Welcome to today’s Midas Report. Your one stop shop for everything important happening within AI.

Here is what is unfolding in AI today, January 27th, 2026….

Our top story of the day kicks off the big race for “physical ai”.

Microsoft just pulled back the curton on Rho-alpha, a new model that doesn’t just talk but rather a vision language action (VLA) model designed for robots operating in the physical world.

This isn't another chatbot tweak.

VLA models translate natural language commands into actual robotic control signals. Tell a robot to stack boxes or sort components, and it perceives the environment, reasons through the task, and executes.

Microsoft is already testing Rho-alpha on dual arm systems and humanoid robots, with early access rolling out soon and broader availability through Microsoft Foundry on the horizon.

The strategic angle here is sharp.

While competitors race to refine digital assistants, Microsoft is planting a flag in embodied AI. That means warehouses, fulfillment centers, manufacturing floors, and lab environments where tasks are messy, unstructured, and very much physical.

Ashley Llorens, who leads Microsoft's Research Accelerator, said VLA models are enabling systems to "perceive, reason, and act with increasing autonomy alongside humans." That's not aspirational fluff. That's a product roadmap.

For Microsoft, Rho-alpha opens a wedge to extend Copilot beyond your inbox and spreadsheets into your supply chain and operations. If they nail this, the same company managing your documents could soon be managing your robots.

What makes this more than vaporware is the infrastructure behind it. Microsoft trained Rho-alpha using Nvidia's Isaac Sim and their own BusyBox benchmark, a testing ground for physical interaction tasks. Future versions will add force sensing and other modalities, making robots more responsive to real world chaos.

If you're building in B2B SaaS, infrastructure, or IoT, start positioning for VLA integrations now.

And if you're an early stage founder, stop thinking apps only. The next wave of MVPs won't just run in the cloud. They'll move boxes, sort inventory, and operate in three dimensions.

Physical AI is no longer theoretical. It's shipping.

Thanks for reading today! See you tomorrow.

MIDAS AI

Today’s signal is clear… more compute, more enterprise pull, more public sector rollout, and louder warnings from the people building the thing.

If you’re a founder, operator, or investor trying to stay sane in AI, today is a good reminder that the market is being shaped by two forces at once. One is industrial scale execution, where infrastructure and distribution deals decide who ships. The other is legitimacy, where safety, governance, and pricing models determine who gets trusted and paid. Here’s what moved.

Anthropic’s CEO turns up the volume on existential risk 

When a frontier lab leader starts sounding less like a product marketer and more like a risk officer, policymakers tend to listen.

Anthropic CEO Dario Amodei issued a stark warning that powerful AI could pose catastrophic risks to humanity in the near term. The specific details in the reporting are light, but the tonal shift is the point. This is not the usual vague “AI is important” rhetoric. It is a high profile lab CEO signaling that the downside case is not theoretical and not far away.

Why it matters is partly political and partly financial. Politically, stronger language from a leading builder provides cover for faster regulation, more aggressive model evaluations, and potentially tighter rules around deployment in sensitive domains. Financially, it changes what “responsible scaling” means for investors and boards. If the people closest to the capability curve are warning about catastrophe, diligence conversations stop being about user growth and start including governance, red team rigor, and incident response plans.

What to watch next is whether this rhetoric translates into concrete asks. Expect pushes for mandatory evaluations, reporting standards, and licensing regimes for the largest models. Also watch how competitors respond. If one lab frames itself as the adult in the room, others either match that posture or risk looking cavalier.

CGI and OpenAI go after the enterprise services crown

The AI services war is getting crowded, and OpenAI just picked another army to send into the field.

CGI announced a global go to market alliance with OpenAI to help clients deploy AI securely and at enterprise scale. The partnership centers on rolling out ChatGPT Enterprise across CGI and using CGI’s Responsible Use of AI Framework, plus an AI literacy program incorporating OpenAI training resources. The message is that this is not a slide deck partnership. CGI plans to equip tens of thousands of consultants and domain experts with the platform, and position itself as “Client Zero” by embedding the tools internally before selling the playbook externally.

This matters because enterprise adoption is increasingly a services led motion. Most large companies do not fail at AI because they lack a model. They fail because they lack workflow redesign, security patterns, data access controls, and change management. CGI, like Accenture, is trying to be the adult supervision that turns pilots into production. OpenAI benefits because distribution through consultancies can be faster than building a massive direct sales and delivery arm in every geography and vertical.

What to watch is differentiation. Every consultancy now claims “responsible” AI and “secure” deployments. The winners will be the ones who productize delivery into repeatable patterns, tie value to measurable outcomes, and avoid turning every engagement into a bespoke science project. Also note the competitive frame: CGI plus OpenAI is an implicit challenge to the Accenture plus Microsoft gravitational pull in big enterprise accounts.

McKinsey says AI is rewriting software pricing  

AI is turning software from a predictable subscription into a meter that never stops running.

McKinsey highlighted a shift in enterprise software pricing as AI introduces variable costs tied to usage. The punchline is that the industry’s long march from perpetual licenses to SaaS is now being followed by an even more consequential move toward consumption based and outcome based models. McKinsey notes that the number of software companies using consumption pricing more than doubled between 2015 and 2024, a trend that AI is accelerating.

The operator takeaway is that unit economics are becoming harder and more important at the same time. When your product includes model calls, agents, and compute heavy features, margin is not guaranteed by default. Vendors will increasingly price in ways that pass through costs or capture value created. Buyers will push back, because procurement teams hate uncertainty and love predictability. Expect a messy middle where contracts include hybrid constructs like base platform fees plus usage tiers, or outcome linked bonuses when measurable value is delivered.

What to watch is who learns to sell the new bill. Companies that can explain cost drivers clearly, provide guardrails, and map usage to business impact will win trust. Companies that surprise customers with runaway usage charges will create churn, even if the product is magical.

Anthropic and the UK government bring Claude to GOV.UK 

Agentic AI is moving from demos to citizen facing infrastructure, and the compliance bar just went up.

Anthropic was selected by the UK Department for Science, Innovation and Technology to help build and pilot an AI powered assistant for GOV.UK, starting with employment related services like finding work, accessing training, and understanding available support. The assistant will be powered by Claude and is described as an agentic system that maintains context across interactions to provide tailored guidance through government processes.

This matters because it is a serious public sector deployment, not a sandbox. Governments have the highest stakes mix of sensitive data, vulnerable users, and accountability requirements. Anthropic says the project will comply with UK data protection laws, give users the ability to manage and opt out of personal data usage, and follow a phased “Scan, Pilot, Scale” rollout. It is also collaborating with the UK AI Safety Institute for testing and evaluation, and embedding engineers alongside civil servants and Government Digital Service teams to build capability inside government.

What to watch is precedent…

If this goes well, it becomes a template for how foundation models get integrated into civic services with explicit safeguards and evaluation regimes. If it goes poorly, it becomes ammunition for regulators arguing that agentic systems are not ready for the public. Either way, the public sector is no longer merely regulating AI. It is becoming a customer, and that will shape standards faster than whitepapers ever could.

The AI trade woke up with two competing impulses today. On one hand, public markets are still addicted to compute and the companies feeding it. On the other, the China risk narrative is back in fashion and it is messing with valuations. Meanwhile in crypto, hype met gravity, again.

📈 AI Market Movers & Trends 📈

Wall Street is heading into mega cap earnings with the kind of nervous optimism that only shows up when everyone is already long. Reuters framed it bluntly. Big Tech earnings are about to test the AI rally, with Alphabet seen as a resurgent leader. Translation for investors. The market wants proof that AI spending is turning into durable revenue, not just bigger capex slides.

The most telling signal of the day was not a single stock candle. It was Nvidia reportedly putting about $2 billion into CoreWeave, the AI infrastructure provider that has become a proxy for the new power stack. Chips are only half the story. The other half is who controls the GPUs once they ship, and how quickly they can be turned into billable compute. Nvidia leaning harder into CoreWeave is a reminder that the AI economy is becoming vertically interlocked. Hardware, cloud capacity, and model builders are increasingly the same ecosystem, just with different logos.

At the same time, the market is still jumpy about China. Barrons revisited how DeepSeek once sent stocks plunging and why Chinese AI remains a live threat. The Wall Street Journal added the sharper edge. AI stocks still face a China risk. That is not just geopolitics. It is competition risk, pricing risk, and acceleration risk. If Chinese labs keep improving models while Western firms keep raising prices to fund compute, the margin math gets ugly fast.

Investors should watch one thing into earnings. Not whether companies say AI fifty times. Watch whether the guidance implies higher utilization of AI infrastructure and clear monetization. The AI rally does not need more excitement. It needs receipts.

💸 Funding Watch 💸 

Private markets kept doing what public markets are debating. They are still paying up for anything that looks like production ready AI, especially when it plugs into enterprise workflows.

The standout deal was Synthesia raising $200 million in a Series E that reportedly doubled its valuation to about $4 billion, with Nvidia and Google among the backers. That investor mix matters. When Nvidia shows up in a late stage round, it is rarely charity. It is distribution, compute alignment, and a bet that the category will drive GPU demand. When Google shows up, it is also a subtle claim on the application layer, where video generation is quickly becoming a default feature inside sales, training, and support stacks.

Sequoia also led a $10 million raise for Pace, an enterprise AI startup colliding with insurance. Small check, big signal. This is the new pattern. VCs are not just chasing foundation models. They are chasing regulated workflow wedges where AI can actually earn. Insurance is a slow moving beast, which is exactly why it pays. If you can automate underwriting, claims triage, or customer intake without blowing up compliance, you get sticky revenue and high switching costs.

On the industrial side, CVector raised $5 million to build what it calls an industrial nervous system. That is the quieter but arguably more durable frontier. Industrial AI is less sexy than chatbots, but it is where ROI is easiest to measure. Downtime avoided. Yield improved. Inventory optimized. The theme across these rounds is clear. The market is moving from AI demos to AI deployment, and capital is following companies that can plug into real operations without rewriting the entire enterprise.

Also worth noting, the broader conversation about seed stage unicorns is getting louder. It is a sign of froth, yes. But it is also a sign that investors believe the winners will compound faster than the old venture pacing models can keep up with. That mindset will keep valuations elevated until something forces discipline.

🪙 Crypto Moves 🪙 

Crypto had its daily reminder that AI narratives can print money and vaporize it just as fast. A fake ClawdBot AI token reportedly hit $16 million in value before crashing about 90 percent. This is not an isolated clown show. It is a structural weakness in the AI token meta. The brand power of AI makes it easy to launch a token, slap on a bot story, and ride retail momentum. But without real distribution, real users, and some verifiable link between token value and product usage, these charts are just leveraged storytelling.

At the more legitimate end of the spectrum, Mesh announced a $75 million Series C and reached a $1 billion valuation to build what it calls a universal crypto payments network. This is not purely an AI story, but it matters for AI builders because payments plumbing is the missing layer for autonomous agents that actually transact. If agents are going to buy data, pay for compute, or execute micro purchases inside apps, the rails need to be seamless. Mesh is chasing that future, and the valuation says investors think the wallet and payments layer is about to be re-rated.

On the exchange side, LBank said it will list apM. New listings are not inherently meaningful, but they keep highlighting the same dynamic. Liquidity arrives faster than legitimacy. Traders should treat most AI labeled micro caps as marketing until proven otherwise.

The connective tissue with today’s equity and funding headlines is compute and distribution. Nvidia is investing where compute gets rented at scale. VCs are funding AI apps that can be deployed into real workflows. Crypto is still trying to figure out how to tie tokens to actual utility without getting drowned in scams. The convergence is real, but the quality filter is brutal. If it does not ship, it will not survive.

📊 Stay tuned for tomorrow’s MarketPulse and sign up to our daily Midas Report newsletter.

Good day and welcome to today’s Midas Report. Your one stop shop for everything important in AI that matters to your bottomline.

Here is what is unfolding in AI today, January 25, 2026….

Our top story of the day brings us a little closer to the mighty “agentic ai”.

It’s just a small $4B bet…

This aspect of AI is exactly what we build in our custom builds that makes them so valuable. It’s not just creating automations… it’s creating automations that examine what it already did and improves upon it.

Recursive Intelligence just hit a 4B valuation on a simple promise that would have sounded like sci fi a year ago. AI that improves AI.

That detail matters because it moves the bottleneck. For most teams, the hard part is no longer getting access to a model. It is making the model behave in your product, stay reliable as users change, and get cheaper without breaking everything.

Recursive AI is basically an attempt to turn that messy, manual tuning work into a machine that learns how to do it better over time. If it works, the pace of improvement stops being tied to your team size and starts being tied to how quickly your system can run feedback loops.

This is where the competitive gap opens.

When development cycles compress, today’s best model becomes tomorrow’s baseline. The winners are not the ones who picked the fanciest model in January. They are the ones already shipping adaptive agents that can keep adjusting in the wild while everyone else is still scheduling quarterly prompt cleanups.

The 4B valuation is also a neon sign about where platform value is expected to land next.

VC money is signaling that “using AI” is getting commoditized, while automating the boring parts of making AI work is where the moat forms. Not the chat interface. The layer that decides which model to use, how to tune it, how to route tasks, and how to keep quality up as costs go down.

Founders should take this personally.

If your team is spending real time on model tuning, evaluation, routing, cost control, or LLM ops firefighting, start mapping what can be offloaded or auto optimized. Even if you do not switch tomorrow, you want clean instrumentation and feedback signals so you can plug into these tools when they mature.

Investors should widen the lens too.

The more interesting bets may be the infrastructure that makes recursive AI usable for midsize builders, not just frontier labs with armies of researchers.

Thanks for joining today. See you tomorrow for more good stuff.

MIDAS AI

Today’s signals say the next enterprise winners will pair governed agentic playbooks with real infrastructure and a clearer story about how work actually changes

If you squint, most AI news falls into two buckets. People building bigger engines, and people figuring out how to steer them without driving into a wall. Today we got both. IBM wants to be the adult in the room for enterprise AI scale. McKinsey is trying to reframe AI from replacement to amplification. Operators are being reminded that shipping “more AI” is not the same as making better decisions. And the infrastructure arms race keeps accelerating, even when the headlines can’t decide whether they are about factories or molecules.

IBM launches Enterprise Advantage to industrialize internal AI playbooks

IBM is packaging consulting, governance, and reusable agent assets into a cloud and model agnostic platform service that looks built for the messy middle of enterprise adoption.

IBM announced IBM Enterprise Advantage, an asset based consulting service designed to help organizations build, govern, and operate internal AI platforms at scale, available January 26, 2026. The positioning is pragmatic. You do not need to change your cloud. You do not need to swap models. You do need a way to turn scattered pilots into something secure, repeatable, and actually shippable across business units.

Under the hood it rides on IBM Consulting Advantage, IBM’s internal AI powered delivery platform that IBM says has supported more than 150 client engagements and delivered up to a 50 percent productivity increase for consultants using it. Enterprise Advantage also includes a marketplace of industry specific AI agents and applications, which is IBM quietly acknowledging what buyers want now: not another model, but a catalog of governed capabilities that can be deployed with less reinvention and fewer compliance surprises.

For founders and investors, this is IBM making a serious bid in the post foundation AI consulting wars where the prize is not “who has the best LLM,” but “who owns the operating system for enterprise change.” Watch for two things next. First, whether IBM can translate marketplace agents into measurable time to value for specific vertical workflows. Second, whether cloud and model neutrality holds up once clients ask for deeper performance tuning, cost controls, and observability across mixed stacks like AWS, Azure, Google Cloud, and watsonx.

AI deployment alone will not win the decision layer will  

A CIO.com argument lands at the right moment because the competitive edge is shifting from rollout volume to decision quality and velocity.

The article’s core warning is simple: enterprises can deploy AI everywhere and still lose if their decision making does not improve. That sounds obvious until you’ve watched a company celebrate “thousands of Copilot seats” while approvals, pricing, hiring, and incident response remain as slow and political as ever.

This matters because the next wave of enterprise software positioning is moving up stack. Infrastructure and tools are table stakes. The durable moat is decision advantage: who can sense changes sooner, recommend better actions, and shorten the loop from signal to action without creating new risk. That is less about model IQ and more about how you instrument the business, define decision rights, and encode policies into workflows that humans actually trust.

What to watch is vendors and internal platform teams racing to own that layer. Expect more talk about decision intelligence, evaluation frameworks tied to business outcomes, and systems that can justify recommendations with traceability rather than vibes. If your product cannot show how it changes a specific decision and improves a measurable KPI, you are competing in the land of features and procurement fatigue.

Enterprise AI is hitting three walls and they are mostly human 

OutSystems CEO Woodson Martin says the real blockers are talent scarcity, confusion about use cases, and the internal narrative required to get adoption.

Martin’s thesis is refreshing because it is operational, not theoretical. First, there is a deep hunger for talent that can build applications with a deep understanding of AI, meaning not just prompt literacy but the mix of domain knowledge, workflow design, and risk awareness that turns models into products. Second, vendor messaging has become so broad that it is creating confusion, collapsing meaningful distinctions between platforms that enable AI and tools that merely attach it.

The third blocker is the most painful and the most real: internal traction. CIOs often need help evangelizing, building a credible story for P and L owners, and proving momentum. Martin even quotes a CIO asking for help positioning OutSystems as an agent platform, with Workbench as a way to get traction with an agentic system. Translation: the buyer is not only purchasing software, they are purchasing political cover and a narrative that makes the business say, “wow, I could dramatically improve productivity.”

For operators, the takeaway is that enablement is becoming part of the product. The winning vendors will not just ship agent builders, they will ship repeatable demos that map to business outcomes, implementation paths that reduce perceived risk, and proof points that help champions sell internally. For founders, this is a go to market reminder. Your buyer needs a story that survives the finance meeting, not just a slick sandbox.

McKinsey says superagency is the workplace model that scales

McKinsey frames the winning enterprise pattern as AI that amplifies employees, with leadership action as the limiting reagent.

McKinsey’s 2025 report on superagency argues that the dominant model is not replacement, it is empowerment, and it backs that with survey data from 3,613 employees and 238 C level executives collected in late 2024. The headline stats are telling. Ninety two percent of companies plan to increase AI investments, yet only 1 percent of leaders call their companies mature in AI deployment. Employees are also three times more likely than leaders estimate to use gen AI for at least 30 percent of daily work, which suggests adoption is already happening, just not through official channels.

The report also highlights the trust and measurement gap. Cybersecurity and inaccuracy top employee concerns, and only 39 percent of companies use benchmarks to evaluate AI systems. Even among those that benchmark, only 17 percent focus on fairness, bias, transparency, and compliance. That is a lot of spend chasing ROI with limited instrumentation, which is a polite way of saying many companies are driving fast with foggy windshields.

What to watch is whether “superagency” becomes a practical operating model rather than a keynote theme. McKinsey points toward executive alignment, federated governance, workforce training, and modular architecture as prerequisites to scale. For founders selling into enterprise, the implication is clear: products that slot into human workflows, provide explainability, and support governance will scale faster than tools that assume the organization will reorganize itself around your UI.

Nvidia and CoreWeave expand the AI factory narrative but the real signal is verticalized compute

Even when the details blur, the direction is consistent: specialized infrastructure and tight hardware software integration are becoming competitive weapons.

The stated theme is an expanded collaboration to scale AI factories, meaning data centers purpose built for accelerated AI computation. The strategic implication is straightforward. Demand for AI workloads keeps rising, and the winners will be the ones who can deliver capacity with predictable performance, cost, and deployment speed.

The provided source details also highlight NVIDIA BioNeMo adoption in life sciences, including companies like Amgen and AstraZeneca, running on NVIDIA DGX Cloud. That may read like a different story, but it reinforces the same infrastructure point: NVIDIA is not just selling chips, it is selling vertically integrated stacks where the platform, tooling, and compute come bundled into something closer to an outcome. In pharma, that outcome is faster model development and inference for drug discovery workflows.

For investors, the thread to pull is how “AI factory” providers like CoreWeave change the hyperscaler calculus. Vertical integration and workload specific optimization can win share when speed to capacity matters more than vendor consolidation. For founders, this is a procurement reality check. Your inference cost curve, latency, and security posture may depend less on model choice and more on where and how you run it.

The next advantage is not just intelligence. It is industrial grade delivery of it.

Today felt like the AI economy snapping back into focus.

Public markets rotated toward proven cash engines, startups priced the next wave of enterprise adoption, and crypto stayed oddly quiet, which is its own kind of signal. Nvidia kept the crown, Cisco caught a bid, and Synthesia’s new valuation screamed one message. Real revenue is back in fashion.

📈 AI Market Movers & Trends 📈

Nvidia was the day’s cleanest expression of the current AI trade. Not hype, not vibes, not “one day we will monetize.” Just brutal demand pull through.

Zacks tagged Nvidia as Bull of the Day, and the subtext is what matters. The market is rewarding companies that sit on the critical choke points of AI compute and infrastructure, even as investors debate whether this is a bubble or simply a long cycle with periodic shakeouts.

That bubble conversation is getting sharper. VanEck’s CEO basically waved away the “waiting for the bubble to burst” crowd by arguing the sector already purged the weakest names. That framing matters because it is exactly how money managers justify re risk sizing into leaders while avoiding the long tail. Translation. The garbage already got taken out, so the survivors get premium multiples again.

The tension is timing. FXEmpire called this a make or break week for the AI trade, which tracks with what you can feel in positioning. Bulls want confirmation from earnings, guidance, and macro prints. Bears want a single crack in demand to turn “AI infrastructure cycle” into “capex hangover.” This is why Nvidia strength matters beyond Nvidia. If the king looks stable, the whole court breathes easier.

Meanwhile Cisco popped on an upgrade, with Barron’s framing it as a big cap bargain tied to AI networking and enterprise plumbing. This is the grown up rotation. When investors start treating networking and pipes as AI plays, it signals belief that inference and deployment are spreading beyond a handful of hyperscalers.

One more thread worth watching is the “fallen angel” angle. Finance headlines flagged at least one AI stock down 14 percent to start 2026 that now looks interesting. That’s a reminder that the market is done paying infinite prices, but it’s still willing to recycle quality names once valuations stop being ridiculous.

💸 Funding Watch 💸 

If Nvidia is the public market anchor, Synthesia was the private market loudspeaker. The UK AI avatar company raised a massive round that priced it at about 4 billion, with coverage converging around a roughly 400 million raise. Nvidia’s venture arm participated, and Alphabet’s VC arm was also in the mix, according to CNBC and others. That combo is not just a trophy cap table. It is a distribution thesis. Nvidia wants workloads, and Google wants enterprise AI adoption that keeps cloud demand sticky. Synthesia sits right in the monetizable middle, corporate video, training, marketing, internal comms, the stuff that actually has budgets.

The most important detail is not the valuation headline. It’s that Synthesia is effectively selling an enterprise wedge product that makes AI visible to non technical buyers. AI avatars are not a toy when the buyer is HR, sales enablement, or customer support training at scale. This is where “AI transformation” stops being a slide deck and becomes a procurement line item.

Synthesia also signals a broader pattern in 2026 funding. The market is paying for applied AI with clear unit economics, not foundational model moonshots without distribution. Even the optics matter. A British company nearly doubling valuation to 4 billion is a message to European founders that the exit ramp is not only in Silicon Valley, as long as you show traction.

On the other end of the map, Chinese AI firm Stepfun reportedly raised about 719 million to push foundation models and terminal agents. That is the parallel universe trade. The US and UK are funding applications and enterprise adoption at premium prices, while China is still writing enormous checks for core model capability and agent deployment. Investors should read that as competitive pressure on model commoditization. If more capital flows into base models globally, the edge shifts to whoever owns distribution, data, and enterprise relationships.

Emergent’s 70 million Series B also stood out, especially with the claim that ARR hit 50 million in seven months. If that number holds up, it’s another neon sign that buyers will pay fast when the product is tied to productivity or automation outcomes, not “AI for AI’s sake.”

🪙 Crypto Moves 🪙

Crypto was the quiet room today, and that’s not an accident. When AI equities and AI venture rounds are hot, AI tokens need a sharper story than “decentralized compute someday.” The lack of major AI token headlines is a kind of market verdict. Traders are still skeptical that most AI tokens capture real cash flows, and in 2026 that skepticism is getting healthier.

But do not confuse quiet with irrelevant. The convergence is still happening, it’s just shifting from speculation to infrastructure decisions. Nvidia’s dominance in compute and the global push into agents, like Stepfun’s stated focus on terminal agents, highlight the same bottleneck that crypto has been trying to monetize for years. Access to compute, access to data, and coordination of workloads across participants. The difference is that the public market is paying Nvidia for it right now, while token markets are still trying to prove they can deliver comparable reliability, pricing, and performance.

Here’s the connective tissue founders and builders should notice. When Nvidia’s VC arm backs a company like Synthesia, it is not charity and it is not branding. It is demand cultivation. Nvidia wants more real workloads that justify the next generation of GPUs. If crypto based AI networks want relevance, they need to plug into that reality by supporting real inference demand, verifiable service quality, and enterprise friendly compliance. Otherwise they remain a trading venue, not an economic layer.

For traders, the implication is blunt. In periods where public AI leaders are strong and private funding is rewarding revenue, AI tokens need concrete catalysts like meaningful integrations, clear fee models, or enterprise adoption signals to outperform. Without that, they drift, or they pump briefly on headlines and then fade as capital rotates back to equities and late stage venture where the cash flow story is clearer.

Crypto’s AI moment will return, but it is going to look more like plumbing and less like memes. Today was another reminder that the market is done paying for narratives that cannot invoice someone.

📊 Stay tuned for tomorrow’s MarketPulse and sign up to our daily Midas Report newsletter.

Good day and welcome to the Midas Report!

Here is what is unfolding in AI today, January 23, 2026….

South Korea just did what most governments have been circling for years and actually shipped a full set of AI laws.

That matters because it turns AI regulation from a vague future risk into a real operating constraint, with a timeline and consequences that teams will have to design around.

For founders, the bigger signal is not just Korea. It is the direction of travel. Once one major economy sets a comprehensive framework, everyone else gets a template, and the global market starts to fragment into rule sets that look similar in spirit but different in the details.

If you are building for global scale, you cannot treat compliance like a late stage checkbox. You are effectively choosing whether your product will be easy to sell across borders or constantly stuck in the custom build penalty box.

There is also a quiet sales upside hiding in the paperwork.

In APAC, especially in fintech, health, and government adjacent sectors, enterprise buyers do not just ask if your model works. They ask if it is defensible, auditable, and safe to deploy without making them the headline.

Early compliance is how you turn trust into a feature. It shortens security reviews, reduces procurement friction, and helps a buyer say yes without betting their career on your roadmap.

So what do you do this week, not this quarter.

Audit where your AI stack touches user data, where it makes or influences automated decisions, and where you would struggle to explain why a model produced an output.

Those are no longer just product questions. They are legal risk zones now, and Korea will not be the recent jurisdiction to spell them out.

Treat this like you would treat payments compliance. Build the muscle early, document the choices, and make transparency and controls part of the product… not a scramble after your first big customer asks hard questions.

Investors should read the same signal. Startups that are fluent in global compliance are becoming more fundable assets, because in regulated markets, speed without trust is just churn with better branding.

Thats us for the day. Thanks for reading.

MIDAS AI

Enterprise AI is entering its next phase, and it looks a lot less like “innovation theater” and a lot more like supply chains, procurement, and operating model rewrites.

Today’s stories draw a clean line through the market. The old winners sold advice and rented compute. The new winners bundle execution, infrastructure, and distribution into something sticky enough to survive the CFO.

Accenture, EY, and a newly consolidated Lightning AI are all making the same bet from different angles. AI is no longer a feature. It is the product, the delivery mechanism, and the moat.

Accenture makes a 1B wager that implementation is the real bottleneck

Accenture is paying up to buy applied AI muscle because the hardest part of AI is not the model, it is getting it into production.

Accenture is acquiring an applied AI organization in a deal pegged at around $1 billion, signaling that the consulting giant sees enterprise AI as a scale game, not a boutique craft. Even without a lot of verified deal specifics in the source writeup, the strategic intent is clear. Accenture wants more repeatable delivery capacity across data, systems integration, governance, and change management, the unglamorous stuff that turns pilots into operating leverage.

Why this matters to founders and operators is simple. When Accenture shifts spend from “we can build it” to “we must buy it,” it is admitting that the market has moved from experimentation to industrialization. Large enterprises are increasingly asking for outcomes with timelines, not slide decks with frameworks. If you sell into the enterprise, expect services partners to be both your accelerant and your competitor, especially if your product still requires heavy customization to land.

What to watch next is how aggressively Accenture productizes the acquisition. The fastest path to margin is turning bespoke AI integration into reusable patterns, reference architectures, and managed services. If you are building in this ecosystem, the opportunity is to be the component they standardize on. The risk is being the thing they decide to replicate in house once they have enough delivery reps trained on your category.

EY.ai positions the Big Four as AI operating system vendors 

EY is bundling consulting, risk, and tax into a unified AI platform because AI is becoming a structural service line, not a team of specialists.

EY is pushing its EY.ai platform as a unified approach that cuts across business functions, including consulting, risk, and tax. The important signal is less about the branding and more about the organizational shape. The Big Four are trying to turn AI into a coherent delivery stack with shared tools, shared governance, and shared commercialization, instead of scattering “AI centers of excellence” across practices.

For enterprise buyers, that is a promise of consistency. One platform, one playbook, fewer handoffs between advisory and implementation. For founders, it means procurement dynamics are shifting. When a firm like EY sells an “AI transformation” motion, it will prefer toolchains that plug into its platform narrative. You may be evaluated not just on model quality or features, but on how cleanly you support auditability, controls, documentation, and regulatory posture.

What to watch is whether EY.ai becomes a real internal platform with reusable assets or a portfolio label for multiple tools. The winners in this category will be the firms that can demonstrate measurable deployment velocity and risk containment, especially in regulated sectors where AI adoption is constrained less by capability and more by accountability.

The circular deals behind the AI gold rush are becoming a lockout mechanism

Partnerships between chipmakers and cloud titans are creating a flywheel that concentrates advantage and makes “choice” feel optional.

Bloomberg’s breakdown of circular deals in AI maps a reality most operators feel in procurement but rarely articulate. Compute supply, model availability, cloud distribution, and enterprise contracts are increasingly intertwined. When a cloud platform locks up access to the newest GPUs, it attracts the hottest workloads. Those workloads attract more enterprise customers. Those customers justify more capacity reservations. And around it goes, with smaller players paying more, waiting longer, or settling for second best.

This matters because the AI stack is not just competitive. It is interdependent. Builders choose a model based on availability and cost, then discover their inference bill is really a cloud dependency. Enterprises choose a cloud based on perceived safety and procurement ease, then learn their AI roadmap is tied to a vendor’s hardware deals. Investors should read this as a moat story. The defensibility is not always product. It is access.

The practical takeaway is to model vendor risk like you model interest rates. Who controls your marginal compute supply in a shortage? How exposed are you to a single hardware generation or a single cloud’s quota policy? If you are an early stage company, the goal is to keep portability credible until you have leverage. Portability is not a technical virtue. It is negotiating power.

Lightning AI and Voltage Park merge to sell the full stack, not just the shovels  

The new Lightning AI wants to be the integrated AI cloud, combining software and GPU infrastructure so customers stop duct taping their toolchain together.

In a move that fits perfectly inside the “circular deals” era, Lightning AI and Voltage Park have merged into a single company now named Lightning AI. The pitch is vertical integration. Instead of selling raw GPU capacity like many neoclouds or depending on third party clouds like many AI platforms, the combined company says it is software first and infrastructure native, designed end to end for AI workloads.

The numbers attached to the story are bold. The company claims more than 400,000 users, annual recurring revenue above $500 million, and a valuation north of $2.5 billion. On infrastructure, it points to over 35,000 Nvidia GPUs available, including H100, B200, and GB300, spread across six US data centers. The strategic framing is also clear. William Falcon, its founder and CEO, compares today’s AI tooling to carrying separate gadgets instead of using an iPhone, a tidy metaphor for bundling that every platform company loves because it is usually right.

For founders and operators, this is a credible alternative pattern to hyperscalers and to “GPU resellers with a dashboard.” If Lightning AI can genuinely reduce friction across training, deployment, and production operations while keeping pricing predictable, it can win teams who are tired of stitching together a model host, a data pipeline, an orchestrator, and a cloud billing maze. The detail to watch is whether the integrated experience stays clean as enterprise requirements pile up, and whether customers can actually maintain leverage and portability even as they adopt more of the stack.

The deeper implication is consolidation pressure. If full stack AI clouds become the expectation, point solutions in MLOps and infrastructure tooling will need to either specialize sharply or integrate tightly. The middle ground, a tool that does a little of everything but owns nothing, is about to get very uncomfortable.

That’s the pattern for the day. Consulting is buying execution. Advisory is turning into platforms. Infrastructure is consolidating into vertically integrated stacks. And the incumbents are wiring deals that make the flywheel spin faster for them than for everyone else. The only sane response is to build with optionality until you can build with power.

Today was a clean reminder that AI is not a single trade, it is a stack. Intel showed what happens when you promise an AI comeback and deliver a soft forecast. Meanwhile TSMC and ASML kept looking like the toll collectors of the entire AI arms race. In crypto, listings and presale hype tried to front run real adoption.

📈 AI Market Movers & Trends 📈 

Intel’s stock took the hit after a cautious outlook underscored the uncomfortable truth. Not every legacy chip giant gets to print money just because the word AI is in the earnings deck. The market is rewarding execution and capacity, not ambition. Intel’s stumble matters because it highlights where the AI value pool is concentrating. The winners are the companies that already sit inside the buildout, shipping the picks and shovels that hyperscalers actually need this quarter, not next decade.

That is why TSMC keeps getting framed as the highest conviction AI play in public markets. In an AI arms race, the foundry wins because everyone shows up at the same factory door. Nvidia, AMD, Apple, and the next wave of custom silicon builders do not scale without leading edge nodes and packaging capacity, and TSMC is the choke point. Investors are basically paying for the privilege of owning the infrastructure that the entire model economy runs on.

ASML sits one step upstream and the setup is similar. A rich valuation only looks rich until you remember there is no serious leading edge manufacturing ramp without EUV tools. Demand can wobble around the edges, but the strategic direction is one way. More compute, more wafers, more tools. The broader bubble chatter is getting louder again, with dot com comparisons floating around, but the key difference is that AI has a visible revenue engine already. Cloud spend and enterprise deployments are real, even if some multiples are not.

💸 Funding Watch 💸 

The startup tape is sending a blunt signal. The market is splitting into two buckets.

There are companies raising enormous rounds because they sit close to distribution or infrastructure, and everyone else fighting for oxygen with smaller checks and tougher terms. Lists of 2025 mega rounds show a recurring pattern. Investors were willing to write nine figure tickets for teams building model tooling, developer platforms, data infrastructure, and vertical AI products with clear ROI paths. The theme is not novelty. It is adoption.

The Replit founder’s ongoing messaging in the ecosystem captures the mood. AI is turning product development into a speed contest, and the winners are the ones that can ship faster, onboard users faster, and compound iteration. That sounds inspiring until you look at the other side of the trade. Talent markets are getting sharper, with more people effectively locked into roles while others get cut loose. It is not just cost trimming. It is companies reallocating toward AI native output and away from work that does not scale.

One underappreciated thread is how deeply AI is reshaping financial plumbing. AI driven lending and capital markets workflows are moving from experimental to operational, because when margins compress, automation stops being a nice to have. That links directly back to the public market winners. When enterprises industrialize AI, they buy compute, they buy tooling, and they buy the chips that make it all possible. The startup world is racing to capture the workflow layer, while the public market supply chain keeps collecting rent.

🪙 Crypto Moves 🪙 

Crypto AI was in its usual split personality mode today. On one side you have legit exchange activity where listings act like jet fuel. The ELSA listing on Upbit drew attention because Upbit liquidity can turn a quiet token into a full blown momentum trade overnight. Listings are not fundamentals, but they are a distribution event, and traders treat them like an earnings surprise. On the other side you have the presale machine pushing another wave of AI branded tokens, with DeepSnitch AI grabbing headlines and the usual narrative that smart money is loading up. That may move price, but it does not automatically create a product.

Sentient also popped up again in the prediction industrial complex, with multi year price forecasts making the rounds. This is where builders and investors should stay disciplined. Price predictions are entertainment unless they tie back to usage, revenue, or at least measurable network demand. The real signal is whether these projects are integrating into developer workflows, compute marketplaces, inference networks, or data pipelines that actual teams rely on.

The connective tissue to today’s equity story is simple. The market is paying for infrastructure. In stocks, it is fabs and lithography. In crypto, the only AI tokens that deserve durable attention are the ones that create credible infrastructure, not just a ticker. The next leg of crypto AI will not be won by louder marketing. It will be won by projects that make inference cheaper, data more available, or coordination easier in a way Web2 cannot easily replicate.

📊 Stay tuned for tomorrow’s MarketPulse and sign up to our daily Midas Report newsletter.

Welcome to today’s Midas Report! We keep you up to date on what’s happening in AI that matters to your bottom line.

Here is what is unfolding in AI today, January 22, 2026….

All eyes are on Claude's 'constitution layer'…

Anthropic just rewrote Claude’s constitution, and the timing tells you who the real audience is.

This new document launched January 21, 2026, replacing the original 2023 version, and it reads less like a developer note and more like an enterprise buyer’s comfort blanket.

The pitch is transparency, but not the hand wavy kind.

The constitution now uses a four tier priority system that puts safety, ethics, compliance, and helpfulness in order, so when Claude hits a messy edge case, it has a way to justify what it does instead of shrugging and guessing.

That matters because enterprise deployments do not fail in the obvious situations.

They fail in the weird ones, when a customer request collides with policy, regulation, brand risk, or a human being having a bad day.

Gartner analyst Arun Chandrasekaran basically says the quiet part out loud… the goal is judgment in unforeseen situations, and unexpected scenarios are inevitable in enterprise settings.

This is the real shift in AI buying.

Companies are less impressed by a flashy demo and more interested in whether the system can behave predictably when the environment gets unpredictable.

Or as Futurum’s Bradley Shimmin frames it, this is closer to philosophy and ethics than pure engineering, because alignment and trust are now product features, not side quests.

Zoom out and you can see the market hardening.

Open source heavyweights like IBM, Nvidia, Meta, and AI2 are pushing transparency through training data and methodology disclosures, while Anthropic is packaging transparency as a usable interface for decision making.

Different routes, same destination… trust plus performance is becoming the infrastructure layer that decides who wins regulated and high stakes deals.

If you build AI tools for organizations, take the hint.

Document behavior, not just results, and write it like your compliance lead, customer success manager, and legal counsel will all read it in one sitting.

The second order effects are coming next, too.

If constitutions become table stakes, expect audits, benchmarks, and governance layers to turn into the next battleground for differentiation.

Prompt of the Day

Once again, I find myself playing with Nano Banana… it’s just that good.

Today, we do a food nutritional breakdown with infographic style… all from one prompt. Worked well as you can see below.

This concept could work for nearly any niche on social media with some creativity.

Here’s the prompt used:

“Ultra-clean top-down (90° flat lay) infographic composition with playful candy aesthetic. A perfectly centered white ceramic plate sits on a seamless light background. On the plate: chocolate covered donut, shown in a precise cross-section or layered cutaway, revealing internal ingredients, textures, and structure with extreme clarity. PRIMARY FOCAL POINT (HERO ELEMENT): Above the plate, occupying a large portion of the upper canvas, place an oversized floating infographic bubble displaying: TOTAL CALORIES (research nutritional info for single chocolate covered donut) The calorie bubble features a playful candy design with bright colorful gradient background (pink to purple to yellow), candy sparkle effects, confetti elements, and fun playful typography with rounded shapes and bright colors. This is the first thing the eye sees. SECONDARY ELEMENTS: Surrounding plate, fill the remaining space with a dense, high-information infographic system featuring: – Playful candy style charts (protein, carbs, fats) with colorful data visualization and fun shapes – Playful percentage bars with candy patterns and textures – Candy-inspired icons with cute and fun designs – Playful temperature and time displays with colorful numbers and fun fonts – Measurement scales with candy grid patterns and colorful borders – Micro-diagrams with candy elements and fun details Use thin playful connectors with colorful lines, fun arrows with cute designs, candy pictograms, and radial playful charts clearly linking data to specific food components. Visual hierarchy rules: Calorie bubble = dominant Plate + food = secondary anchor Surrounding data = tertiary but dense and readable Design language: premium playful infographic, Apple UI × candy design, extremely clean, structured, and modern with fun and colorful elements. Color palette: natural food colors, monochrome playful UI with bright colorful accent colors (pink, purple, yellow, blue) for the calorie bubble and playful elements. Lighting: soft studio lighting with candy sparkle glow effects, zero clutter, flat background for maximum contrast. Output: 4:5, ultra-crisp, social-feed optimized, no watermark.”

Thanks for reading today’s Midas Report. We will see you tomorrow with what matters to you.

MIDAS AI

Today’s signal is loud and oddly consistent across five very different stories…

The AI market is maturing into two real arenas founders can actually build around. One is interface and trust, where voice is turning into the default UI and “emotional intelligence” is the new battleground. The other is infrastructure and operations, where the winners are the teams who can turn models into measurable business value without getting buried by integration debt and compliance.

Here’s what moved, why it matters, and what to watch.

Google hires Hume AI CEO as emotional voice becomes a product wedge 

Google DeepMind just made “emotionally aware” voice a first class feature, not a research curiosity.

Google DeepMind signed a licensing deal with Hume AI and is bringing over Hume CEO Alan Cowen plus roughly seven engineers. It is not a formal acquisition, which matters in 2026 because regulators are watching talent deals that look like acquisitions wearing a trench coat. The FTC has said it plans to examine these “aqui hire” structures, and Google has already been here before with its reported $3 billion Character.ai licensing arrangement.

Hume’s core is emotionally intelligent voice interfaces. They train models using expert annotation of emotional cues in real conversations, then use those models to detect emotion in a user’s voice and generate more realistic, adaptive voice experiences. That capability is exactly where voice agents are headed. You can have a brilliant model that still feels useless if it cannot read the room, especially in customer support, healthcare, education, and any workflow where “helpful” is more than just factual accuracy.

The near term watch item is integration. DeepMind plans to fold Hume’s tech into its frontier models, which likely means Gemini gets better at voice and prosody in ways that are hard to copy with prompt engineering alone. Also note the timing with Google’s multiyear partnership to integrate Gemini into Siri. If voice becomes the primary interface, as Hume’s incoming CEO Andrew Ettinger predicts, then emotional adaptation becomes a differentiator you can feel in five seconds.

Jensen Huang says AI is driving the largest infrastructure buildout in history 

Nvidia is selling a narrative that conveniently aligns with its business, but the numbers are still sobering.

Nvidia CEO Jensen Huang says the global AI expansion is fueling what he called the largest infrastructure buildout in human history, citing projections of $85 trillion in AI related infrastructure investment over the next 15 years. His framing is that AI is a layer cake starting with energy, then chips, then data center infrastructure, then cloud infrastructure, land, and power. If you are a founder building “just software,” this is your reminder that software margins still ride on very physical constraints.

He pointed to the H200 as newer, more efficient, and more affordable hardware, while emphasizing a steady march toward higher energy efficiency and lower token cost each year. That combination matters because it resets the feasible set of products. Every big drop in token cost unlocks a new tier of workflow automation that used to be uneconomical, particularly for long context enterprise use cases and always on voice agents.

The geopolitical layer is not subtle. Huang referenced export approvals under the Trump administration and argued that Nvidia cannot “concede any market,” while also saying China’s military does not rely on H200 chips because they build on their own chips. Investors should hear two things at once. Demand is real and structural, and policy risk is now part of your cap table whether you like it or not.

Enterprise AI adoption moves from pilots to profits 

The playbook is shifting from “try some tools” to “build internal systems that people actually use.”

Bain Capital Ventures gathered operators from Ford, Mammoth Brands, and Box, and the through line is that 2025 became a turning point for enterprise AI value realization. Ford’s internal platform Ford LLM has about 50,000 weekly active users, which is the kind of adoption number that stops being innovation theater and starts being operating system.

What is notable is where the value is landing. Ford’s “AI Big Bets” include a supply chain risk assist using a multi agent system, generative tooling that turns sketches into 2D and 3D renderings, and proprietary models that simulate aerodynamic drag, cutting computation from 16 to 18 hours down to seconds. That last one is a reminder that “AI” is not only chat. Sometimes it is an internal model that simply compresses time, and time is still the most universal KPI.

Box and Mammoth show the organizational side. Box ran a hackathon after giving employees early access to Box Agents for unstructured data, then backed it with a company wide AI certification course. Mammoth piloted over 30 AI tools in 2025, and two thirds of those vendors had fewer than 100 employees, which is good news for startups who can clear security and integration hurdles. The watch item is procurement. Ben Kus at Box said the normal rules around procuring software are no longer relevant. Translation for founders is simple. Win fast, integrate cleanly, and expect constant re evaluation.

Leaders expect major AI revenue boost by 2030 

Executives are bullish on AI revenue, but most cannot explain where it will come from.

IBM’s Institute for Business Value surveyed 2,000 C suite executives globally and found that 79 percent believe AI will significantly contribute to enterprise revenue by 2030. Only 40 percent say it is currently boosting revenues, so leaders are effectively budgeting for a future that has not fully arrived yet. That gap is opportunity, but it is also risk.

The most revealing stat is the one nobody wants to lead with. Only 24 percent can clearly identify where AI driven revenue will come from, and 68 percent worry integration issues will cause AI efforts to fail. This is why “AI strategy” is quietly becoming “systems integration with a model attached.” Mohamad Ali at IBM Consulting said AI will define businesses, and Morgan Stanley CEO Ted Pick warned there will be teething pain. Both can be true, and usually are.

Capital is following the expectation curve. Gartner projects global AI spend rising 44 percent year over year in 2026 to $2.52 trillion. Spending focus is expected to move from productivity and efficiency today toward product and service innovation from 2026 to 2030. Founders should interpret that as a roadmap. Sell efficiency now to get in the door, but design for revenue features that can graduate into the product.

AI in finance scales fast and agentic systems start to show up in production 

Banks are not “experimenting” anymore, which should terrify and delight vendors in equal measure.

Nvidia’s sixth annual State of AI in Financial Services report surveyed more than 800 professionals and found nearly 100 percent plan to increase or maintain AI budgets next year. Active AI use jumped to 65 percent from 45 percent, and 61 percent are using or assessing generative AI. Finance is compliance heavy and reputation sensitive, so this kind of acceleration is a strong signal that the tooling and governance stack is finally usable.

The business outcomes being reported are unusually concrete. 89 percent said AI is increasing annual revenue and decreasing annual costs, with 64 percent reporting revenue up more than 5 percent and 61 percent reporting costs down more than 5 percent. In finance, measurement is the product. If you improve authorization rates by basis points, the P and L notices.

Two additional shifts matter for builders. Open source is now central, with 84 percent calling it important to their strategy, largely because institutions want to fine tune on proprietary transaction and customer data. And agentic AI is moving from conference slide to deployment. 42 percent are using or assessing agents, and 21 percent have already deployed them. The watch item is governance and auditability. The bank that lets an agent act without a tight control plane is the bank that will fund your competitor’s postmortem.

If you are building this week, the map is getting clearer. Voice is becoming the front door, infrastructure is becoming the tax, and enterprise value is increasingly gated by integration, enablement, and the ability to prove ROI without pretending everything is “AI transformation.” The hype cycle is still here, but the buying behavior is finally growing up.

Today felt like a power shift, not a headline grab. Public markets rotated within AI semis, mega seed rounds kept getting bigger and weirder, and crypto reminded everyone that AI hype plus thin liquidity is still a dangerous cocktail. The throughline was simple. Distribution wins and the rest is just branding.

📈 AI Market Movers & Trends 📈  

Alibaba stock popped after it laid out plans to rival Nvidia on AI infrastructure, leaning into in house chips and a broader stack rather than just selling cloud capacity. The market liked the ambition. It also liked the timing. US export controls have turned self sufficiency into a forced innovation program for China, and Alibaba is one of the few with enough data, capital, and cloud demand to justify custom silicon at scale.

Meanwhile Broadcom kept doing what Broadcom does. It wins quietly, invoices loudly. Reports pegged its AI related revenue trajectory with expectations pushing toward a 50 billion run rate over time as hyperscalers keep designing custom accelerators and networking becomes the real bottleneck. Nvidia still owns the mindshare, but the supply chain winners are spreading out. When every model run needs more bandwidth, more switching, more interconnect, the “shovels” are not only GPUs. They are the plumbing.

This is why the hidden winner narrative around Nvidia’s boom keeps resurfacing. The best positioned names are often the ones selling picks, power, and pipes into the same gold rush. Investors are starting to price that in, and it shows up in how the market is treating the broader semiconductor complex rather than only Nvidia. Bulls did not abandon the GPU king. They just stopped pretending the ecosystem has a single profit pool.

One more tell came out of China. Moore Threads reportedly tripled revenue as domestic demand leans harder into local AI compute. That is not a Nvidia killer story. It is a volume and sovereignty story. And volume has a habit of becoming competence faster than incumbents expect.

💸 Funding Watch 💸 

If you want to understand 2026 venture, look at one number. Humans& reportedly raised a 480 million seed round at a 4.8 billion valuation. Seed. Not Series D. Seed. That is either peak confidence or peak narrative premium depending on how cynical you are feeling today.

The pitch is “human centric AI tools” which sounds soft until you remember where budgets are going. Enterprises are not only buying models. They are buying adoption. Workflow change, trust layers, and products that reduce risk for humans who have to live with the outputs. A giant seed round here is a bet that interface and governance will matter as much as raw model quality. Investors are effectively saying the next moat is not just intelligence. It is integration plus credibility.

On the more grounded end, AppliedAI raised a pre Series B to scale an enterprise AI platform. That is the classic signal that the market is still paying for implementation reality, not only frontier dreams. In the background, Genspark reportedly hit 100 million in ARR and is weighing new funding. That is a sharp reminder that revenue is back in fashion, even in generative AI, and buyers are willing to pay when the product is tied to outcomes rather than demos.

Then there is OpenEvidence, which reportedly doubled its valuation to 12 billion in a new round. Healthcare continues to attract premium pricing because the data is sticky, the value of correct answers is huge, and the switching costs are real. The pattern across these deals is consistent. The big checks are going to companies that sit close to regulated or operationally critical workflows. Less “look what the model can do.” More “here is where it ships and why it stays.”

🪙 Crypto Moves 🪙 

Crypto’s AI corner delivered its usual split personality. On one side, traders chased the latest AI coin narratives, with DeepSnitch AI getting attention across major exchange commentary and hype driven prediction pieces. When you see “100x countdown” language, you are not looking at fundamentals. You are looking at liquidity games, attention arbitrage, and retail adrenaline. That does not mean there is no opportunity. It means your risk model should be brutally honest.

On the other side, the more credible convergence story kept building around Ethereum. Reporting highlighted why BlackRock remains bullish on Ethereum in 2026 even with price action that has not matched the grand narratives. The institutional argument is not that ETH is a meme. It is that tokenized finance, stablecoins, and onchain settlement keep creeping into real infrastructure. AI slots into that because agents need rails. Payments, identity, provenance, and audit trails are not optional when software starts acting on behalf of humans and businesses.

Then Binance reminded everyone that exchanges can and will pull the plug. A delist shakeup removed a batch of spot trading pairs including AIBTC, a move that matters less for the specific tickers and more for what it signals. Distribution is fragile. If your token depends on a couple of venues and a hype loop, it is not an ecosystem. It is a house of cards with a logo.

The darkest AI crypto link came from the real economy. AI enabled impersonation scams reportedly drove a record 17 billion in losses. That is not bullish for any token, but it is bullish for one category. Verification. Identity. Forensics. The next “AI x crypto” winner might not be compute or inference marketplaces. It might be the boring trust layer that keeps money from walking out the door.

📊 Stay tuned for tomorrow’s MarketPulse and sign up to our daily Midas Report newsletter.

Good morning and welcome back to today’s Midas Report. Your one stop shop for everything you need to know that’s happening in AI today that matters to your bottom line.

Here is what is unfolding in AI today, January 20, 2026….

CGI just made a quiet but meaningful bet that agentic AI is ready to leave the lab.

The company expanded its global alliance with Google Cloud around Gemini Enterprise, with a multi year plan to equip tens of thousands of CGI consultants with it. That is not a pilot budget. That is a staffing decision.

Agentic AI has been easy to talk about and hard to land.

The promise is simple… Instead of AI that answers questions, you get AI that can actually do the work across systems and teams.

Google Cloud even called Gemini Enterprise a front door to AI that can orchestrate complex work across entire organizations, which is a polite way of saying it is meant to run real workflows, not just write nice drafts.

CGI’s CTO put the stakes plainly. This is about accelerating the delivery of outcomes with Agentic AI, not just pilots, and clients want it to be enterprise grade, secure, reliable, and measurable. That last word is doing a lot of work.

This is the tell for the broader market. When a major IT services firm moves from experimentation to embedding agent tech into managed IT and business services, it signals that buyers are about to raise the bar.

The question is shifting from can it do things to did it change anything.

IT buyers are not going to fund a fleet of shiny agents that create more tabs, more alerts, and more meetings. They will fund fewer incidents, faster cycle times, cleaner handoffs, and lower cost per transaction. They will ask for proof in metrics, not screenshots.

The other tell is how CGI is structuring the rollout. This is not just tooling. It includes co investment in go to market, plus workshops, training, hackathons, and scaled delivery through CGI’s global network. In other words, the product is the service design.

If you are building in enterprise AI, tune your pitch accordingly. Anchor on one or two org wide workflows you can measurably improve, define the before and after, and make security and reliability part of the story from day one.

Fancy features are nice. Measurable outcomes are what gets deployed.

Prompt of the Day

Today, we start playing with infographics. I love NanoBanana can pop these out so well. I use them all the time in my social media marketing and now it takes about a minute to make them instead of an hour.

Here is the prompt used:

“Create an infographic image of 2025 Apple MacBook Air 13 and much with a handsketch line paper style using blue and red pen aesthetic. Combine a realistic photograph or photoreal render of the luxury laptop with technical annotation overlays placed directly on top.

Design Style: Handsketch line paper aesthetic with blue and red pen style. Use a textured white or cream-colored line paper background with visible pencil grid lines and paper texture. The annotations should look like they were drawn with blue and red pens on this line paper.

Technical annotations include:
•Key component labels in blue pen with elegant, handwritten style
•Internal cutaway or exploded-view outlines with blue and red pen lines, showing the movement mechanism, rotor, and winding system
•Measurements, dimensions, and scale markers in blue pen with handwritten style
•Material callouts with blue and red pen annotations and subtle icons
•Arrows indicating function, force, or flow (compute power, hard drive, water resistance, power reserve) in blue and red pen with handwritten style
•Simple schematic or sectional diagrams with blue and red pen lines and handwritten annotations
Place the title "MacBook Air” inside a handsketch-style technical annotation box with blue and red pen lines in one corner.

Style & layout rules:
•The real MacBook Air remains clearly visible beneath the annotations
•Annotations feel handsketch, handwritten, and authentic
•Clean composition with balanced negative space
•Educational, museum-exhibit / engineering-manual vibe with a handsketch twist

Visual style:
Minimal technical illustration aesthetic, blue and red pen linework over realistic imagery on textured line paper, precise but slightly hand-drawn feel with authentic sketching style. The paper texture and pencil grid should be visible.

Color palette:
White or cream-colored line paper background, blue and red pen annotation lines and text. Handsketch and authentic aesthetic.
Output: 4:5, ultra-crisp, social-feed optimized, no watermark.”

Thanks for reading today’s Midas Report. Got lots of good stuff coming.

See you tomorrow,
MIDAS AI

Today’s signal is clear... Enterprise AI is getting audited, agents are growing up, cloud pricing is turning into a knife fight, and energy is becoming the real moat.

The AI conversation is quietly moving from demos to discipline. Founders are being asked to prove ROI, not just capability. Operators are being forced to govern agentic systems that touch real workflows. And investors are watching the stack harden into something that looks less like SaaS and more like utilities, supply chains, and balance sheets. Here is what matters today, and what to watch before the market makes the decision for you.

Enterprise AI must tie to P and L or risk becoming shelfware 

The enterprise AI hangover is here, and the cure is visibility, governance, and a direct line from models to money.

Bizzdesign is amplifying a message that is spreading fast across large enterprises. AI programs that cannot show measurable business impact are about to get cut, deferred, or quietly starved. Forrester is already forecasting a correction, with 25 percent of planned 2026 enterprise AI spend expected to slip into 2027. That is not an AI winter. It is a budgeting committee finally asking for receipts.

The data behind the warning is ugly in a familiar way. Forrester found only 15 percent of AI decision makers reported a positive impact on profitability in the last 12 months, and fewer than one third can link AI outputs to concrete business benefits. EY says more than 70 percent of organizations claim they have scaled or integrated AI, yet only about one third have governance protocols to guide or evaluate those initiatives. S and P Global adds that just over one third have an AI policy, and only 21 percent measure impact. Gartner piles on with a root cause that every operator recognizes. Nearly two thirds lack the data management practices needed for effective AI implementation.

Bizzdesign, unsurprisingly, positions its Enterprise 

Transformation Suite as the antidote. The pitch is not magic models. It is architecture repositories and portfolio management that map business activities to data, technology, and change initiatives, so AI work stops living as scattered experiments and starts behaving like a program with accountability.

For founders selling into enterprise, the takeaway is blunt. You will win deals faster if you show how your AI initiative plugs into existing process metrics, risk controls, and systems of record.

For operators, the playbook is to treat AI like transformation, not like tooling. If you cannot describe which processes changed, which costs moved, and which controls apply, you are funding AI sprawl.

ByteDance declares AI war on Alibaba’s cloud empire 

Cloud competition in China is reportedly turning into a price and product blitz, with AI agents as the new wedge.

The Financial Times frames a fresh escalation in China’s cloud market, with ByteDance pushing aggressive discounts and leaning into AI agent capabilities to challenge Alibaba’s cloud position. Even without clean, verifiable details from the source extract here, the strategic direction fits a broader pattern we have seen elsewhere.

When model access starts to look similar, vendors compete on distribution, integration, and pricing. And when workloads shift toward inference and agent driven automation, the cloud vendor that owns the workflow layer can dictate the economics.

If this dynamic holds, it is not just a China story. It is a preview of how cloud margins can compress when AI becomes table stakes. Discounting is the obvious lever, but the more interesting one is bundling agent platforms into cloud contracts so switching costs rise.

For founders, that means your go to market may get harder if hyperscalers or national champions bundle agent tooling, orchestration, and hosting into one line item that procurement already understands. For investors, watch for pressure on standalone platform providers unless they have a differentiated data advantage, a vertical moat, or a distribution channel that clouds cannot easily replicate.

The operator takeaway is practical. Treat your cloud vendor relationship as a strategic dependency, not a commodity. If pricing wars begin, the short term savings can hide long term lock in through proprietary agent frameworks, observability, and identity. A cheaper bill this quarter is not a win if your workflow layer becomes unportable next year.

AI agents are becoming strategic business partners

The new enterprise advantage is not automation. It is trusted agentic systems embedded in decision making.

The World Economic Forum highlights the shift from task tools to agentic AI that acts more like a collaborator inside the business. KPMG’s work is a useful anchor here.

In 2024, KPMG launched a study using its patent pending AI Value Assessment to quantify labor productivity impact. Their estimate is bold. Fully embracing agentic AI could unlock roughly 3 trillion dollars in global productivity gains, equivalent to about a 5 percent improvement in average profitability measured by EBITDA for Fortune 1000 companies. That is the kind of number that makes CFOs pay attention, and also the kind that forces teams to prove causality.

One concrete example is a KPMG client that built an AI powered Career Companion for 15,000 employees. The reported output is not just convenience. It generated personalized career paths and skill development plans, created more than 650,000 skills, and cut the time required to generate skills and job architectures by 99.75 percent.

Whether your org is excited or terrified by that, the point is that agents are moving into high leverage, high context internal workflows, not just customer support macros.

The gating factor is trust and control. Stephen Chase, KPMG’s Global Head of AI and Digital Innovation, puts it plainly. AI only scales safely and sustainably when trust is built in from day one. In practice, that means agent control systems, identity and access management designed for agents, real time observability dashboards, centralized AI registries, and human in the loop oversight. Governance is also becoming standardized through regulation and norms, with the EU AI Act and ISO 42001 showing up as reference points.

For founders, the opportunity is to sell the missing layer between model capability and enterprise comfort. Controls, auditability, and integration into daily workflows are increasingly the product.

For operators, expect new roles like orchestration engineers and responsible AI or trust engineers to become real headcount items.

For investors, the winners will be the teams that ship agents that can be supervised, measured, and revoked, not just agents that can talk.

Microsoft CEO says energy is now the gatekeeper to AI scale  

Compute was the constraint. Now power and the ability to secure it are the real differentiators.

CNBC reports Satya Nadella warning that energy consumption will dictate who can afford to lead in AI infrastructure and training, especially as AI capex climbs into the hundreds of billions. The specific verifiable details are limited in the provided extract, but the thesis is consistent with what anyone building at scale is feeling. You can rent GPUs. You cannot instantly rent new grid capacity.

This reframes the competitive map… The advantage is no longer just better models or more GPUs. It is access to power, permitting, data center buildout, and the engineering discipline to make tokens cheaper through efficiency.

For founders, this increases the value of architectures that reduce compute per outcome, whether through smaller models, better retrieval, smarter caching, or tight agent orchestration that avoids wasteful loops.

For operators, energy is going to show up as a board level KPI, because it is directly tied to cost of goods sold for AI heavy products.

And for investors, the next wave of moats may look unsexy on the surface. Grid partnerships, efficient inference stacks, and infrastructure siting decisions could matter as much as model benchmarks.

The through line today is simple. AI is growing up. The companies that win will be the ones that can explain, control, and power what they build.

Today’s AI tape had a familiar split personality. Public markets kept chasing upside narratives like aggressive AMD targets and billionaire backed winners, while the quieter constraint got louder in the background. Power. Not funding. Not hype. Power. Crypto felt that too as AI tokens lost momentum, then perks like new futures listings tempted traders back in.

📈 AI Market Movers & Trends 📈

The most telling signal today was not another breathless AI stock list. It was the market’s ongoing tug of war between conviction and caution.

On one side, you have analysts tossing out ambitious upside cases for AMD, the kind of price target that dares momentum traders to reload. That matters because AMD is the cleanest liquid bet on AI acceleration that still has a catch up narrative versus Nvidia. Bulls love a comeback story, especially one wrapped in datacenter GPUs and inference demand.

On the other side, the adults in the room keep pointing at what actually governs AI scaling now. Electricity and grid capacity. The Yahoo driven theme that power, not capital, will decide who wins the AI wars is not a cute thought experiment. It is a real constraint that turns AI from a pure software dream into an industrial supply chain fight. Investors are starting to price that in with more scrutiny on hyperscaler capex efficiency and on chipmakers being careful with expectations.

Meta sits right in the crosshairs. The market is increasingly demanding a clean line from heavy AI capex to 2026 ROI, and the coming earnings framing is basically a referendum on whether Meta can convert spend into durable monetization while Reality Labs keeps eating cash. If Meta can show credible returns from better ad targeting, engagement, and creator tooling, the multiple can hold. If not, the capex story shifts from investment to indulgence fast.

Underneath it all is the bubble question creeping back into the discourse. Not because AI is fake, but because the market has a habit of paying tomorrow’s winners in full today. The next phase is about who can actually deliver compute, power, and profit at the same time.

💸 Funding Watch 💸

Funding headlines were oddly quiet in the input today, and that in itself is a signal.

In 2024 and 2025 you could not refresh your feed without a new foundation model round. Now the center of gravity is shifting toward applied AI, infrastructure, and regulated workflows where ROI is easier to prove and distribution is king.

The broader venture context helps explain why. Fintech has climbed back to roughly 11 percent of global venture funding, second only to AI. That is not just a fun league table stat. It is a clue that investors are pressing AI teams to ship into revenue rich verticals with clear paybacks, and finance is the obvious playground. Think underwriting automation, fraud, compliance, and back office agents that save real dollars. When fintech funding rises alongside AI, you should expect more AI native fintech and more fintech incumbents quietly buying or partnering with AI startups to defend margins.

Another trend hiding in plain sight is how “infrastructure” is being redefined. It is no longer just chips and cloud credits. It is energy procurement, datacenter siting, and the ability to lock down reliable power. That is why the power constraint theme bleeds into startup strategy. The winners will be the teams that can secure compute predictably and price it sustainably, not the ones with the prettiest demo.

If you are a founder, the fundraising pitch is shifting. Investors are less impressed by parameter counts and more interested in unit economics, deployment pipelines, and distribution hooks. If you are an operator, expect more M&A curiosity around tools that compress inference costs, manage GPU fleets, or reduce model drift. The next big deal wave is likely to look boring on the surface and extremely profitable underneath.

🪙 Crypto Moves 🪙

AI crypto had a rougher tone today, and it was the useful kind of rough. The sector could not hold early momentum and liquidations punched the usual weak leverage out of the system. That is healthy, even if it hurts.

When AI tokens only go up, builders get complacent and traders get reckless. A reset forces everyone to ask the only question that matters. What is the token actually doing.

Two pockets showed relative strength. ICP and DePIN names held up better while parts of the AI token complex sagged. That tracks with the power and compute narrative from equities. DePIN is effectively the crypto world trying to solve infrastructure constraints with market incentives, whether that is compute, storage, bandwidth, or coordination. When the public markets start talking about energy as the bottleneck, it is not crazy that crypto traders rotate toward networks that claim to make infrastructure more accessible.

The other notable catalyst was Binance listing AIA USDT perpetual futures. Futures listings are not fundamentals, they are adrenaline. They increase liquidity, invite leverage, and can turn a sleepy chart into a casino table. Traders will treat it as a volatility opportunity. Builders should treat it as a distribution moment, because liquidity brings attention, and attention can turn into users if the product is real.

Zooming out, the bigger convergence story is market structure. The Wall Street Journal report that NYSE is moving toward a 24/7 trading platform for blockchain based securities is not directly an AI token headline, but it is a permission slip for the next decade.

If tokenized assets and always on markets become normal, AI driven market making, risk management, and automated compliance become inevitable. The irony is delicious. The more finance goes on chain, the more AI becomes the interface layer that makes it usable.

📊 Stay tuned for tomorrow’s MarketPulse and sign up to our daily Midas Report newsletter. 📊 

Welcome to your daily AI intelligence brief.

Let’s dive into what is unfolding in AI today, January 19th, 2026…

Let’s start with the big story of the last couple days… OpenAI announcing ads in ChatGPT.

The age of ad ree artificial intelligence just ended.

On January 16th, OpenAI announced it will begin testing advertisements within ChatGPT for U.S. users on its free and newly expanded "Go" tiers, marking the most significant shift in the platform's business model since its 2022 launch.

For investors, this isn't just a product update… it's a signal that the AI gold rush is entering its monetization phase.

With an estimated 800 million weekly active users and projected infrastructure costs of $1.4 trillion over the next decade, OpenAI needed a revenue engine that could scale beyond subscriptions alone.

How It Works…

Advertisements will appear at the bottom of ChatGPT responses when there's a relevant sponsored product or service based on the current conversation.

These "Sponsored Recommendations" will be clearly labeled and separated from organic answers. A user asking for keto dinner recipes might see an ad for avocado oil or grocery delivery services… highly targeted, contextually relevant, and potentially lucrative.

The company has drawn clear lines… Plus, Pro, Business, and Enterprise subscriptions will remain ad free, and users under 18 won't see any advertisements.

Sensitive topics including health, mental health, and politics are off-limits for ad placement.

The Financial Picture…

OpenAI's move comes as the company reportedly burns through capital at an extraordinary rate, with projections suggesting it could consume up to $115 billion by 2030.

Despite reaching $20 billion in annualized revenue, the subscription model alone couldn't support the massive computational demands of serving nearly a billion users while developing increasingly sophisticated AI models.

CEO Sam Altman, who previously called ads in AI "uniquely unsettling," appears to have made peace with necessity.

His recent statement was direct: "A lot of people want to use a lot of AI and don't want to pay, so we are hopeful a business model like this can work".

What This Means for Markets…

This pivot puts OpenAI in direct competition with Google's $200+ billion search advertising empire.

Microsoft, OpenAI's primary backer and cloud provider, stands to benefit significantly as advertising drives increased demand for Azure infrastructure.

NVIDIA remains the ultimate winner… more ads mean more compute power, which means more chips.

The losers? Traditional content publishers face further decline as AI provides direct, monetized answers without sending traffic to source websites.

The "Great Decoupling" just accelerated.

The Bottom Line…

OpenAI's advertising test, launching in the coming weeks, represents a calculated gamble.

The company promises that ads will not influence ChatGPT's answers and conversations will remain private from advertisers. Whether users believe this as ad revenue pressures mount remains to be seen.

For investors, the message is clear… AI companies are transitioning from growth at all costs to sustainable monetization.

The free ride is over.

As someone who has run paid ad campaigns across the web for the past decade, this will be a huge opportunity for businesses to put their product in front of the right people… at least until the ad blindness kicks in.

Prompt of the Day

Today, we use Nano Banana to help us pump out some infographics. These images do incredibly well on social media and are a consistent part of our daily facebook marketing. Nano Banana has made them as easy as a single prompt where it would take at least an hour to pump these out prior…

Here is the prompt I used. You could do even better splitting the research and the image creation… but I wanted to showcase how solid Nano Banana can be.

Prompt:

“Create this image, and do these steps…

1. Research and create an isometric infographic illustration titled “AI Tools That Actually Save Time” (Top 10) showing the top ai tools that in order of time saved with usage

Visual Style:
Create 20 adorable 3D clay-style miniature dioramas in a kawaii “blind box collectible” aesthetic. Each scene should feature:
• A unique isometric building representing the brand
• The brand’s mascot or iconic elements
• Tiny chibi-style customers and staff
• Clean borders with soft plastic textures and bright studio lighting

Layout Requirements:
• Arrange all dioramas in a flowing, connected layout on a warm cream/beige background
• Connect scenes with dotted lines to show ranking flow
• Label each with rank number, brand name, and exact location count in clear black text
• Add decorative red circular elements scattered throughout
• Include “Official reports” citation at bottom
• Use harmonious colors with a playful, educational infographic feel

Overall Composition: The final image should look like a professional, shareable social media infographic that’s both informative and visually delightful - perfect for viral engagement. Make it in 9x16 aspect ratio”

Thanks for joining us today! We have a lot of cool stuff coming for you guys from AI Automation Kits to some free training to some extra goodies.

Great to have you. Talk to you tomorrow.

MIDAS AI

Enterprise AI is getting stricter, mobile AI is getting richer, and hardware breakthroughs are quietly rewriting the cost curve

If you feel like AI is simultaneously becoming more industrial and more everywhere, you are not imagining things. Today’s lineup spans a millionfold energy cut in training, a sharper enterprise playbook for reliability, a fresh wave of APAC budget expansion, IBM packaging agentic AI into a consulting product, and a reminder that consumers are already voting with their thumbs and wallets.

The theme is execution over excitement. The winning teams in 2026 will not be the ones with the flashiest demo. They will be the ones who can ship systems that behave, pencil out economically, and scale without melting the planet or the org chart.

New training method cuts AI energy use by a millionfold 

A memristor training update trick turns energy into a rounding error, and makes edge AI feel a lot less imaginary.

Researchers from Zhejiang Lab and Fudan University published an approach called error aware probabilistic update that trains neural networks directly on memristor based analog in memory computing hardware while slashing training energy use by nearly six orders of magnitude versus GPUs. The work appeared in Nature Communications, and it tackles a very specific pain point that has held analog training back: write errors in memristors that worsen over time due to device relaxation.

Their move is conceptually elegant. Instead of performing dense, constant weight updates like standard backprop, the method probabilistically transforms the update magnitude so the average update stays the same but the number of actual writes plummets. In experiments and simulations, it drives update frequency below 0.1 percent, with one reported run updating only 0.86 per thousand parameters during training of a 152 layer ResNet. Fewer writes mean less energy and longer device life, and the team reports around a 1,000 times lifetime extension thanks to reduced write operations.

Why founders and investors should care is not just the headline number. If training can happen on low power analog hardware, you can imagine a future where certain models are trained or fine tuned closer to the edge, where power budgets and infrastructure are constrained. It also reframes the “AI is capped by energy” argument.

Watch whether this approach generalizes cleanly beyond memristors to other non volatile memory tech like ferroelectric transistors or magnetoresistive RAM, and whether the tooling ecosystem emerges to make analog training feel less like a lab project and more like a product roadmap.

Salesforce targets enterprise friction in GenAI adoption 

Salesforce is betting the next phase of enterprise GenAI is not smarter models, but stricter systems.

Salesforce is leaning into the reality that pilots are easy and production is a compliance nightmare. Its message is that strong benchmark performance does not automatically translate to consistent business outcomes, and that enterprises do not want 97 percent correctness when the workflow needs to work 100 percent of the time. That framing is refreshingly unromantic and deeply accurate.

The strategy is to combine generative AI with deterministic systems so rules, standard operating procedures, and audit constraints remain non negotiable. In practice that means a blend of large and small language models, used where flexibility, reasoning, and empathy are needed, with rule based logic wrapping the parts that cannot be allowed to drift. This is less “LLM as the app” and more “LLM as a component inside an engineered workflow,” which is how regulated industries and serious back offices actually buy software.

The signal here is market maturity. Vendors who keep selling general purpose copilots will get squeezed by buyers asking for reliability, governance, and measurable ROI. Watch 2026 for procurement language shifting from model choice to system guarantees, escalation paths, and proof that edge cases are handled without a human babysitter on call.

APAC enterprises to lift AI budgets by 15% in 2026 

ASEAN spend is climbing, hybrid is the default architecture, and agentic AI is moving from curiosity to cautious trials.

Lenovo’s CIO Playbook 2026 study run by IDC surveyed 920 APAC tech leaders and found 96 percent of ASEAN organizations plan to increase AI investment in 2026, averaging 15 percent growth. Even more telling is where the money goes. Hybrid AI infrastructure is being adopted by 86 percent of APAC orgs, reflecting a pragmatic mix of public cloud, on premises, and edge deployments driven by cost control and data sovereignty.

The economics are forcing the architecture. Lenovo points out inference can cost 15 times more than training over a model’s lifetime, and its APAC leadership expects 75 percent of AI compute will be dedicated to inference by 2030. That is a quiet but profound shift. If your product relies on heavy inference, your margin and your customer’s cloud bill will become the same conversation. This is why hybrid, quantization, caching, and model right sizing are turning into board level topics rather than engineering trivia.

Agentic AI is also edging forward. Sixty percent of APAC organizations are exploring or planning limited deployments, but only 10 percent are ready to scale, largely due to governance, integration, and data quality. The opportunity for builders is clear: win by making hybrid operations and agent governance boring and manageable, especially as non IT departments increasingly fund AI initiatives.

IBM rolls out Enterprise Advantage to scale agentic AI

IBM is packaging its internal AI delivery playbook into a productized consulting lane for governed agentic deployments.

IBM launched IBM Enterprise Advantage, an asset based consulting service available January 19, 2026, aimed at helping clients build, govern, and operate internal AI platforms at scale. The pitch is low disruption: clients can scale agentic AI without changing cloud providers, models, or core infrastructure, and can integrate across AWS, Google Cloud, Azure, IBM watsonx, plus open source and closed source models.

This is IBM recognizing the true constraint in enterprise AI is not model access, it is platform operations. The offering builds on IBM Consulting Advantage, which IBM says has supported over 150 client engagements and boosted consultant productivity by up to 50 percent. In other words, IBM is trying to sell the assembly line, not just the parts, and to do it with reusable assets rather than bespoke consulting hours.

For operators, the watch item is whether these packaged services become the “agentic AI platform in a box” that wins budget faster than DIY internal builds. For startups, this is both competition and validation: buyers want governed platforms and migration free adoption. If you sell into enterprise, your differentiation needs to be sharper than “we orchestrate agents,” because IBM is now saying that too, with procurement friendly wrapping.

Chart shows generative AI apps are dominating mobile 

Consumer GenAI is no longer a novelty category, it is becoming a top tier mobile business line.

Visual Capitalist, citing Sensor Tower, shows generative AI apps on track to approach 4 billion downloads, $4.8 billion in in app purchase revenue, and more than 43 billion hours of time spent in 2025. The forward projection is even louder: consumer spending is expected to exceed $10 billion in 2026, with generative AI rising from the number 10 category in downloads to number 4, and jumping to number 3 in in app purchase revenue, surpassing Dating and Social Discovery. By time spent, it is projected to hit number 5 globally.

ChatGPT is a useful benchmark: as of Q3 2025 it ranked number 2 globally by in app purchase revenue across iOS and Google Play, behind only TikTok. That is not “people are trying AI.” That is “people are paying for AI at scale.” The monetization proof matters because it suggests consumers will subscribe for utility, not just entertainment, and it gives builders permission to price for value.

The strategic takeaway is that mobile is becoming a primary distribution channel for GenAI, not a companion. Watch for the next wave to be less chat and more embedded workflows: education, creation, search replacement behaviors, and vertical tools that feel like apps, not prompts. If you are investing, this is where retention metrics will start to look like real software businesses, not viral experiments.

Today’s AI tape had a familiar split personality. Chips stayed in the driver’s seat, but the road got narrower after TSMC signaled a capacity crunch. Meanwhile software names kept sliding as investors re priced what “AI powered” really means when it hits the income statement. In crypto, AI tokens leaned hard into hype, and the scam warnings got louder.

📈 AI Market Movers & Trends 📈 

The most important AI headline today was not another Nvidia product tease. It was TSMC essentially telling Nvidia and Broadcom that the buffet line is getting long. The foundry warned customers about a capacity squeeze as AI chip demand keeps surging, which sounds like a nice problem until you remember supply tightness can turn into delayed shipments, pushed out revenue, and awkward guidance language.

For investors, this is the subtle shift from pure demand story to constraints and pricing power story. If TSMC stays tight, the companies closest to the allocation spigot win, and everyone else gets to explain why bookings are strong but delivery is later.

The broader market backdrop is also changing. The Wall Street Journal noted the Magnificent Seven trade is losing its one way momentum. Translation. The same mega caps that dragged indices higher on AI optimism are now getting judged on execution and capex math, not just narrative. That matters because it pulls oxygen from the “everything AI goes up” phase and forces a rotation toward companies that can actually monetize AI rather than just talk about it.

On the other side of the ledger, software stocks kept selling off as “AI fears” deepened. That phrase is doing a lot of work. What it really means is investors are worried AI will compress pricing, increase churn, and make switching costs weaker for certain SaaS categories. Bargain valuations are not enough if the market thinks your moat is a prompt away from collapsing. Chips look like scarcity. Some software is starting to look like a commodity.

💸 Funding Watch 💸 

Startup land was quieter on big round announcements in the research feed, but one theme stood out. Everyone wants to be an ecosystem. NeoBio.ai launched what it called an AI native ecosystem designed around long term value circulation, which is a fancy way of saying it wants to be the platform where multiple participants create and capture value rather than a single point solution that gets copied in six months.

Founders are responding to the same pressure public software is feeling. If standalone features get eaten, you either build distribution, build network effects, or you get priced like a plugin.

This also maps cleanly onto what public market investors are rewarding. Hardware and infrastructure keep their leverage because the bottlenecks are physical and the demand is real. Pure software has to show either proprietary data, locked in workflows, or a wedge that expands into a durable suite. The funding market is internalizing that shift even when the headlines look like product launches instead of giant checks.

There is another under discussed angle. The TSMC capacity warning is a quiet fundraising tailwind for alternative compute and efficiency startups. If top tier GPUs stay constrained or expensive, anything that improves utilization, inference efficiency, model compression, or workload orchestration becomes more investable. The next wave of meaningful AI deals is likely to sit in the boring middle layer between model makers and apps. The layer that makes AI cheaper, faster, and predictable enough for enterprises to sign multi year contracts.

🪙 Crypto Moves 🪙

AI crypto had a very on brand day. Big promises, shiny presales, and a growing chorus of adults in the room telling people not to get rugged. Multiple pieces pushed the familiar “could this AI coin change your life” framing, with tokens like DeepSnitch AI being marketed with extreme return scenarios. When you see specific numbers like turning 1000 into 200k, you are not looking at product market fit. You are looking at a liquidity story and a marketing funnel.

At the same time, Bitget published a 2026 guide on identifying and avoiding AI crypto scams, and that is the tell. The more the sector leans into AI narrative, the more it attracts opportunists selling black box bots, fake model dashboards, and “agent platforms” that are basically a token wrapper around a Telegram group. The signal for traders is simple. If the pitch is mostly upside and referrals, treat it like a hot stove. If the pitch is verifiable usage, transparent token utility, and real integrations, it at least deserves a second look.

The interesting convergence point is this. Public markets are obsessing over real supply constraints in compute, while crypto markets are minting synthetic stories about compute and AI agents.

The winners in AI crypto will not be the loudest presale. They will be the projects that can credibly connect tokens to something scarce in the real world, like verifiable compute, model access, or distributed infrastructure that enterprises can actually use. Until then, the best trade might be skepticism.

Welcome to today’s Midas Report.
Your one stop shop for everything that matters in AI.

Here is what is unfolding today, January 16, 2026….

Our top story comes down to one of my favorites.

This week Anthropic shipped Cowork as a research preview inside the Claude app on macOS, and it is only for the 100 a month plan.

That pricing and gating are the tell…

They are testing with early adopters because this is the kind of feature that either becomes a daily habit or gets ignored forever.

Cowork is built on Claude Code, but it is dressed for people who do not want to touch a terminal. The pitch is simple. Give it a task, let it handle the busywork, and keep you in the loop.

The substance is in what it actually does. It can organize files into folders, convert file types, shrink a too big PDF, turn 20 JPEGs into one PDF, and generate reports from a pile of stuff. It can also use a browser to help manage emails or gather information, but only after it asks permission for each action.

That permissioned design matters because it is where agents stop being a fun demo and start being usable in real workflows.

Cowork runs in a virtual machine sandbox and you explicitly grant access to specific folders. If you do not grant access, it literally cannot see them.

That is the right direction for trust, even if Anthropic still warns you not to expose sensitive data and suggests a dedicated nonsensitive folder due to risks like prompt injection from sketchy web content.

Zoom out and you can see the strategy. Anthropic is not chasing the loudest model benchmark. It is trying to own the work surface where plugins and ecosystem apps will inevitably attach.

Once users get used to an AI that can take actions, the next question becomes which tools it can connect to and which workflows it can automate.

Claude Cowork is Anthropic quietly admitting what most AI tools have been pretending not to be.

Not a replacement for you. A seatmate.

If you run a team, the move is to start testing agent based workflows now, internally, on low risk tasks. Pick repeatable ops, document wrangling, and lightweight analysis, then see what sticks. If you are building on Claude, pay attention to Cowork’s interaction design. The sticky part is not magic. It is the way it asks, waits, and then does the thing.

A bit of a versus today… I wanted to make a whole bunch of art styles of the same person.

I tried standard ChatGPT against Nano Banana.

Nano Banana did exactly what I wanted… ChatGPT got the layout okay, but actually giving different art styles… it failed.

Nano Banana Pro

ChatGPT

Here is the full prompt used along with a reference image of my face…

“Create a comprehensive art style showcase featuring a male portrait (reference the subject from the provided image) displayed in 45 different art styles arranged in a grid layout. Each style should be clearly labeled and visually distinct. The styles include:
1. Future City Style - Futuristic urban aesthetic with neon lights and digital elements
2. White Box Art Style - Minimalist white background with clean geometric shapes
3. Traditional Chinese Ink Wash Style - Brush strokes with ink wash technique
4. Rococo Style - Ornate, decorative with pastel colors and intricate details
5. Ethnic Style - Cultural patterns and traditional motifs
6. 3D Animation Style - Rendered with depth and dimensional effects
7. Ink Wash Illustration Style - Watercolor-like ink wash technique
8. Handicraft Style - Textured, handmade appearance with visible brush strokes
9. Wild Art Style - Abstract, expressive, energetic brushwork
10. Sci-Fi Black and White Style - Monochrome with futuristic elements
11. Mechanical Style - Industrial, gears, metal textures
12. Naturalism Style - Realistic, detailed, natural colors
13. Rural Idyllic Style - Peaceful countryside aesthetic with soft colors
14. Abstract Composition Style - Non-representational geometric shapes
15. Light Luxury Style - Elegant, sophisticated with gold accents
16. Sports Style - Dynamic, energetic with bold colors
17. Neon Style - Vibrant neon colors and glowing effects
18. Virtual Reality Style - Digital, holographic, futuristic interface elements
19. Red Art Style - Dominant red color scheme with bold strokes
20. Classical Art Style - Renaissance-inspired with rich colors and details
21. Cartoon Style - Simplified, colorful, playful
22. Realism Style - Photorealistic with high detail
23. Watercolor Style - Soft, flowing watercolor technique
24. Black and White Film Style - Vintage film grain, monochrome
25. Cyberpunk Style - Neon-lit, dystopian, high-tech
26. Abstract Art Style - Non-objective, expressive forms
27. Impressionistic Style - Soft, blurred brushstrokes with visible texture
28. Futurism Style - Dynamic, speed lines, modernist aesthetic
29. Oil Painting Style - Rich, textured oil painting technique
30. Modern Minimalist Style - Clean, simple, geometric
31. Romanticism Style - Emotional, dramatic, rich colors
32. Romantic Oil Painting Style - Romantic themes with oil painting technique
33. Surrealism Style - Dreamlike, impossible imagery, unexpected juxtapositions
34. Decorative Art Style - Ornamental, patterned, decorative elements
35. Understated Realistic Style - Subtle, refined realism
36. Ultra-Modern Style - Cutting-edge, avant-garde, innovative
37. Retro Style - Vintage, nostalgic, 1970s-1990s aesthetic
38. Punk Style - Edgy, rebellious, bold colors and graphics
39. Anime Illustration Style - Japanese anime aesthetic with vibrant colors
40. Fantasy Style - Magical, mythical, imaginative elements
41. Minimalism Style - Extreme simplicity, clean lines
42. Baroque Art Style - Dramatic, ornate, rich colors and details
43. Flat Design Style - Modern, clean, two-dimensional
44. Realistic Oil Painting Style - Detailed oil painting with rich textures
45. Light and Shadow Photography Style - Dramatic lighting with strong contrast

Arrange all 45 styles in a clean, organized grid layout with clear labels for each style. Use a professional, gallery-like presentation. The subject should be recognizable across all styles while maintaining the unique characteristics of each art style. Ensure the layout is balanced and visually appealing in a 4:5 aspect ratio.”

Thanks for reading today’s Midas Report! See you tomorrow.

MIDAS AI

Four signals today that video, retail, and hardware are turning AI from demos into durable businesses

If you are building in AI right now, today’s headlines read like a progress report on a bigger shift. Consumer traction is accelerating, enterprise adoption is operational not experimental, and the biggest lab in the room is quietly assembling a supply chain. Meanwhile, a top VC firm is basically waving a flag that says your old SaaS instincts might get you hurt.

Here’s what mattered today, and what founders and investors should be watching next.

Higgsfield hits a 1.3B valuation and makes AI video feel inevitable

A Snap veteran’s new company just proved that generative video can scale like a consumer app and monetize like a real business.

Higgsfield, founded by former head of Generative AI at Snap Alex Mashrabov, extended its Series A with an additional $80 million, bringing the round to $130 million total and putting the company at a reported $1.3 billion valuation. The product is a consumer and creator focused tool for creating and editing AI generated video, and it is moving fast from novelty to workflow.

The traction is the story. Higgsfield says it crossed 15 million users nine months after launch and is already at a $200 million annual revenue run rate, up from $100 million just two months prior. That kind of revenue velocity is rare even in the frothiest consumer cycles, and it suggests two things operators should internalize. First, AI video is becoming a core marketing surface, not just a creative toy. Second, time to monetization for breakout AI consumer apps is compressing because the output is directly tied to revenue generating activities like ads, social, and commerce.

Strategically, Higgsfield also signaled a shift in who uses the product, from casual creators toward professional social media marketers. That is where budgets live and where retention gets sticky, because content calendars do not take days off. With Accel, Menlo Ventures, AI Capital Partners, and GFT Ventures participating, the market is effectively voting that the category will produce large outcomes. What to watch next is whether Higgsfield can defend on distribution and workflow integration, because in video, models commoditize but channels and habits compound.

OpenAI starts shopping for US hardware suppliers  

OpenAI is acting less like an API company and more like a future device and robotics platform with a domestic supply chain.

Bloomberg reports that OpenAI is soliciting proposals from US based hardware makers as it plans an expansion into consumer devices, robotics, and data center technology. There is not much subtlety here. This is a signal of intent to move down the stack and into physical product development, which changes the competitive map for everyone who assumed the frontier labs would stay purely software.

For founders, this matters because vertical integration creates new choke points and new partner opportunities at the same time. If OpenAI builds or heavily influences the reference architecture for devices and robotics, it can define the default interfaces, the safety and control layers, and the distribution relationships. That can be great if you are building components, sensors, actuation, testing infrastructure, or specialized manufacturing capability. It can be existential if your “device startup” is mostly a wrapper around someone else’s model without a differentiated hardware advantage.

For investors, the US supplier angle is equally strategic. It hints at a supply chain posture shaped by geopolitics and export controls, plus a desire for tighter iteration loops between model teams and hardware teams. What to watch is whether this becomes a quiet ecosystem play, similar to how smartphone platforms cultivated suppliers, or a more direct push into branded consumer hardware. Either way, OpenAI is broadening the battlefield.

Menlo says there are no AI markets, only proto markets 

If you are applying the SaaS playbook to AI, Menlo’s argument is that you are optimizing for a world that no longer exists.

Menlo Ventures put out a clear thesis that “AI markets” today are proto markets, meaning they can generate meaningful revenue but lack stable boundaries, durable moats, and settled unit economics. The practical implication is blunt. Category definitions are still melting, product expectations mutate fast, and foundation model releases routinely force application companies to rebuild architecture and workflows.

The examples are telling. Claude Sonnet 4.5 landing reportedly pushed Cognition to rebuild Devin. Manus has reconstructed its agent framework multiple times. On the enterprise side, offerings like Salesforce Agentforce and ServiceNow NowAssist blur lines between categories like sales, IT, support, and search. The point is not that the products are bad. The point is that the ground under them is moving, so “ship once then defend” is not a real strategy right now.

Menlo also flags a more uncomfortable truth. In areas like code and voice AI, pricing is sometimes at or below cost, which means classic SaaS margin assumptions are not yet reliable. Their prescription is to compete on product plasticity and learning velocity. Build systems that absorb user behavior, evolve quickly, and become “hackable by design,” borrowing language from Anthropic’s Claude Code creator Boris Cherny. What to watch is which companies turn this into a defensible loop, where usage teaches the product faster than competitors can copy, and where distribution and workflow lock in replaces feature lock in.

Walmart’s Sparky shows AI at scale 

Walmart is done tinkering and is using agents in customer journeys, store operations, and fulfillment where impact is measurable.

Walmart is leaning hard into AI transformation, including a partnership with Google’s Gemini platform and continued work with OpenAI. Its in house agent, Sparky, already fields customer service questions and nudges repeat purchasing with reminders for appointments and prescriptions. Walmart also enabled purchases inside ChatGPT in Fall 2025, and the Gemini integration lets users access membership perks and sync purchases into a live Walmart cart.

This is what operationalized AI looks like in a physical first business. Walmart associates are using a backend agent to prioritize tasks like restocking and responding to in store issues like spills. Fulfillment centers are using AI to predict which products to store. None of that is flashy, and that is the point. It is the unglamorous automation of thousands of micro decisions that produces real margin and customer experience gains.

The strategic layer is distribution. Walmart is positioning itself to show up inside AI driven inquiry results that may not start as commerce questions but can convert into sales. Daniel Danker, Walmart’s EVP of AI acceleration, product and design, summed it up with unusual clarity.

This is the year where tinkering becomes transformation, and the bigger risk is not being out front.

What to watch is the retail arms race this triggers. Once a retailer the size of Walmart treats AI agents as a core interface, every competitor is forced to match on speed, personalization, and presence inside the new AI shopping funnels.