrastructure companies to stay private.

2021 funding used to be the high mark, but no more. In 2024, Q4 funding for AI infra & apps really starts to kick in, well above the frothy levels of 2021 ZIRP investing. And in 2025, every month sets a record.

LinkSV 2026 Tech, Talent & Investment Trends webinar, noon PT on January 9^th

AI ruled in 2025 bringing with it an unprecedented level of funding for both infrastructure and app companies. What trends coming into 2026 will impact startups and their ability to raise capital – and how?

Pre-seed and seed fundings surged, while graduation to series A slowed.
Round sizes, valuations increased at all stages.
M&A exploded, more than 2X in numbers and dollars over 2024.
VCs raised fewer funds, capital more concentrated: 30 firms raise 74%.

Register here to join us and get it all in the January edition of the LinkSV Tech, Talent & Investment Trends Friday January 9 @noon PT.

Follow us on LinkedIn to stay on top of SV funding intelligence and key players in the startup ecosystem.

Early Stage:

Cultivo closed a $14M Series A, accelerates investment into nature at scale.
Traini closed a $7.5M Series A, innovates the way to communicate with pet dogs using generative AI and a cognitive wearable device.
Double Blind Bio closed a $5.6M Seed, developers of DeBBy, the world's first AI-powered growth agent built to help research teams close more trials with less effort.
Aravolta closed a $5.1M Seed, provides data centers with infrastructure management (DCIM) software, and digital twins, regulatory reporting, and autonomous control.
Givefront closed a $2M Seed,the Financial Management Platform built for nonprofits

Growth Stage:

uLink Labs closed a $8.6M Series B,developing a groundbreaking wireless implantable device and remote monitoring system designed to enable personalized heart failure management
Resynergi closed a $6.4M Series B, an advanced recycling technology company accelerating plastic circularity to protect human health and the environment.

You.com is an AI-powered search and assistant platform that gives users control over how they discover and interact with information. The platform blends conversational AI with structured search results so users can get answers, explore resources, and complete tasks in one place.

What You.com Delivers
• Personalized search that adapts to user preferences
• AI assistant features that generate summaries, insights, and responses
• Multiple views of results including direct answers and web sources
• Privacy-focused design that minimizes data tracking

Why It Matters
Traditional search engines often prioritize ads and rankings over relevance and user choice. You.com reimagines search so users can get information and assistance that is more transparent, customizable, and aligned with their goals.

Marianne Bekker, Head of Dev Rel at You.com

Who It Serves
Consumers, professionals, students, and anyone who wants a more intuitive and privacy-oriented way to search, explore, and interact with AI tools.

You.com delivers a modern search experience built for the age of AI.

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

The Data Center Capacity Crisis is Here💰 Bay Area Startups Collectively Secured $15B+ 🐣 Super Early Birds for AI INFRA SUMMIT 🎅🏻 Happy Holidays 🎄

Sat, 27 Dec 2025 16:20:48 +0000

🏗️ The Data Center Capacity Crisis is Here.

_Author:_{Gregg Scharfstein}_{, Infracore}

The data center market does not feel like it is going through a temporary squeeze. It feels different at a basic level, more constrained and less forgiving. A few years ago, most requests still sat in a range that allowed some flexibility. Five megawatts was common. Ten was notable. Today, initial conversations routinely start at ten or twenty megawatts, with clear expectations that those numbers will increase.

Many buyers already know where they want to end up, even if they are unsure how to get there or how long it will take finding usable capacity has become its own problem. Buyers are working through multiple brokers and comparing availability lists that look comprehensive until they are cross checked line by line. The same facilities show up under different names, often with slightly different assumptions attached.

Time goes into calls, site visits, and spreadsheets, only for the same issues to surface. Power is not actually secured. Cooling designs do not support current density requirements. Some sites technically exist but sit in locations that fail latency, regulatory, or staffing constraints. When a site remains open, there is usually a reason that becomes obvious during diligence, sometimes within the first few technical questions.

Utility timelines have stretched far beyond what most deployment plans assumed. Power delivery that once took two years now regularly runs three or more. Transformer availability is inconsistent and hard to predict. Substation upgrades push schedules out to the end of the decade. These delays are not unusual anymore. They are part of nearly every serious discussion. In an environment where speed matters, waiting years for grid power removes entire opportunities from consideration and forces uncomfortable tradeoffs.

How capacity is used has changed as well. Inference workloads, edge use cases, and data sovereignty requirements have narrowed where deployments can realistically happen. Broad geographic flexibility disappears quickly once latency, compliance, and workforce access are accounted for. What sounds flexible at the start often becomes highly constrained within a few weeks of planning, sometimes after resources have already been committed.

Your 2026 Data Centre Infrastructure Roadmap
Signal #12 (of 12): "Path to 2030" 🏁

Speed now includes everything that happens before construction. Months spent searching followed by a year or more of build time means long stretches where hardware sits unused. Revenue projections slip. Market windows close. Teams lose momentum internally while waiting for external dependencies to resolve. These costs rarely show up cleanly in early models, but they are felt later, often after decisions can no longer be reversed.

As a result, more teams are looking toward existing buildings and alternative power arrangements. Brownfield facilities in well connected areas reduce uncertainty and remove layers of dependency. Behind the meter power avoids long utility queues and simplifies scheduling risk. InfraCore is seeing demand for these approaches because they allow projects to move from contract to operation in under a year, sometimes closer to half that.

The capacity shortage is shaping decisions already being made. Success over the next cycle will not come from having the most options on paper. It will come from securing capacity that actually works, in locations that support deployment realities, and committing before delays compound.

🎄 𝐇𝐚𝐩𝐩𝐲 𝐇𝐨𝐥𝐢𝐝𝐚𝐲𝐬 𝐟𝐫𝐨𝐦 𝐈𝐠𝐧𝐢𝐭𝐞𝐆𝐓𝐌🎄

Upcoming Events

Join us for AI INFRA 5 on May 01. Grab your Super Early Bird tickets!

Grab Your Discount Today

Early Bird Pricing

Bay Area Startups Collectively Secured $15B+ in December Week 4

The frenetic pace of funding slowed down this week in the two days before the holidays kicked in. The $15B strategic investment in Anthropic by Nvidia and Microsoft spiked December's total to $20.59B.

Exits, M&A: The pace of tech acquisitions is more than 2X that of 2024 in numbers and in dollars. ServiceNow's $7.75B acquisition of Armis was the 7th billion-plus acquisition of a local company in December. Those seven acquisitions total more than $29B, a shot in the arm for investor, employee and founder liquidity.

Acquisitions in the AI sector have been muted in 2024 and 2025, with multiple potential acquisitions replaced by “reverse acqui-hires”. The latest of these is reportedly Nvidia paying ~$20B to license competitor Groq's AI chip technology and hiring Groq's CEO and senior execs. They did a similar deal in September with Enfabrica, avoiding the regulatory scrutiny expected in acquiring the companies. Like other reverse acqui-hires, the companies and remaining employees are left to pivot, having been stripped of their talent and the exclusive right to the technology they developed.

Follow us on LinkedIn

Early Stage:

AIO closed a $17M Series A, AI platform for restaurants that seamlessly integrates order and pay devices with marketing, staff and inventory management while automating back-office operations and accounting.
Known closed a $9.5M Seed, intelligence for real life connections, starting with dating.
Dazzle AI closed a $8M Seed, closing the gap between what people want and what they can do with AI.
Internet Backyard closed a $4.5M Pre-Seed, automates quoting, billing, payments, and financial workflows for compute providers.
Drafted closed a $1.6M Seed, designs your dream home instantly with AI.

Growth Stage:

Anthropic closed a $15B Series Unknown, an AI safety and research company developing AI systems that are helpful, honest and harmless.
Kargo closed a $42M Series B, creating a universal interpreter for supply chains by using computer vision to connect the physical world of freight to the digital systems that manage it.

Novos Power is an energy technology company focused on low-cost, low-carbon power solutions for industrial and data center applications. The company develops modular power systems that can supply reliable electricity while reducing overall energy costs and environmental impact.

https://www.novospower.com/

What Novos Power Delivers
• Integrated power plants that produce electricity at competitive, industrial-scale rates
• Low-carbon energy solutions that help companies reduce emissions
• Scalable systems designed for high-demand workloads, including compute and critical infrastructure
• A modular approach that accelerates deployment and minimizes construction risk

Why It Matters
Energy costs and carbon intensity are key challenges for industries with heavy power demand, such as AI infrastructure and manufacturing. Novos Power’s systems provide a viable alternative to traditional grids, helping organizations control costs and improve sustainability.

Novos power at AI INFRA SUMMIT

Who It Serves
Data centers, industrial operations, energy buyers, and organizations seeking dependable, cost-competitive, and cleaner power.

Novos Power delivers energy solutions to support high-growth technology and industrial ecosystems.

Contact them today.

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

Introducing the first AI-native CRM

Connect your email, and you’ll instantly get a CRM with enriched customer insights and a platform that grows with your business.

With AI at the core, Attio lets you:

Prospect and route leads with research agents
Get real-time insights during customer calls
Build powerful automations for your complex workflows

Join industry leaders like Granola, Taskrabbit, Flatfile and more.

👉Try Attio Pro for free

🤖 You Should Be Scared About Inferencing 💰 Bay Area Startups Collectively Secured $2.76B 🤝 AI Tech Leaders Club Recap 👨🏽‍💻

Sat, 20 Dec 2025 16:00:30 +0000

Read full article: here All eyes turned to computation. Few saw the consequences.

🤖 You Should be Scared of Inferencing
by Jay Chiang, CTO of Pascaline Systems

Training was the easy part.

I know it doesn’t sound like it. Especially not against the backdrop of billion-dollar clusters, megawatt power draws, and an entire economy built around GPU scarcity. But training is a project. It has a beginning, a budget, and an end.

Inference only looks easy.

Yes, each query is lighter, each model smaller. But that illusion hides a harder truth: inference never stops. It’s everything, everywhere, all at once. It scales not with chips, but with society. It doesn’t just compute. It’s what turns machine intelligence into social reality.

The GPU boom created the impression that the AI race is over. It seems that hyper scalers own the stack and the rest of us merely consume it. But training was the sprint. Inference is the marathon that follows, and few are prepared for the distance

The Infrastructure Illusion

Investors have poured hundreds of billions into AI infrastructure over the past three years. According to McKinsey’s 2024 “AI Infrastructure Outlook”, data-center CapEx is projected to exceed $400 — 450 billion by 2025, with more than 70 percent dedicated to training-centric capacity. (Deloitte AI Infrastructure 2024 and Dell’Oro Group 2024 reports show similar magnitudes.) The playbook was simple: build ever-larger campuses where power is available or cheapest. The physics of inference breaks that logic.

“Many believe they’ve mastered the playbook. What they have is the wrong chapter.”

Training sites thrive on concentration — hundreds of megawatts, predictable loads, batched traffic. Inference sites thrive on distribution — low latency, regulatory locality, variable demand. The first are built where energy is available; the second must be built where data and users actually are.

“We are entering a phase where inference will demand smaller, geographically distributed facilities — not just larger ones.” — Synergy Research Group, AI Data Center Update 2025
“The next big constraint is not compute power but where that compute can be located.” — International Energy Agency, Electricity 2025 Report

And that’s the illusion: billions committed to megawatt campuses while the next wave of demand splinters into thousands of smaller nodes. What happens when the world realizes the training build-out can’t host the inference economy? Do investors double down on sunk assets or pivot to where usage is already racing ahead?

Our AI Inference Activity Index (AIAI) (see Issue 1) already shows usage leading capital by nearly two years. It’s a wake-up call.

The Enterprise Gap

Every enterprise fears to be left behind. Generative AI is now boardroom doctrine. Executives race to show movement — hire a team, rent some GPUs, issue a press release. For many, that is the plan.

Then comes the fear of not being ready. Pilot projects stall. Integrations break. Costs mount. According to the MIT report The GenAI Divide: State of AI in Business 2025, “The outcomes are so starkly divided across both buyers… and builders… that we call it the GenAI Divide. Just 5% of integrated AI pilots are extracting millions in value, while the vast majority remain stuck with no measurable impact”.

Most companies quietly realize that adding AI to a software stack is not the same as running an intelligent system end-to-end. As soon as they start to address the technology challenges, the fear of the talent gaps arrives. Everyone knows the story by now. There simply aren’t enough experts. Over half of IT leaders report a worsening AI skills shortage. (CIO Dive, 2025)

One study estimated only about 22,000 true AI specialists worldwide, with millions of unfilled positions across industry. (Keller Executive Search, 2025) NVIDIA’s Jensen Huang warned recently that China is training more than a million AI engineers each year — a scale Western economies cannot yet match. (Reuters, 2025) That warning rattled markets, helping trigger a $1 trillion pullback in tech valuations as investors questioned whether enterprises could execute fast enough.

But these are the visible fears. The most underestimated is the fear of inference itself. Ironically, the more successful you are, the more likely you will face this one.

Inference isn’t an algorithmic milestone. It’s a continuous economic function. Every chat reply, every recommendation, every automated decision is a live transaction involving data movement, model execution, retrieval, and expectation. Each consumes compute and energy. None scale linearly. Unlike training, inference never ends. Every successful deployment multiplies its own cost base.

“Inference must win on economics.” — Liran Zvibel, CEO of WEKA IO

Yet most enterprises can’t even measure that economy.

They worry about headcount when they should worry about headroom — how to serve intelligence sustainably under volatile, always-on demand.

And underneath the spreadsheets lies the social layer no one models: how decisions change workflows, how automation reshapes teams, how trust shifts between humans and algorithms. Inference is not just technical. It is social, organizational, and moral infrastructure. Almost no enterprise is prepared for that.

The Technologist’s Blind Spot

Technologists are most comfortable when they can predict what comes next. This time, they can’t. Inference doesn’t behave like the systems they’ve mastered. Its rules are still being written while the machines are already running.

Training rewards averages. Inference punishes outliers. At hyperscale, one slow shard becomes the user experience. Jeff Dean and Luiz Barroso called this out a decade ago in The Tail at Scale: “Temporary high-latency episodes may come to dominate overall service performance.”

Inference lives permanently in that tail. Every request can differ wildly in length, context depth, and memory footprint. A single cache miss or congested hop can saturate a GPU, stall a pipeline, or trigger cascading retries. Adding GPUs hides inefficiency — it doesn’t cure it.

Serving keeps context alive. Each session brings its own history, embeddings, and KV cache, stretched across GPU, CPU, and storage. Keeping that state coherent decides your p99 latency and cost. Adding GPU can’t resolve it. We need memory. And I don’t mean the hardware. Each inference session carries states, such as conversation history, vector embeddings, cached retrievals , practically fragments scattered across GPU HBM, CPU DRAM, NVMe SSD, and networks. Keeping all that coherent is not a solved problem.

Recent experiments such as PagedAttention (UC Berkeley, 2024) and Pensieve (NYU, 2024) underscore how inference memory is still a research problem and not an engineered one. Both highlight the same fragility: state management, not raw compute, is the new bottleneck.

2025 is being called the Year of Agentic AI, which means we have to consider When Models Talk Back. Longer context windows gave engineers false comfort. It is a relief when humans were generating prompts. Then came agentic AI. They don’t just answer. They think aloud, branch, call tools, verify, and retry. They spend tokens the way supercomputers spend FLOPs.

Recent research on test-time compute (Anthropic 2024, DeepMind 2025) shows that multi-agent reasoning can amplify token generation 10–25× per query — a complete inversion of inference economics.

This new frontier isn’t about size alone. It’s about control — how far to let agents think, how deep to recurse, how to prune search trees in flight. Every decision changes cost, latency, and sometimes the answer itself.

“We’re learning that inference is no longer passive. It plans, acts, and spends.” — MIT CSAIL Workshop 2025

That leads a cruel fact. The Data Plane Is Breaking. Inference now generates its own gravity. Each query creates new embeddings, logs, feedback, and governance trails, data that must persist for learning and compliance. Add localization laws and retention mandates, and the old cloud pattern, stateless microservices over object stores, collapses.

Enterprises now need semantic memory planes: fast retrieval objects, compressed context stores, and policy-aware caches that bridge secure domains. Inside an organization, AI memory must be partitioned by trust, teams, customers, auditors, yet remain instantly searchable. That architecture barely exists today.

That is not all. Real-World Inference Is Physical. When AI leaves the browser and enters cars, robots, headsets, and factories, latency stops being cosmetic. In AR/VR, motion-to-photon must stay under 20 ms. In vehicles, perception-to-actuation must be deterministic. Variance, not speed, is the killer. Expect Time-Sensitive Networking (TSN), edge colocation, and deterministic fabrics to become as critical as GPUs. Inference at the physical edge has to think — and respond — on a schedule the cloud was never built for.

In summary, technologists should consider

Tail-first design — Engineer for p99 (99 percentile) from day one. Throughput can wait.
State as infrastructure — Treat conversation, KV, and embedding caches as tiered, observable systems, not black boxes.
Govern inference — Cap agent recursion depth, budget “thinking tokens,” and verify outputs.
Build deterministic paths — For AR, robotics, and automotive, latency jitter kills trust.

We’ve learned how to train large models. Now we have to learn how to serve them.

The Slow Policy Machine

Policy always trails technology, but inference has turned the lag into a chasm. Training lived in data centers, under lock and audit. Inference lives everywhere — in hospitals, banks, schools, public spaces, and the phones in our pockets.

Governments are starting to wake up. In 2025, the European Union’s AI Act entered its enforcement phase, classifying real-time inference systems in critical domains as “high risk.” The U.S. Executive Order on Safe, Secure, and Trustworthy AI mandates reporting of compute clusters above a training threshold, yet says little about inference, the phase actually touching citizens. China’s Generative AI Provisions (2024) require content filtering and provenance, but again, not energy, latency, or localization.

Everyone regulates the model. No one regulates the behavior.

“We’ve built a system where AI acts faster than governments can think.” — MIT Technology Review, Policy 2025 Roundtable

That’s the real fear for policymakers: inference is not just a workload, it’s a decision factory — one that operates in microseconds and everywhere at once.

And the infrastructure beneath it is straining. The International Energy Agency (IEA, 2025) now projects global data-center electricity use will reach 950 TWh by 2030, nearly doubling from 2023, with AI the main driver. Yet permitting cycles for new capacity still take 3–5 years. Grid expansions move in fiscal decades, not milliseconds.

Sovereignty laws like the EU’s Data Act, India’s Digital Personal Data Protection Act, and Saudi Arabia’s National Data Management Regulations are forcing inference to occur within borders, turning compliance into a physical siting problem.

Every millisecond of inference now carries a passport, a carbon cost, and a moral implication. And none of those fit neatly into existing frameworks of law, labor, or liability. Policy isn’t just slow — it’s built for a world that assumed decisions were human.

The fear isn’t only for ministers and regulators. Citizens should be scared too, because inference isn’t an app you install. It’s an ambient power that shapes how you see, decide, and act, before anyone has agreed on who governs it.

Where Fear Meets Design

Every fear points to a missing design principle.

The investor’s fear: misaligned capital, which is a spatial problem.
The enterprise’s fear: runaway OpEx, which is an architectural problem.
The technologist’s fear: unpredictability, which is a systems problem.
The policymaker’s fear: loss of control, which is a temporal problem.

All of them converge in the same place: how we serve intelligence. The risk isn’t that AI becomes too powerful. It’s that we mistake its continuity for simplicity.

Inference was never meant to be handled by the old web stack. Its units aren’t requests; they’re relations — between users, agents, data, and time. Each relation carries memory, regulation, and risk. Serving it safely demands new layers of awareness in both hardware and software.

The emerging blueprint is taking shape:

Spatial locality — inference must live closer to data and users.
Temporal persistence — context must survive across sessions, not reset with every query.
Semantic control — memory must know what it stores and why.
Economic observability — systems must track joules and dollars per decision, not per GPU-hour.

This is not a feature list. It’s a manifesto for stability in an intelligent world.

“The next revolution in AI won’t be smarter models — it will be infrastructure that knows it’s intelligent.” — Pascaline AIAI Research, 2025

When design absorbs fear, fear becomes discipline. That’s how every new infrastructure era begins — when the engineers start designing for the things everyone else is afraid of.

Inference is the economic heartbeat of the AI era, but we’re still measuring it with training-era instruments.

“The Consulting Game” Workshop

AI Tech Leaders Club / Dec16

AI Tech Leaders Club brought together a packed room of senior technical decision makers on December 16 for an evening of focused, peer-to-peer problem solving on AI integration. Hosted by Jay Chodagam and Vlad Pavlov, the session centered on shared challenges in deploying AI across complex organizations. Participants represented leading Silicon Valley teams driving AI strategy, engineering, and operations.

The workshop blended structured small-group discussions with open conversation, giving leaders space to compare real-world use cases and constraints. The room stayed engaged well into the evening as teams traded lessons on technical architecture, talent models, and responsible rollout patterns.

Linda Yang of Supermicro leads her workgroup

Why it mattered: Conversations surfaced concrete paths to move from experimentation to production-grade AI, highlighting both quick wins and longer-horizon bets.The momentum coming out of the session reinforced the value of AI Tech Leaders Club as a system for ongoing signal-sharing across the Valley.

Upcoming Events

Grab Your Discount Today

Early Bird Pricing

Bay Area Startups Collectively Secured $1.57B in December Week 3

Bay Area startups closed on $1.57B in the third week of December, bringing the MTD to $5.49B. 60% of the total came from six megadeals, Homebound ($300M), MoEngage ($180M), Chai Discovery ($130M), Mythic ($125M), Addition Therapeutics ($100M) and Nirvana Insurance ($100M).

IPOs - The fourteen 2025 Silicon Valley tech IPOs raised more than $7B, a significant uptick from 2024. Wealthfront's IPO earlier this month was the last, while Figma's $1.2B IPO was the largest of 2025 and the only one to raise more than a billion dollars.

Follow us on LinkedIn

Early Stage:

Link Cell Therapies closed a $60M Series A, developing technologies that allow for safe targeting of multiple antigens, reducing the risk of on-target, off-tumor toxicity.
Valerie Health closed a $30M Series A, using AI to modernize patient and provider communication interfaces for independent practices.
HEN Technologies closed a $20M Series A, building the world’s first end-to-end intelligent fire suppression ecosystem, powered by AI, IoT and advanced fluid dynamics.
Vybe closed a $10M Seed, a "Lovable for internal apps", empowering semi-technical folks to ship internal apps.
Interface closed a $3.5M Seed, providing the energy sector's AI-native management system.

Growth Stage:

Homebound closed a $300M Series D, the first tech-powered homebuilder, making building your perfect home simple, human, and transparent.
MoEngage closed a $180M Series F, providing insights into customer behavior, enabling users to engage customers across web, mobile, email, social, and messaging channels.
Chai Discovery closed a $130M Series B, builds frontier AI to predict and reprogram the interactions between biochemical molecules, the fundamental building blocks of life.
Mythic closed a $125M Series D, accelerating computing with Analog Processing Units, a different approach to AI inference that collapses compute and memory into a single plane.
BluePath Finance closed a $75M Series C, a leading financier of sustainable infrastructure.

https://hyperaccel.ai/

HyperAccel is a semiconductor company building purpose-designed hardware for generative AI inference. The company focuses on accelerating large language models with architectures optimized for latency, throughput, and energy efficiency.

What HyperAccel Delivers
• Specialized AI processors designed for large language model inference
• Server systems optimized for high-throughput, low-latency workloads
• Hardware architectures that balance memory bandwidth and compute efficiency
• A supporting software stack that integrates models with custom silicon

Why It Matters
As generative AI workloads scale, traditional hardware struggles with cost, power consumption, and latency. HyperAccel addresses these constraints with inference-first silicon built for real-world deployment at scale.

HyperAccel team at AI INFRA SUMMIT NOV7

Who It Serves
Cloud providers, data center operators, enterprises, and AI builders seeking more efficient hardware for large-scale inference workloads.

For more information contact: Jay Kim jayckim@hyperaccel.ai

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

What investment is rudimentary for billionaires but ‘revolutionary’ for 70,571+ investors entering 2026?

Imagine this. You open your phone to an alert. It says, “you spent $236,000,000 more this month than you did last month.”

If you were the top bidder at Sotheby’s fall auctions, it could be reality.

Sounds crazy, right? But when the ultra-wealthy spend staggering amounts on blue-chip art, it’s not just for decoration.

The scarcity of these treasured artworks has helped drive their prices, in exceptional cases, to thin-air heights, without moving in lockstep with other asset classes.

The contemporary and post war segments have even outpaced the S&P 500 overall since 1995.*

Now, over 70,000 people have invested $1.2 billion+ across 500 iconic artworks featuring Banksy, Basquiat, Picasso, and more.

How? You don’t need Medici money to invest in multimillion dollar artworks with Masterworks.

Thousands of members have gotten annualized net returns like 14.6%, 17.6%, and 17.8% from 26 sales to date.

Explore offerings

_{*Based on Masterworks data. Past performance is not indicative of future returns. Important Reg A disclosures:}_{masterworks.com/cd}

⚙️ The AI Upheaval Nobody Saw Coming 💰 Bay Area Startups Collectively Secured $2.76B 🤝 Enterprise AI Day Recap 💃

Sat, 13 Dec 2025 16:05:24 +0000

⚙️ The AI Upheaval Nobody Saw Coming

AI is finally reaching the parts of the real economy that software mostly bounced off for forty years, including restaurants, logistics, agriculture, and construction. Our friend Devansh wrote a fantastic piece that is chock full of good info. Anyone building or investing in AI for physical operations should read the full article here:

The core argument in his article is that a new AI native stack has quietly flipped the economics of digitizing thin margin, chaotic, brownfield industries, while simultaneously tightening control in the hands of platforms, hardware vendors, and policymakers.

Why software failed the real economy

For decades, digitization in sectors like trucking or full service restaurants was a rational non choice, because on 3 to 5 percent net margins, a multi million dollar software rollout with a one third success rate was effectively a bankruptcy bet. High upfront cost, high failure rates, and extreme workforce turnover meant the safest system was still a clipboard and messaging apps, not an enterprise resource planning platform.

The article frames this as an evolutionary valley, where operators could see a better digital future, but crossing the valley meant a temporary and often fatal drop in performance that thin margin businesses simply could not survive. In other words, brownfield industries were not anti technology; they were doing basic expected value math.

The new AI stack that changes the math

The upheaval starts when the cost and reliability AI systems finally match the constraints of physical work. Devansh points to four shifts: AI assisted development collapsing build cost, edge computing providing offline deterministic behavior, small inexpensive models making inference economically viable, and agentic systems that adapt to messy human workflows instead of forcing standardization upfront.

Together, these turn software from a rigid capital intensive bet into something closer to a configurable appliance that can live on a small edge device in a kitchen or warehouse. The economics are made concrete by walking through camera driven use cases in quick service restaurants, showing how cloud vision models can be dozens of times more expensive than running quantized models on local edge hardware when amortized over millions of monthly frames.

A time boxed and uneven opportunity

On the demand side, the article introduces a Digital Viability Index that fuses 2025 gross margins with an automation readiness score to identify where AI is effectively a forced adoption. Agriculture, food wholesale, grocery, construction, and trucking emerge as sectors where margins are brutal, workflows are structured enough to automate, and operational pain is acute.

Across millions of establishments and trillions of dollars in activity, he estimates a five year annual recurring revenue window in the mid single digit billions for brownfield AI stacks, using adoption curves calibrated on historical technology diffusion like modern payments and cloud migration. Yet this upside is structurally skewed, because franchises and multi site operators with capital, IT staff, and training infrastructure are positioned to adopt sovereign edge stacks early and capture most of the value.

The hidden AI tax and recentralization

The second half of the piece warns that the cost collapse is not a free lunch but a shift from upfront build cost to the ongoing cost of supervising probabilistic systems. Even a small hallucination rate becomes double digit workflow failure when chained across many operational steps, forcing investments in oversight, human in the loop controls, and fallback logic. At the same time, AI generated code brings higher churn, complexity, and worrying security statistics, which makes audits and refactoring unavoidable operating expenses.

Read the full article here

Overlay this with market structure, and the picture gets darker as incumbents bundle AI features at no extra charge to crush smaller entrants, while public procurement and subsidy programs tend to favor large enterprises that already have compliance and grant writing capacity. The net effect, as he argues, is that brownfield operators are being pushed into AI adoption to survive, but much of the long term value flows upward to model vendors, hardware providers, and policy favored platforms.

For founders, investors, and operators, the article is both a map and a warning, showing that AI is finally economically compatible with the messy, low margin real economy, but unless architecture and strategy are designed for sovereignty from day one, the upheaval may end with many small players as tenants in someone else’s stack.

Enterprise AI Day Recap:
From Hype to Production Reality

On December 11th, leaders across the AI infrastructure stack—from GPU orchestration to edge video AI to enterprise platforms—gathered for Hotel Valencia Santana Row to tackle a hard question: how do we turn today’s AI capabilities into real, scalable enterprise value?

A few themes surfaced repeatedly:

Data, not models, is the bottleneck. Making sense of complex, fragmented enterprise data (often 15–17 sources per use case) and building robust ontologies is where most of the cost and time live—far more than prompt engineering.
Economics are about to matter a lot more. Token spend, GPU scarcity, and low utilization (often ~30–35%) mean that Twice the performance at half the price is becoming a requirement, not a nice-to-have. Right-sizing compute (GPUs for training, cheaper accelerators/CPUs for inference) and moving workloads closer to data—at the edge—are emerging best practices.
Agents are the next frontier. 2025 has been the experimentation year; 2026 will be about production, agentic workflows. Enterprises are already deploying AI assistants for IT, finance, and customer support, with serious focus on governance, observability, and access control as the number of agents explodes.
Open, hybrid, multi-vendor will win. CIOs want flexibility: multiple GPU vendors, hybrid cloud/on-prem, and “bring your own model/data/channel” instead of lock-ins.

The consensus: the models are ready. The winners will be the teams that master data, economics, and governance to deliver measurable business outcomes, not just AI demos.

View the full Gallery

Upcoming Events

AI Tech Leaders Club Workshop

Early Bird Pricing

Bay Area Startups Collectively Secured $2.76B in December Week 2

Bay Area startups closed on $2.76B in the second week of December. There were five megadeals this week, making up almost 80% of the total: OpenAI ($1B), Unconventional AI ($475M), AirWallex ($330M), Harness ($245M), and fal ($140M). We're seeing an increasing number of AI companies closing their second and third rounds in the same year. The latest was today's $75M Series B funding for Serval, less than three months after their series A, here's seven more that received multiple rounds this year and were funded again this month, page through using the arrows at the top: December Multiple rounds 2025.

Exits - M&A rolls on with 20 more acquisitions this week, but only two of note: the biggest, Confluent's acquisition by IBM for $11B and Newfront Insurance's acquisition by WTW for $1.05B.

For startups raising capital and sales people and service providers looking for who's just closed new fundings: The Pulse of the Valley weekday newsletter keeps you current with capital moving through the startup ecosystem in SV and Norcal. Startups raising rounds and their investors; investors raising and closing funds; liquidity (M&A and IPOs) and the senior executives on both sides starting new positions. Details include investor and executive connections + contact information on tens of thousands of fundings. Check it out with a free week's trial, sign up here..

Follow us on LinkedIn to stay on top of what's happening in 2025 in startup fundings, M&A and IPOs, VC fundraising plus new executive hires & investor moves.

Early Stage:

Unconventional AI closed a $475M Seed, rethinking the foundations of a computer to optimize energy efficiency for AI.
ProLynx closed a $70M Series A, a biotechnology company dedicated to developing ultra-long-acting medicines for obesity and other metabolic diseases.
Runware closed a $50M Series A, delivers AI-as-a-Service at 90% lower cost and with higher speed than competitors.
Kilo Code closed a $8M Seed, the all-in-one, agentic platform for software developers.
Rotostitch closed a $1M Pre-Seed, transforming apparel manufacturing powered by proprietary hardware and software.

Growth Stage:

Harness closed a $240M Series E, the AI DevOps Platform™ company, enabling engineering teams to build, test, and deliver software faster and more securely.
fal closed a $140M Series D, the leading platform for real-time generative media, offering cutting-edge inference solutions for AI-powered video, image, 3D, and audio applications.
Serval closed a $75M Series B, the AI platform for IT teams, combining an AI-native ITSM with integrated workflow automation and access management.
Solve Intelligence closed a $40M Series B, AI that helps you write high-quality patents quickly and in all stages of the patent life cycle, from drafting and filing to prosecution and opposition.
SafeinHome closed a $25M Series D, a leading provider of Remote Supports, empowering people with disabilities and older adults to live more independently and safely at home and in their communities.

Questions? Contact: Dawn DeBruyn

https://www.octopyd.com/

Octopyd is an AI-first recruitment platform that helps companies hire top talent through a rigorous, structured vetting process that blends automation with human expertise.

What Octopyd Delivers

AI-driven candidate sourcing and screening
A multi-stage vetting process combining technical assessment and expert review
Consistent evaluation for skills, experience, and team fit
Faster time to hire with reduced hiring risk

Amir Pourmousa & Mashid Fardadi at AI INFRA SUMMIT

Why It Matters
Hiring remains one of the most expensive and time-consuming challenges for growing companies. Octopyd reduces uncertainty by delivering candidates who are thoroughly vetted, aligned with role requirements, and ready to contribute from day one.

Who It Serves
Startups, mid-market companies, and enterprises looking to scale teams efficiently without sacrificing quality, along with HR and talent leaders focused on long-term retention and performance.

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

Get a Senior Growth Team to Review Your Marketing Strategy.

Galactic Fed’s senior growth team has scaled brands like Varo, Edible, and Quiznos. And now they’re opening a limited number of free 1:1 sessions with these growth geniuses to help ambitious founders find their next breakthrough.

They’ll dig into your traffic, funnels, and ad performance to surface what’s working, what’s wasting money, and where your fastest growth levers really are.

Spots are limited, but one session could change your next quarter or your entire fiscal year.

Reserve your session before they fill up. →

✅ AI Reality Check for Enterprise? BA Statups Secured $1.5B in Dec 💰 Enteprise AI Day Dec 11 📍

Sat, 06 Dec 2025 16:00:50 +0000

Enterprise AI: Reality Check

At AI INFRA SUMMIT 2025 in San Francisco, many discussions returned to a single truth. Enterprise AI adoption looks promising on the surface, but the gap between ambition and execution remains wide. Surveys show that while most large companies now report some form of generative AI deployment, only a subset are consistently achieving measurable ROI at production scale. Companies want AI to transform operations, yet many struggle to move beyond experiments. The maturity of infrastructure, governance, and internal alignment determines success far more than the model itself.

The Illusion of Progress

Enterprises often report active AI initiatives. Many have prototypes, early pilots, or small deployments inside isolated departments. These efforts create a sense of momentum, but they rarely reflect full institutional adoption. Most organizations still rely on manual processes, disconnected systems, and legacy data flows that are not ready for continuous AI workloads.

This disconnect shows up in the operational layer. AI systems require predictable data quality, consistent compute access, and reliable monitoring. IDC reports that spending on compute and storage hardware for AI deployments grew sharply in 2025, with the bulk of that investment still concentrated among hyperscalers and major cloud providers rather than traditional enterprises. Without these elements, even well designed models fail to deliver long term value. Experiments become stalled projects instead of engines for transformation.

The Missing Foundations

True enterprise adoption depends on the often overlooked components. Governance, telemetry, and capacity planning are no longer optional. They form the operating framework that allows AI systems to run safely and predictably.

Governance sets the rules for data usage, model oversight, and risk. Telemetry provides visibility into performance, cost, and behavior. Capacity planning ensures that compute and storage can handle peak demand and that models do not fail during critical operations. Without these foundations, enterprises struggle to scale AI beyond isolated demonstrations.

The Shift Toward Operational Discipline

The summit highlighted a clear shift in how enterprises approach AI. Instead of focusing solely on experimentation, many are investing in infrastructure that supports repeatable and reliable outcomes. Late‑2025 infrastructure outlooks project global AI infrastructure spending climbing toward the high hundreds of billions of dollars by the end of the decade, with accelerated servers making up the vast majority of AI server spend. This includes unified data pipelines, standardized deployment frameworks, and monitoring systems that track every step of the model lifecycle.

These investments move AI from the world of innovation labs into the core of business operations. Once enterprises treat AI as part of their production environment, adoption accelerates. Teams receive consistent inputs, leadership gains predictable outputs, and projects develop clear accountability.

Why Adoption Still Stalls

Even with progress, many enterprises struggle with culture and alignment. AI requires cross functional partnership. It depends on collaboration between technical teams, operations leaders, compliance groups, and business units. Many organizations continue to work in silos. As a result, they deploy tools without redesigning the workflows around them.

This is the root cause of many failures. Enterprises try to add AI onto old processes instead of shaping new processes around the capabilities of AI. Without structural change, gains remain limited.

Looking Ahead

The enterprise AI landscape is entering a more realistic phase. Enthusiasm has been replaced by assessment. Total AI spending is now measured in the trillions of dollars globally, with a growing share earmarked specifically for infrastructure, data center modernization, and governance rather than just model development. Companies are learning that adoption requires discipline, structure, and long term planning. The organizations that succeed will be those that treat AI as a core operational layer rather than a side project.

Upcoming Events

On Dec 11th Supermicro and AMD will host Enterprise AI Day at the stunning Hotel Valencia Santana Row, a half-day gathering featuring enterprise-leading panels and holiday-season networking.

Finally, on December 16, we’re teaming up with the AI Tech Leaders Club to close out the year with a hands-on workshop designed specifically for senior Technical Decision Makers in Snowflakes office in Menlo Park.

AI Tech Leaders Club Workshop

Aiify.io has increased our HealthTech Week discount to 30% in honor of Thanksgiving Week! Join us at HealthTech Week 2026 (coinciding with JPMorgan Healthcare Conf), which features the 3-day HealthTech Summit (Jan 14-16). Thanksgiving discounts are available, and you can use our code, “IGNITE30” for an additional 30% off.
This link automatically adds the code (discount is visible at checkout): https://luma.com/HTSummit=IGNITE30

Visit the event site for more information, including confirmed speakers and special events like their Startup Pitch Competition and the world's first HealthTech Wearables Fashion Show: http://HealthTechWeek.org

Bay Area Startups Collectively Secured $1.1B in December

Bay Area startups closed on $1.1B in the first week of December. There were five megadeals this week, making up more than 60% of the total: Harvey, XLight Inc., Angle Health, Axiado, and Verkada. Harvey stands out as this is the third round the company has closed this year, each one at a multiple-billions increase in valuation. Their total capital now stands well above $1B.

Exits - M&A has continued at an increased and steady pace since the beginning of the third quarter. There have now been 440 acquisitions, but only 16% (70) have disclosed prices. For those 70, the total stands just over $238B (excluding today's Netflix acquisition of Warner Bros.). Tech and biotech IPOs this year have been 3X last year's numbers, and one of each – Figma and BilliontoOne – did really well. We still have Wealthfront's IPO coming up later this month, but we'll be waiting until next year for an AI IPO (OpenAI? Anthropic?) to test the public market waters.

Follow us on LinkedIn to stay on top of what's happening in 2025 in startup fundings, M&A and IPOs, VC fundraising plus new executive hires & investor moves.

Early Stage:

Vinci closed a $46M Series A, brings physics-accurate design and simulation to the desk of any hardware engineer, at full resolution and up to 1000x faster than legacy tools.
Ricursive Intelligence closed a $35M Seed, a frontier AI lab building the compute foundation for the next generation of AI.
Mixx Technologies closed a $33M Series A, a deep-tech company solving the data-movement bottleneck for AI infrastructure.
Lemurian Labs closed a $28M Series A, a universal platform that works across any hardware and spans compiler technology and runtime orchestration, enabling organizations to write code once and deploy it seamlessly across edge, cloud and on-premise.
Raindrop closed a $15M Seed, an applied AI research company building "Sentry for AI agents" - monitoring infrastructure that catches when AI agents fail silently in production.

Growth Stage:

Harvey closed a $160M Series F, provides a unified and intuitive interface for all legal workflows, allowing lawyers to describe tasks in plain English.
Angle Health closed a $134M Series B, connects thousands of employers and their employees to transparent, affordable healthcare.
Verkada closed a $100M Series F, a cloud-based B2B physical security platform company with six product lines — video security cameras, access control, environmental sensors, alarms, workplace and intercoms — integrated with a single cloud-based software platform.
Axiado closed a $100M Series C, an AI-first company redefining platform security and system management at the silicon level.
Ripple Foods closed a $17M Series E, a leader in developing innovative and delicious dairy-free products.

Questions? Contact: Dawn DeBruyn

Classie AI builds an enterprise platform that gives organizations complete visibility and control across every interaction that happens between users, AI agents, data, and tools. As businesses adopt faster and more autonomous agent workflows, Classie provides a clear operational layer that keeps everything governed and secure.

What They Deliver

Real time detection and onboarding of all AI agents and applications
Full monitoring and audit trails with usage summaries and event notifications
Policy controls that reduce risk and prevent data leaks
Explainability tools that help teams understand how agents behave and why

Why It Matters
AI agents are moving into production across every industry. Teams need a way to supervise these systems, protect sensitive data, and maintain trust while still moving quickly. Classie gives organizations a stable foundation to scale AI without losing control.

Who It Serves
Enterprises that are deploying GenAI agents and need governance, compliance, and safe operations across their entire AI environment.

Learn more at Classie.Ai

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Share the newsletter

Run ads IRL with AdQuick

With AdQuick, you can now easily plan, deploy and measure campaigns just as easily as digital ads, making them a no-brainer to add to your team’s toolbox.

You can learn more at www.AdQuick.com

Valley Recap

🏗️ Imasons: AI runs on software, but it scales on infrastructure ⏱️Two Weeks To Go fo AI INFRA 5! 💵 Bay Area Startups Collectively Secured $3B+ in April MTD

🏗️ AI runs on software, but it scales on infrastructure

Featured Panel: AC to DC: Powering the AI Era

Spotlight: Infrastructure Masons

What iMasons Does (PIE: People, Innovation, Efficiency)

Why It Matters

Who It Serves

Bay Area Startups Collectively Secured $3B in April MTD

Your Feedback Matters!

Logan Lemery Head of Content // Team Ignite

The Biggest Names in Food Are Testing This New Tech

🦞NemoClaw: The Missing Layer Under Agents? 📈Largest Quarter of VC Investments Ever: $216B💰

🦞NemoClaw: The Missing Layer Under Agents?

Most people look at NVIDIA NemoClaw and see another agent framework. That misses the point.

That changes how agents can be used

This is Where AI gets Real.

Highlights from Enterprise AI & HPC Summit NYC

The South Bay is ground zero for AI infrastructure—and we’re turning it into a full week of high-signal events leading into AI INFRA SUMMIT.

Builders Pass — Powered by Micas Networks

CLAWCAMP // APR16

Bay Area Startups Collectively Secured $213B+ In Q1 2026

Your Feedback Matters!

Logan Lemery Head of Content // Team Ignite

Stack BTC while you sleep

🔥 SIGN UP FOR AI INFRASTRUCTURE WEEK. 💰Bay Area Startups Collectively Secured $20B in March MTD

AI infrastructure in 2026 doesn’t respect boundaries anymore. Power decisions hit model performance. Cooling limits shape rack density. Model behavior drives how hardware is actually used. Everything is tied together, and when one layer slips, the rest follow.

SambaNova will be joining us at AI INFRA SUMMIT, bringing its full-stack AI platform built to run large models efficiently in production. By combining purpose-built hardware with an integrated software layer, SambaNova enables a seamless path from model development to real-world deployment.

The South Bay is ground zero for AI infrastructure—and we’re turning it into a full week of high-signal events leading into AI INFRA SUMMIT.

CLAWCAMP // APR16

Bay Area Startups Collectively Secured $7.2B in March MTD

Your Feedback Matters!

Logan Lemery Head of Content // Team Ignite

Market Volatility Exposes Weak Delegation

🔥 From GTC: The Infra Era Has Arrived 💰Bay Area Startups Collectively Secured $7.2B in March MTD

🔥 From GTC: The Conversation Has Moved Below the Model

Plugged in with PLUG AND PLAY

Bay Area Startups Collectively Secured $7.2B in March MTD

Early Stage:

Your Feedback Matters!

Logan Lemery Head of Content // Team Ignite

Your $4T Wealth Multiplier: Secure an 18.8% return

🔥 GTC’26: Where AI Factors Meet Human Chemistry ⚛️ 🤖 Are Bots Your New Users? 💰Bay Area Startups Collectively Secured $5.8B in March MTD

🔥 GTC’26: Where AI Factors Meet Human Chemistry ⚛️

Bay Area Startups Collectively Secured $5.8B in March MTD

Early Stage:

Your Feedback Matters!

Logan Lemery Head of Content // Team Ignite

Here’s how I use Attio to run my day.

🔊Announcing First Wave of AI Infra Summit 5 Speakers 💪🏼 Who's Ready for GTC?! 💰Bay Area Startups Collectively Secured $4.5B in March MTD

READY FOR GTC WEEK?

Bay Area Startups Collectively Secured $4.5B in March MTD

Early Stage:

Your Feedback Matters!

Logan Lemery Head of Content // Team Ignite

Your Tax Data, Finally in One Place

✍️AI Workloads Are Being Rewritten💰Bay Area Startups Collectively Secured $193B

GTC Week is almost here and San Jose is about to become the epicenter of the AI world.

Bay Area Startups Collectively Secured $193B in 2026 YTD

Early Stage:

Your Feedback Matters!

Logan Lemery Head of Content // Team Ignite

Want to get the most out of ChatGPT?

🪫Data Center Capacity Is Disappearing💰Bay Area Startups Collectively Secured $136B MTD in Feb

AI INFRA SUMMIT x PLUG AND PLAY

Bay Area Startups Collectively Secured $163B MTD In February

Early Stage:

Your Feedback Matters!

Logan Lemery Head of Content // Team Ignite

Stop Drowning In AI Information Overload

🔥Heat Is Killing AI 💰Bay Area Startups Collectively Secured $52.5B MTD

The Density Explosion

The Retrofit Crisis

Liquid Cooling as Operational Mandate

The Revised Facility Stack

Thermal Implications for Operators

Upcoming Events

Bay Area Startups Collectively Secured $52.5B MTD In February

Early Stage:

Your Feedback Matters!

Logan Lemery
Head of Content // Team Ignite

Logan Lemery
Head of Content // Team Ignite

Logan Lemery
Head of Content // Team Ignite

Logan Lemery
Head of Content // Team Ignite

Logan Lemery
Head of Content // Team Ignite

Logan Lemery
Head of Content // Team Ignite

Logan Lemery
Head of Content // Team Ignite

Logan Lemery
Head of Content // Team Ignite

Logan Lemery
Head of Content // Team Ignite

Logan Lemery
Head of Content // Team Ignite

Logan Lemery
Head of Content // Team Ignite

Logan Lemery
Head of Content // Team Ignite

Logan Lemery
Head of Content // Team Ignite

Logan Lemery
Head of Content // Team Ignite