WEBVTT

1
00:00:00.080 --> 00:00:11.560
On y'all. Welcome back to Blockspace Live, presented by CleanSpark. For our lead story today, Iron co-founders, the Roberts brothers, have just been awarded 5%

2
00:00:12.460 --> 00:00:30.760
of the outstanding stock in the form of RSUs, and some Iron shareholders have a few questions here, Charlie, about the alignment between founders and shareholders, especially considering that Iron is taking a dump today along with other neo clouds, as the meta news we covered earlier has shaken up the sector.

3
00:00:31.160 --> 00:00:43.420
Following that, we have Jay Patel of Lagos Finance on to talk about private credit potentially showing some more cracks, specifically in the tech sector, and also what's going on with Strive, CETA, and the digital credit ecosystem.

4
00:00:44.000 --> 00:01:02.600
Following that, we will cover Anthropic's reported deal striking with Samsung to develop its own custom AI chip, and we will end on Nvidia's rev share agreement with AI cloud platforms, a pretty new and breaking development that has us questioning the conventional wisdom for compute demand.

5
00:01:03.900 --> 00:01:17.300
Blockspace goes live Monday, Tuesday, Wednesday, Thursday, and Friday, every weekdays at 1:00 PM Eastern, featuring quick hits on AI data centers, emerging tech, Bitcoin, and markets.

6
00:01:17.400 --> 00:01:29.680
If you like what you hear, you'll love the newsletter, newsletter.blockspacemedia.com to sign up, delivered to your inbox every single day. This live stream turns into a podcast.

7
00:01:29.740 --> 00:01:38.680
I mean, where podcasts are streamed, so if you miss it when we're live at 1:00 PM Eastern, you can listen to it later on in the day or the next morning.

8
00:01:38.920 --> 00:01:54.480
This show is brought to you by CleanSpark, NASDAQ listed at ticker C-L-S-K. More on CleanSpark later on in the show. Colin, every single day I wake up and I'm like, "I'm in the wrong business. I- I'm a talking head.

9
00:01:55.180 --> 00:02:01.000
I should have been issuing shares- You should have- "... and then taking those shares."

10
00:02:01.540 --> 00:02:14.160
You should have co-founded a Bitcoin mining company and then pivoted to AI, and fair enough, great track record of execution, and then get yourself awarded roughly $400 million worth of stock grants.

11
00:02:14.980 --> 00:02:27.060
So we'll go ahead and share this. I hate to share noted, uh, you know, AI there. Oh, I also had the tweet pulled up. I also had Jim Chanos pulled up. This is, this is the tweet. So this is James Chanos'

12
00:02:28.000 --> 00:02:37.000
depiction of the news item here. He says, quote, "So Iron just announced an $800 million stock grant, not options, to its two co-CEOs.

13
00:02:37.040 --> 00:02:47.800
To put that into perspective, it is a 17% of the estimated cumulative adjusted net income of 4.7 billion over the term of the grant fiscal year 2027 to fiscal year 2030."

14
00:02:48.440 --> 00:03:05.280
So what specifically happens here was the Iron board just approved co-CEOs William and Dan Roberts to receive roughly 9.1 million restricted stock units each for a combined 18.2 million restricted stock units across both accounts.

15
00:03:05.870 --> 00:03:22.180
This is roughly 5% dilution according to the most updated outstanding shares for Iron, according to latest disclosures, and also some estimates because they have an ATM open, so there have been some estimates backing out how much have been issued under that ATM and how much dilution that has equated to.

16
00:03:22.620 --> 00:03:30.680
We'll know the full number during their most recent or dur- in, in their next quarterly updates. Some notes on this.

17
00:03:31.080 --> 00:03:39.540
There's a four-year annual vesting at 25% a year, and there's a two-year post-vesting lockup before the CEOs can sell or transfer these shares.

18
00:03:39.580 --> 00:03:47.780
So the combined effect of that is six-year total runway from the grant to full liquidity, with the last tranche not fully tradable until their fiscal year 2033.

19
00:03:48.800 --> 00:03:55.620
Uh, in exchange, no further equity grants to either co-CEO until fiscal year 2031. And also,

20
00:03:57.130 --> 00:04:09.300
th- so the, the, th- I, I would like to frame this with the co-CEOs voting power, because they each hold roughly 21.8% of voting power with their Class B shares, which are rated at 15 to one compared to others.

21
00:04:10.240 --> 00:04:19.519
And so together, these brothers have over, you know, 42%, 43.6% of the voting shares in the company.

22
00:04:20.079 --> 00:04:32.700
And the reason why I threw that in there is because with this, with, with these RSUs, you have even longtime Iron bulls kinda asking questions about, you know,

23
00:04:33.800 --> 00:04:46.090
who's really being prioritized here, the shareholders or the insiders in Iron, which granted have created, have minted some incredible generational wealth for early Iron bulls. But

24
00:04:47.740 --> 00:04:59.060
he also states here, this, this, uh, JP Insights guy on X, who calls himself a former holder. He says, "This isn't automatically bad." Imply- implying that he's sold since then. I mean, yeah. I mean, I...

25
00:04:59.240 --> 00:05:07.600
If the stock was up like 1000x- Yeah... since I bought, I would probably- I would also be a former holder, but I don't know, like Iron, you kinda had to be in a little bit of like a true believer camp.

26
00:05:07.660 --> 00:05:16.000
So that implies- You did... you're still holding some, you know, even- Yeah, probably are. But he says, quote, "This isn't automatically bad, but I don't love it at first glance.

27
00:05:16.120 --> 00:05:20.650
The good part is the runway, six years total. It's important to point that out. This isn't something they can sell immediately.

28
00:05:21.060 --> 00:05:29.180
Four years of vesting and then a two-year hold on top for a company where the Roberts brothers basically are the identity that locks them in. Not that I thought they were going anywhere anyway.

29
00:05:29.200 --> 00:05:40.070
The problem is that it's time-based rather than performance-based, and the size is not small. Call it around 5% dilution and something close to 1 billion." Now, the counterpoint to that is they clearly want to do well.

30
00:05:40.100 --> 00:05:49.190
If they tank the company, the shares aren't gonna be worth anything anyway, and they're going to be selling into an illiquid market. But the other thing that this is framed against, Charlie, and then I'll toss it to you,

31
00:05:50.440 --> 00:06:01.580
uh, they have this Golden State Warriors sponsorship deal that a lot of Iron, you know, bulls were kind of... Even the bulls were scratching their heads at. Th- that's...

32
00:06:01.620 --> 00:06:14.170
They were reportedly paid for $50 million, $50 million per year to sponsor the Golden State Warriors and have their logo on their jersey So it leaves w- us with the question, how much is Iron really putting shareholders first?

33
00:06:14.730 --> 00:06:22.870
I would say that they've done probably a very good job of that, considering the stock is the, has been the most explosive of any of the Bitcoin miners pivoting towards AI.

34
00:06:23.990 --> 00:06:33.350
That being said, at a time when your stock's at an all-time high, you also have an ATM out to dilute because the stock's doing really well. I mean, it's not at an all-time high now, it's actually down on the day.

35
00:06:33.370 --> 00:06:41.730
But I think the question for most people is, okay, well, are you really th- looking forward to the next phase of growth? Whether or not that's fair, we'll see in the months ahead. Yeah.

36
00:06:41.830 --> 00:06:58.450
They did the, the Golden State Warriors, uh, sponsorship deal kind of on the heels of the last, I believe the last, uh, quarterly, uh, update. Dan Roberts was saying basically, "Why do we need a marketing budget?"

37
00:06:58.460 --> 00:07:07.280
[laughs] And here we are. They're, they're... He said something to that effect, which is like, "Well, the stock's like the marketing budget." And they are- And then they purchased a media company to- A media company...

38
00:07:07.280 --> 00:07:19.870
handle their marketing [laughs] as well. Yeah. [laughs] And so- So... that plus the Golden State Warriors, and then, um, I mean, here's a tweet that I have, uh, from X Capital Management, "Reaction to Iron Co.

39
00:07:20.110 --> 00:07:31.790
CEO comp package would've been night and day different had they simply tied it to Nvidia investing conditions. Full package vest based on a 70, higher than $70 price per share and 600K GPUs generating revenue.

40
00:07:32.110 --> 00:07:35.970
That's all they had to do. Bar is high for revenue to offset."

41
00:07:36.070 --> 00:07:49.370
I don't know if I agree with that specific claim, but, like, I do think that you would see a, a materially different reaction on at least the Twitter places that we monitor had it been more tied to performance.

42
00:07:50.170 --> 00:08:01.770
In the con- When, when, when, when we talk about performance, using DI Metrics, our favorite weekly haunt here, performance in NeoClouds as a sector is down.

43
00:08:01.810 --> 00:08:18.070
While the S&P is up on the past five days, 1.7%, NeoClouds have gotten absolutely routed post, following the MetaCloud news. Iron down 13.9% in the past five days. Nebu is down 11.7%.

44
00:08:19.210 --> 00:08:33.330
Um, uh, White Fiber also. But, uh, yeah. So NeoCloud's not having a good time the same time that the comp package is announced. And it also, you gotta, you gotta think in context. The semi analysis.

45
00:08:34.170 --> 00:08:42.870
Um, here's the semi analysis, like rankings of the, the quality of different NeoClouds with CoreWeave at the top in the S tier.

46
00:08:43.250 --> 00:08:48.210
Semi analysis, they're obviously the most, like, some of the more technical folks in the space.

47
00:08:48.710 --> 00:09:01.010
Has Iron at an n- underperforming not recommended tier down here with a bunch of other, uh, companies that I haven't heard of. White Fiber obviously, but dStack, dInfra. Is White Fiber also in that red tier?

48
00:09:01.270 --> 00:09:12.850
Yeah, White Fiber's also- Oh... in this red tier down here. But, um, so doesn't even make like a bronze or silver tier. So, um, and, and- Well, for listeners here, like they actually have tested these.

49
00:09:12.890 --> 00:09:17.650
This is what this is based on. Yeah. This isn't based on assumptions or secondhand accounts.

50
00:09:18.270 --> 00:09:34.950
Semi analysis will actually rent out GPUs from these companies and run them and have them perform various tasks, and that's how they a- arrive at these rankings. Yeah. So, um, just not looking very good overall for Iron.

51
00:09:35.150 --> 00:09:40.010
I'd say the optics, the irony being [laughs] that they are spending a lot on marketing right now.

52
00:09:40.110 --> 00:09:50.399
Um, whether or not how this plays out in the market and share price, I'm not qualified to say or really even predict, but, uh, I will say the people are kind of pissed, uh, to say the least.

53
00:09:50.399 --> 00:09:57.250
The people are kind of pissed. And one last thing I'll just say about this with regards to the performance angle.

54
00:09:58.630 --> 00:10:11.890
This is largely based on, I mean, you can almost argue this is like rear-looking rearview performance awards because they, they're almost being awarded this based on the fact that over the past... Ooh, actually

55
00:10:13.450 --> 00:10:22.610
Iron not looking as hot as it were earlier. Iron flipped negative on the year. Yeah. I believe we even had a tweet out about it yesterday where technically they're down in 2026 to date.

56
00:10:22.770 --> 00:10:34.220
Yeah, and there was a time- You, you for- you forget that the Iron run happened like a year and a half ago. Yeah. And it's just been kind of like roughly not really trending, uh, one way or the other since then.

57
00:10:34.670 --> 00:10:46.670
The stock peaked at 76.41 in November of 2025, and it's, it's currently still up 143% to date, but it was up, uh, much, much more than that when, when it was at its peak over the year.

58
00:10:47.250 --> 00:11:02.000
And if we look at the five year, it's only up, if you'd bought Iron in the last bull run with Bitcoin five years ago today, you would, or no, this is post IPO, you would only be up 53, 54% anyway. Only.

59
00:11:02.050 --> 00:11:15.470
But, you know, compared to being down. Compared to what we've seen. Yeah. Right? Well, Charlie, we'll, uh, leave that there, and we have Jay Patel on next. But first, a word from our sponsor, CleanSpark.

60
00:11:15.790 --> 00:11:30.650
[gentle music] We are CleanSpark, America's Bitcoin miner. A publicly traded company with the largest operating hash rate. Powered entirely by self-operated infrastructure across four states.

61
00:11:32.470 --> 00:11:43.950
This is our proof of work, and we are setting the standard for what's next. Learn more about the intersection of energy and Bitcoin at cleanspark.com.

62
00:11:45.590 --> 00:11:56.130
[upbeat music] If Bitcoin's actually the best money, and it's the thing that people should accumulate, and it's the best risk-adjusted asset, I lose zero sleep about whether or not that's gonna happen.

63
00:11:56.330 --> 00:12:06.870
I, I just ask the question of when. It's literally matrix math that you're running on large pieces of data. The Bitcoin miners can absorb that energy.

64
00:12:07.130 --> 00:12:22.840
And, and, and in many ways this feels like a second bite at the apple to build a new internet. All right, we got Jay Patel in the wings.

65
00:12:22.900 --> 00:12:33.020
Gonna bring him up and talk private credit. Jay, welcome back to the show. How you guys doing? Doing good, man. Doing good, man. Great to see you.

66
00:12:33.140 --> 00:12:50.590
Very timely because news just hit today that Blue Owl had to cap its redemptions at 5%, I believe for the second quarter in a row. Now, redemption requests were down in Q2 versus Q1. Q1 was 5.4 billion.

67
00:12:51.340 --> 00:12:58.430
Q2 was 4.7 billion, still a lot, and they're seeing the most redemption requests of any of these big cr- pr- private credit funds.

68
00:12:59.100 --> 00:13:11.940
So that leads me to ask the question, i- is there any reason to believe, Jay, that this is a wider problem, and is it specifically tech-focused with these demands for redemptions now? I'd say, I'd say twofold.

69
00:13:12.040 --> 00:13:14.480
I don't think it's a Blue Owl specific problem.

70
00:13:14.680 --> 00:13:26.760
So we've seen a bunch of departures from BlackRock funds in terms of executives leaving, you know, continued downward performance of their public BDCs, so I don't think it's, like, a specific Blue Owl problem.

71
00:13:27.360 --> 00:13:37.040
I will say it was interesting their phrasing that, like, "Oh, things are better." Um, remember most of these funds we talked about a quarter or two ago, that they have a 5% redemption gate.

72
00:13:37.100 --> 00:13:46.800
And so Blue Owl is effectively saying things are better because of instead of 20-plus percent of the fund trying to redeem, only 18% is trying to redeem. I don't...

73
00:13:46.880 --> 00:13:54.900
You know, my mind, that's like, "Oh, half our house isn't on fire." It's like, half our house minus one room, as if that's a better situation to be in.

74
00:13:55.320 --> 00:14:04.840
Um, but I guess, you know, they're attempting to spin it in a positive light however they might. I think they do have a separate tech-focused vehicle which is doing even worse.

75
00:14:05.200 --> 00:14:08.400
Um, I think it's, like, 30-plus percent redemption requests.

76
00:14:08.480 --> 00:14:19.840
So I don't think it's tech-specific, but I do think that, you know, if we go back to what we were talking about last quarter, like, the software loans were really where the pain originated.

77
00:14:20.160 --> 00:14:31.860
Um, and I think the market is still honing in on that. But obviously, like, all of these funds have, you know, dropped the gates. You know, Apollo, Blackstone, Blue Owl, you know, even when some of the Goldman funds.

78
00:14:31.920 --> 00:14:43.020
Like, so I think it's an industry-wide thing. Um, but you know, if you think about 5% gate every quarter, it could take, you know, years to play out. Yeah, and I believe they have enough cash.

79
00:14:43.080 --> 00:14:49.450
It was something like 12 billion, so they have enough runway if that 5% gate, if they just keep that 5% gate up for at least a year.

80
00:14:50.360 --> 00:15:00.490
And for the tech-focused fund to your point, there were 1.1 billion in redemption requests, 38% of the fund. Um, so to, just to your, your point earlier.

81
00:15:01.400 --> 00:15:07.820
And 70% of Blue Owl's private credit funds are tech-exposed per their Q4 disclosures.

82
00:15:08.270 --> 00:15:20.140
So I wonder if that's why I wanted to ask the question about tech specifically because if we are looking at tech broadly, and Blue Owl is, seems to be one of the most concentrated of these business development co- um, business development companies, then

83
00:15:21.740 --> 00:15:31.860
it, it would make sense to me that, that, that these, the tech-focused, uh, BDCs are the ones that are seeing the most stress. And that Blue Owl- Yeah... would be under the most stress because of that.

84
00:15:32.660 --> 00:15:41.320
I think the, the tech software-focused BDCs are definitely the most exposed. But remember last quarter, there were a bunch of reports that came out that even

85
00:15:42.260 --> 00:15:53.640
these highly concentrated funds were underreporting their exposures to tech because they were classifying, say, healthcare software as healthcare exposure- Mm-hmm... rather than just tech or software exposure.

86
00:15:53.720 --> 00:16:08.400
So, you know, um, I think all of them have a material amount of software exposure given that that's where the industry really grew so rapidly over the last couple of years. Um, you know, 2021 to 2024.

87
00:16:09.120 --> 00:16:15.080
Um, but you know, obviously the ones that have a higher concentration are probably in a even tougher spot.

88
00:16:16.200 --> 00:16:28.040
So Jay, you have told the story a couple times on here that the private credit, credit, like original sin, if you will, was like lending too heavily to software and then trying to kinda,

89
00:16:28.960 --> 00:16:38.800
uh, get it all back or repaint over this by lending to NeoClouds. And then we have the Meta NeoCloud announcement, and NeoClouds are down bad the past two days.

90
00:16:38.940 --> 00:16:54.720
Basically, perhaps like a DeepSeek moment or repricing of the sector. What does this do? Is this, uh, making the NeoCloud lending risk, uh, bigger? Yeah.

91
00:16:54.880 --> 00:17:04.200
Even- So remember a quarter ago, all of these executives switched up their tone of voice, and they were like, "You know, we're not doing software loans.

92
00:17:04.260 --> 00:17:12.560
We're doing loans against," uh, they, they called it like the halo trade, low obsolescence, heavy assets, but it was really just data center financing.

93
00:17:12.680 --> 00:17:23.600
Um, and, and a quarter ago, that seemed like, you know, the trade to be in, right? Uh, CoreWeave, Meta had a facility with Blue Owl. Basically, every NeoCloud was raising debt capital.

94
00:17:23.940 --> 00:17:33.400
There's basically infinite demand for GPUs. Um, and, and you know, it's not like that demand has waned significantly. Um, but I think that's basically where they pivoted their focus.

95
00:17:33.860 --> 00:17:41.100
And now you've seen over the past week with, you know, Meta's announcement, CoreWeave bonds are trading down somewhere between 5% and 6% just this week.

96
00:17:41.260 --> 00:17:49.620
And CoreWeave, you know, that chart that you guys mentioned, they're, they're probably the most well-positioned of any of the NeoClouds.

97
00:17:49.690 --> 00:18:09.900
So you have to think that some of these private, more thinly traded bonds or private loans that these private credit funds have made to, um, not just the NeoClouds, but everyone along the kind of stack in terms of building out data centers, whether it be infrastructure companies or otherwise, uh, they've gotta be in a tougher position than CoreWeave.

98
00:18:09.920 --> 00:18:19.080
And if CoreWeave is down 5%, 6% for the week, I'm sure the others, you know, if you actually mark to market, are down worse. Um, look, I don't know how that plays out longer term.

99
00:18:19.200 --> 00:18:23.980
You know, maybe this is a market overreaction to, you know, Meta's news.

100
00:18:24.460 --> 00:18:39.514
Um- But obviously, if this was the direction that these private credit funds are trying to sh- pivot their focus to, it doesn't look too great in the long term to be able to tell your investors the story that, "Hey, we're out of software loans, and we're now into, you know, data center financing."

101
00:18:40.044 --> 00:18:40.424
Um- Yeah.

102
00:18:40.904 --> 00:18:56.304
Yeah, and with regards to whether or not this is a blip or a kind of crack in the model moment, I can't help but think about the demand story alongside this, because couched in what Meta's, Meta's announcement is like, yeah, now the neo clouds have to compete with Meta.

103
00:18:56.444 --> 00:19:08.644
They're already competing with xAI, 'cause xAI di- kind of fr- front-ran NVIDIA with regards to pivoting their model to sh- selling compute. But to me, it kind of

104
00:19:10.104 --> 00:19:18.944
indicated that, and to most people, that maybe the demand story here isn't as robust as most people think it is. And we're gonna cover this later on in the story too, in the show too.

105
00:19:19.744 --> 00:19:25.874
But NVIDIA's revenue share deal today, saw a lot of people calling it an evolution of the business model.

106
00:19:26.404 --> 00:19:39.004
I can't help but look at it as deepening circular financing that may indicate that demand is really not as robust as most of these tech giants would like us to think. W- what's your read- Yeah. I-... on all of that?

107
00:19:39.184 --> 00:19:44.924
I, I, I'm, I'm definitely on your side there. You know, there, there's, there's...

108
00:19:45.064 --> 00:19:53.324
One thing that you, you know, I think a year ago folks were, you know, shooting down the circular financing argument saying, "Hey, it's just NVIDIA supporting the ecosystem.

109
00:19:53.364 --> 00:19:57.164
They're making small equity investments in these companies. It's a small portion of the capital."

110
00:19:57.564 --> 00:20:08.204
We've basically gone to the far end now where NVIDIA will backstop you with their balance sheet if you split your revenue with them. I don't see how you can paint that as anything other than circular financing.

111
00:20:08.214 --> 00:20:20.784
That's not to say that it isn't a good business decision, right? If there's sustained demand, maybe it works out. Um, but if you're one of these lenders, you have to, you have to wonder what that means.

112
00:20:20.944 --> 00:20:36.564
If, if NVIDIA's, you know, stepping into this position and willing to do this for some of the neo clouds, um, does that tell you that they themselves anticipate that there might be some headwinds, and that's why they're doing this, so that their partners can get better financing terms on these GPUs?

113
00:20:37.084 --> 00:20:44.904
And then the flip side is, do you wanna be lending to a neo cloud who doesn't- Yeah... have NVIDIA as a backstop? Like, that seems like the worst spot to be in. Like- Right...

114
00:20:44.944 --> 00:20:47.524
you know, what if you don't have the NVIDIA stamp of approval now?

115
00:20:49.804 --> 00:20:59.784
And it, there's almost even a consideration to, well, if NVIDIA is quasi-financing these companies, you know, really, what is the credit worthiness for any [laughs] of these? You know what I mean?

116
00:20:59.864 --> 00:21:13.124
And, and, and, and if you see that NVIDIA is, is now glued to the hip with some of these, and also they're making money on the GPU sales, but now their balance sheet is tied to these folks, it's... I don't know.

117
00:21:13.284 --> 00:21:21.744
It makes me sweat a little bit. It makes me wonder if, if this is a kind of oh, shit moment. It will be an oh, shit moment in a year from now, but who knows? Yeah.

118
00:21:22.044 --> 00:21:36.164
The one defense that I've seen that I think, you know, the, an argument that actually has some pretty decent credibility is obviously, like, from NVIDIA's perspective, putting aside the ability to reduce the cost of capital for their partners to buy,

119
00:21:37.064 --> 00:21:46.384
um, you know, GPUs and build data centers, um, obviously they have an incentive for more of the build-out to be NVIDIA GPU focused.

120
00:21:46.864 --> 00:21:58.104
And, you know, the news that, um, you know, Google is gonna start selling TPUs, obviously OpenAI, and I believe Anthropic have their own kind of inference chip efforts, you know, Cerebras.

121
00:21:58.144 --> 00:22:09.124
Maybe this is more just at them trying to reduce the cost of financing NVIDIA GPUs versus other inference chips, rather than trying to backstop the whole industry.

122
00:22:09.784 --> 00:22:14.264
Um, but either way, I don't think it's a great look, and obviously NVIDIA probably understands that.

123
00:22:14.344 --> 00:22:20.624
So for them to take this step, you know, makes me think that they, they might see some cracks in demand behind the scenes.

124
00:22:22.104 --> 00:22:36.524
Uh, w- we're gonna switch topics for a quick hit with you, Jay, on the Stretch Drive SADA preferreds. This is a big, this is all of Twitter on Twitter right now.

125
00:22:37.284 --> 00:22:49.124
Um, and this, the, this tweet is a bit, it's a little bit cold now. It's two or three days old now. But this is Matt Cole, CEO of Strive, talking about their preferreds. Asking for market feedback.

126
00:22:49.344 --> 00:23:05.824
Uh, targeting $100 per SADA remains the objective, but some investors appear to believe Strive will n- always issue new SADA shares at 100, creating an, an effective cap. Um, and while short interest climbs on this,

127
00:23:06.844 --> 00:23:22.224
what's your take on the landscape of the, the two preferreds, Stretch and SADA right now? I'll, I'll say, you know, I think they've been better able to navigate the situation than I thought they would.

128
00:23:22.344 --> 00:23:28.964
I think they're both still trading at pretty deep discounts. But, you know, the commons recovered a little. Um, I'll say one thing.

129
00:23:29.744 --> 00:23:36.784
I think it sucks to be a common shareholder in either of these companies, because it feels like they don't give a shit about you, or at least they don't care about you as much as they should.

130
00:23:37.204 --> 00:23:49.884
Um, and I think specifically for Strive, like, look, I think, um, you know, it's mostly a confidence game, and I guess, you know, Saylor on the Stretch side has done a decent job of bringing back some level of confidence in the capital structure.

131
00:23:50.464 --> 00:23:56.384
But the preferreds only exist so far as they can benefit the common, right?

132
00:23:56.444 --> 00:24:10.364
Like, MicroStrategy and, you know, um, Strive, uh, you know, quickly thereafter only issued these preferreds so that they could raise capital on extremely favorable terms and buy Bitcoin to increase the Bitcoin holdings for the common shareholders.

133
00:24:11.404 --> 00:24:23.684
If you can't do that, like, you know, MicroStrategy and Strive are not some public utility that are going to provide Bitcoiners with stable 12% yields so that we can all bask in retirement, right?

134
00:24:23.764 --> 00:24:27.764
Like, that's not the purpose of these vehicles. And so, I don't know.

135
00:24:27.784 --> 00:24:40.980
From, from my perspective, the Strive announcement seems kind of odd, because if you're saying that you're not gonna issue at 100- What's the, you know, why even step in and raise the rate or, you know, defend the peg to...

136
00:24:41.160 --> 00:24:48.540
Or, you know, I know they don't prefer to use the term peg, prefer, you know, defend par or whatever you wanna say. Target price, yeah. Yeah.

137
00:24:48.580 --> 00:24:52.060
Why defend the target price if you're saying you're not g- you're not gonna issue out $100?

138
00:24:52.100 --> 00:24:58.720
If you're not gonna issue out $100, then this instrument doesn't really have that much utility for you, because you're not gonna be able to raise more cap.

139
00:24:58.780 --> 00:25:06.980
You're saying you're, "If I'm not gonna raise more capital using these preferreds." Then as a common shareholder, I'd be like, "Okay, fuck them." Like, let's, [laughs] you know, excuse my language.

140
00:25:07.020 --> 00:25:17.700
Like, let's, you know, put them aside. If we're not gonna use it to raise more capital, if they're happy with 70 cents on the dollar, 80, I don't care. I want more Bitcoin per share. So I don't know, it seems odd.

141
00:25:18.100 --> 00:25:31.540
But his logic is, I guess, sound, in that if you're saying you're going to get rid of that $100 ceiling, maybe that creates more pressure on the shorts, lifts some, you know, short interest, and then brings the price up.

142
00:25:31.880 --> 00:25:34.900
Um, there's some recursive logic there, but I can kind of see it.

143
00:25:36.100 --> 00:25:45.660
Going back to what you were saying about these companies signaling that they don't really care about the common stockholders currently, it kind of reminds me of the Abraham Lincoln quote, "A house divided even- against itself can't stand."

144
00:25:46.300 --> 00:25:58.840
I, I do just wonder, specifically in a bear market, how sustainable this model actually is. Clearly this is the biggest crisis for them since. And it seems like Strive is weathering it a little bit better than Strategy.

145
00:25:59.720 --> 00:26:06.840
S- SATA is almost back to par and, and Stretch is still below n- 85, below 90.

146
00:26:07.620 --> 00:26:23.340
So- Yeah, I think, I mean, part of Stretch, um, I d- I don't know how big a portion of the sell pressure the last couple days has been, but obviously there's the unwind of the tokenized Stretch-backed stablecoins that we've seen on chain, where they're basically forced sellers of Stretch

147
00:26:24.340 --> 00:26:26.980
continuously to meet redemptions.

148
00:26:27.100 --> 00:26:45.200
Um, I, I do think that kind of gets to another piece of this, though, which is, like, these are really great bull market vehicles, where when people are throwing money at you hand over fist, they don't care about the terms or covenants, and you can tell them, "This is digital credit," but it has none of the protections of credit and it's just digital.

149
00:26:45.300 --> 00:26:55.160
Um, then yeah, they're great. You can buy a lot of Bitcoin. But like you said, in a bear market, I mean, what's the real utility? Like, it's not like they're gonna be able to issue a lot more Stretch right now.

150
00:26:55.680 --> 00:26:58.860
Um, Strive has definitely navigated it a little bit better.

151
00:26:58.940 --> 00:27:12.480
I think, you know, the one advantage they also had was that they had moved to that daily dividend model a lot earlier, um, which, you know, maybe helped them from a, just a public appearance perspective in terms of being more preferred shareholder friendly.

152
00:27:12.540 --> 00:27:29.180
But yeah. Jay, appreciate your insights. You, uh, you do way better than me fumbling around through Claude to try to understand a lot of these, uh, arcane finance topics. So appreciate your time and insight.

153
00:27:30.080 --> 00:27:47.910
Of course, no problem. Cheers. Thank you, Jay. All right. Take care, guys. Shout out, Jay. We are gonna keep talking about Anthropic announced a new chip. Surprise, everybody's going vertical.

154
00:27:48.760 --> 00:27:59.280
Then we are gonna wrap this up, again, a redux on NVIDIA and their rev share program. Before we go to Anthropic, a word from our sponsors, Luxor.

155
00:28:00.040 --> 00:28:11.000
[fireworks] This episode is brought to you by Luxor's Commander, Bitcoin miner management software for enterprise operations.

156
00:28:11.520 --> 00:28:16.990
Commander gives you real-time fleet monitoring, bulk remote commands across your fleet, and Intelligent Miner.

157
00:28:17.400 --> 00:28:29.520
That's an automated profitability engine that runs every five minutes against live hash rate markets and power markets. ERCOT backtests show, uh, [laughs] excuse me. ERCOT backtests show

158
00:28:30.460 --> 00:28:42.300
20, 10% improved profitability with intelligent mining versus binary mining. Commander Pro is $100 per megawatt or a 25 basis point pool fee adder, and you can try it for free for 60 days.

159
00:28:42.360 --> 00:28:56.600
So if you're interested in learning more, go to luxor.tech/commander to get started. All right, so I'll, I'll, I'll do a little intro on this, and I'll throw it to you, Colin.

160
00:28:57.580 --> 00:29:13.440
So Anthropic has begun early stage work on a custom AI chip and has held talks with Samsung Electronics as a potential manufacturing partner, per The Information. It's early, so we don't have a ton of info on this.

161
00:29:13.740 --> 00:29:22.960
Um, no details on design or, like, how any of it's gonna be manufactured has dropped. Um, and it's also not guaranteed that the project will proceed.

162
00:29:23.080 --> 00:29:36.960
However, it is, Anthropic is reportedly looking specifically at Samsung's two nanometer process and its advanced packaging facilities. Colin, I'll tag you in to continue the story.

163
00:29:37.620 --> 00:29:46.690
I would just wanna know what they are gonna call this. OpenAI's is Jalapeno. Yeah. Is this gonna be Habanero? Are they gonna try to one-up the other frontier model? [laughs] The ghost pepper.

164
00:29:46.700 --> 00:29:59.560
Yeah, this scoop is coming from The Information. I believe it was s- their scoop. Information, by the way, really batting a high average with some very good inside scoops on a lot of these companies.

165
00:30:00.300 --> 00:30:12.260
So hats off to the team there. And this comes on the heels of Anthropic recently hiring Clive Chan, an early member of OpenAI's own custom chip team, as part of this engineering build-out.

166
00:30:12.280 --> 00:30:20.960
And like we were just saying, this mirrors what OpenAI is already doing, where they are building their own inference ASIC chip called Jalapeno.

167
00:30:21.600 --> 00:30:31.860
Google, Amazon, Meta, and Microsoft have also built out proprietary silicon to cut their dependence on third-party suppliers. Th- those are not AI chips, as I understand that, uh,

168
00:30:32.920 --> 00:30:44.172
m- m- most of that, of those build-outs for proprietary chips are actually for- How should we call it? Old-fashioned data centers, antiquated data centers, vintage 2001 data centers.

169
00:30:45.172 --> 00:30:53.052
But the interesting thing for me is what Information estimated for Nvidia's market share. That's 74% of the AI market,

170
00:30:54.092 --> 00:31:04.232
and I wonder how much we will see that getting eaten into over the next few years as these AI ASIC and fringe chips become potentially more widespread.

171
00:31:04.292 --> 00:31:23.552
Now, as we covered, Charlie, on one of our prior broadcasts, the reason why we haven't seen AI take, or ASIC chips take off for AI specifically, has to do with combination of Nvidia's moat with their software stack as CUDA is the standard for training and building models.

172
00:31:23.572 --> 00:31:38.272
But also the fact that ASIC chips themselves may not be flexible enough for building a model, for training a model, but they could be just... They could be good or better for pumping out inference.

173
00:31:38.392 --> 00:31:49.812
And Nvidia itself, with the announcement that we're gonna cover next, basically hinted at the fact that as training models kind of take a backseat in demand to inference, we're starting to see a shift

174
00:31:50.772 --> 00:31:56.832
in the landscape of what cloud providers are providing and what end users need.

175
00:31:56.872 --> 00:32:11.072
So I wonder how much the inference story taking over the larger market share for AI compute generally, how much that will shift the scales in favor of Anthropic and OpenAI here. Yeah. Um, I...

176
00:32:11.232 --> 00:32:15.012
There's really not a whole lot of information on this right now, so we can only have limited takes.

177
00:32:15.712 --> 00:32:30.702
I will kind of attempt at one, which is, um, you know, as we cover this story of open models, open weights, Chinese, cheap Chinese models really catching up to these frontier models, uh,

178
00:32:31.652 --> 00:32:36.292
you have to wonder, where does the moat come from? Where does the moat for these frontier models come from?

179
00:32:36.912 --> 00:32:43.132
Um, because how much of a moat can exist if you're only a few months ahead with, with a moat, with a cutting-edge frontier?

180
00:32:44.092 --> 00:32:52.952
Um, maybe the moat is in the relationship between the hardware and the model, and this is where CapEx comes in. This is where the chip integration comes in.

181
00:32:53.512 --> 00:33:06.552
If you've got a proprietary AI chip that you d- that just really works with your model, and you have, uh, IP and that deployed, and you've spent years building and, and rolling that out, that's a moat.

182
00:33:06.712 --> 00:33:08.232
That is a moat like nobody's business.

183
00:33:08.312 --> 00:33:23.932
So, uh, maybe the, maybe the next move is, uh, like may- the ne- the next, you know, step or, or shift in meta of the A- like frontier model meta is just frontier models, uh, are run on proprietary chips.

184
00:33:23.992 --> 00:33:33.212
That could be a thing I could see, and that, um, is a pretty compelling moat thesis. Again, I'm kinda scrambling, kinda trying to imagine, craft a narrative here.

185
00:33:33.332 --> 00:33:49.372
But it's very clear that Anthropic has been pretty agnostic to whose compute they're running on, whether it be AWS, Google's TPUs, their... It'll be Groq's Colossus, uh, Colossus. I mean, it, it... They've been...

186
00:33:49.412 --> 00:34:01.002
It's a- any compute that they can have, 'cause they're very compute constrained. Um, this would be a deeper integration, so we'll have to see. That's kind of the only real take I have.

187
00:34:01.072 --> 00:34:05.812
I've tried to pull some tweets about this, but, like, there's really not, they, they really not the- There's not too much.

188
00:34:05.852 --> 00:34:16.412
The only other take I have is going to Samsung is interesting, and there are probably multiple reasons for that, TSMC being the premier semiconductor manufacturer.

189
00:34:16.421 --> 00:34:25.882
But also, if you're the king, then everyone's coming to you for favors. Samsung, it has much smaller market share. They are competitive in some ways.

190
00:34:26.592 --> 00:34:39.382
MicroBT, for instance, uses Samsung chips, and MicroBT may not be able to compete with Bitmain on efficiency for Bitcoin miners, but they're pretty darn close, and people love MicroBT machines specifically for durability.

191
00:34:39.412 --> 00:34:46.232
So going to Samsung obviously carries risks in the sense of maybe they don't deliver on the two nanometer spec, and then this whole thing goes to waste.

192
00:34:46.632 --> 00:34:54.412
I would imagine that that is a risk, obviously, Anthropic is very privy to, and they've done their research. But there's also an argument to be made that if you...

193
00:34:54.561 --> 00:35:04.292
For Samsung, this is a huge boon for having a frontier model coming to you to build a chip like this, and it also offers an opportunity for these companies to grow together.

194
00:35:04.712 --> 00:35:15.712
And if they really do pull this off, then for Anthropic, Samsung's always probably going to give you priority if you help slingshot them into greater relevance within this arena. So... Yeah.

195
00:35:16.112 --> 00:35:28.592
Um, as much as I do like TSMC, kind of a great narrative and story and crazy company culture over there and geopolitical significance, it's good to have a number two foundry out there. Uh, it's,

196
00:35:29.892 --> 00:35:40.392
uh, this is, uh, in the, in a world where this is the- One of the only-... bottlenecked industries. Yeah. It's not just- Maybe a world where it's the only foundry that the West has access to- Ah, yeah...

197
00:35:40.402 --> 00:35:50.972
if a certain geopolitical event comes to pass. Well, TSMC's got their-- They bui- they're building a foundry in, in Arizona, which is kicking off to a rough but effective start. Okay.

198
00:35:51.732 --> 00:36:01.892
We are gonna wrap up with this story and close out with the Nvidia rev share news coming to you right after a word from our sponsor, Ligos.

199
00:36:02.792 --> 00:36:19.372
[fire sizzles] Hedge funds and lending desks are reeling with Bitcoin just floating above 60K. If you're using a Bitcoin-backed lending service, do you know if your Bitcoin's safe?

200
00:36:22.792 --> 00:36:36.612
Oh, you muted yourself. Right. FT or Celsius victim, if you are using a Bitcoin-backed lender, do yourself a favor and check out Ligos Finance. They are our preferred Bitcoin-backed lender here at Blockspace.

201
00:36:36.952 --> 00:36:44.352
They use Bitcoin native smart contracts to make sure you are always in control of your stack and no one else has access to your Bitcoin.

202
00:36:44.856 --> 00:36:57.096
With Lagos, there's no rehypothecation, there's no wrapping, no bridging, just Bitcoin-backed lending with rates as low as 8.5 to 10% APR. Go to Lagos.finance to learn more.

203
00:36:58.776 --> 00:37:15.396
All right, so kind of a doozy of a day for, uh, compute lending and compute markets, which frankly is a lot to cover, a lot of news to try to follow.

204
00:37:15.436 --> 00:37:32.036
So, um, this story is, uh, again, once again, the largest company in the world, NVIDIA, trillions of dollars large. NVIDIA offers startup customers chance to swap compute power for rev share. This story from CNBC.

205
00:37:32.916 --> 00:37:46.556
Chipmaker, chipmaker NVIDIA says it's entering revenue sharing agreements with fast-growing startups in a move that which will see customers swap access to compute power for a slice of future profits. Wait a second.

206
00:37:47.176 --> 00:37:55.256
These are very different types of, uh, uh, investments, compute versus future profits of NVIDIA.

207
00:37:55.316 --> 00:38:17.256
The artificial intelligence chip leader announced Thursday this partnership program, cloud-based AI firms, model builders, and other enterprises will share both product and cloud revenue with NVIDIA, which is positioning itself as an intermedia, h- intermediary helping startups gain direct access to full stack computing power by NVIDIA chips.

208
00:38:18.116 --> 00:38:24.876
Yeah, so- Tagging you in... there's a lot to unpack here, and I actually think that this story's kind of confusing if you don't know how all of the pieces are working together.

209
00:38:24.976 --> 00:38:37.506
So NVIDIA is basically guaranteeing a backstop for computers, for GPUs that they sell to AI la- or neo clouds and cloud pro- uh, cloud providers. And what they're saying is,

210
00:38:39.176 --> 00:38:49.616
"We're going to help match you with end users as well. We are going to give those end users token credits to use your compute,

211
00:38:50.556 --> 00:39:03.976
and then we are going to have a rev share agreement with you where you can pay us back with the profits that you earn from those end users, assuming that they materialize." So

212
00:39:05.196 --> 00:39:19.116
you, you know, NVIDIA is stepping firmly into a financier role with this in a way that has them double-dipping if it goes well, 'cause the AI cloud platform is buying the compu- or buying the computers, and then they're selling it to end users.

213
00:39:19.856 --> 00:39:28.216
But then they also get to have the revenue share on top of that f- in exchange for the backstop that they're giving these companies.

214
00:39:29.026 --> 00:39:37.936
And like, as we said with Jay, this is probably the clearest example of circular financing within the AI ecosystem.

215
00:39:39.016 --> 00:39:46.796
There's an argument to be made that having an equity investment in one of your customers happens in many industries and isn't that terrible.

216
00:39:47.376 --> 00:39:56.586
Obviously, some of that investment dollar, some of those investment dollars will come back to you as that, as that company uses your services, or in this case, buys your GPUs.

217
00:39:57.116 --> 00:40:10.676
But this is NVIDIA taking on the risk of these companies never actually materializing legitimate revenue. And I saw this framed a- as a l- a revolutionary arrangement. Yeah, the, the takes on- And if, if it goes well-...

218
00:40:10.716 --> 00:40:19.196
Twitter are con- the takes on Twitter are confusing, Colin, 'cause, uh, I struggle to see what the big deal is. Yeah, I mean, I think that

219
00:40:20.176 --> 00:40:31.046
the big deal is in the sense of like if the AI boom continues at the rate it has, NVIDIA gets to cash in twice on their GPUs. So that, in a bull scenario, this is great for NVIDIA.

220
00:40:31.556 --> 00:40:46.056
But what happens if your customer doesn't actually end up... What ha- happens if the cloud don't end up actually generating any revenue? And right now they have two companies that are piloting this with them, Sharon AI

221
00:40:46.976 --> 00:41:02.476
and, and a, a Singapore AI infrastructure company called Firmus. And so yeah, the question just becomes then what happens if revenue never materializes, and NVIDIA takes a huge haircut on backstopping these.

222
00:41:02.496 --> 00:41:11.856
They have to, they, they're guaranteeing a backstop on these GPUs. If they don't f- if they don't end up generating revenue for these companies, they're going to compensate them.

223
00:41:12.636 --> 00:41:19.116
So yeah, I have a hard time seeing a world in which this is good for NVIDIA in a bear scenario, obviously.

224
00:41:19.656 --> 00:41:35.176
And as we talked about with Jay, this creates kind of dual risk for them because their revenue is tied to these companies, and now their balance sheet, at least n- not in this sense, because I don't believe they're taking equity, but when they do take equity in some of these companies, their balance sheet can be at risk.

225
00:41:35.316 --> 00:41:45.645
But they're saying they're doing this also, this is the other part of this. They're, they're saying they're doing this because some of these companies are having a hard time securing financing.

226
00:41:45.696 --> 00:41:55.936
That, you know, potentially credit is starting to tighten and get a little more prudent as we see some of these notes in the secondary market from these corporate bonds taking a haircut.

227
00:41:56.736 --> 00:42:02.776
So what does that say about access to credit in the future for these AI builds? And also, what does that say for the actual demand story?

228
00:42:03.316 --> 00:42:14.036
Because if you read the meta news in this in concert with each other, you can paint a picture to where the tech companies aren't being totally honest about the level of demand.

229
00:42:14.156 --> 00:42:21.816
Or another way to put it is demand is not evenly distributed, because there are so many different players within this ecosystem, right? Yeah.

230
00:42:22.936 --> 00:42:34.416
Uh, I don't have a whole lot of insight to this, but let us know if you've got opinions. Let us know in the comments. Otherwise, thank you for listening to Blockspace Live.

231
00:42:34.476 --> 00:42:45.916
We do this every weekday except for holidays at 1:00 PM Eastern, featuring quick hits on AI, data centers, Bitcoin mining, emerging tech, and markets.

232
00:42:46.056 --> 00:42:55.956
If you like what you hear, you'll love the newsletter, newsletter.blockspacemedia.com. And it turns into a podcast shortly after we hit end on the live stream.

233
00:42:57.096 --> 00:43:13.926
This show is brought to you by CleanSpark, NASDAQ-listed ticker CLSK. Thank you, CleanSpark. I'm Charlie. I'm Colin. And we'll see you on Monday, because tomorrow's a h- national holiday in the United States. Happy 250.

234
00:43:14.906 --> 00:43:20.835
[outro music]