← Blockspace
Blockspace
Hut 8’s $19.6B Beacon Point Deal, IREN’s $2.8B AI Cloud Contracts, Uncle Same Moves on Kimi K3
00:50:16
Loading summary
Transcript80 lines
- [00:00]
A
What's up, y'? All?
- [00:00]
B
Welcome back to Blockspace live, presented by CleanSpark.
- [00:04]
A
Charlie, we're so back. We are so back, baby. We are so back.
- [00:09]
B
After last week's bloodletting, markets are green today. So much so that I think like 500 billion roundabouts has been added to the stock market in the opening hours of this morning. So for today's lead segment, a quick market update as well as what FactSet is saying about earnings expectations for the S&P 500 cohort now that earnings season is upon us. Following that, we will jump to the deal flow. Stocks back up deals are coming through. Hut 8 just fully commercialized its Beacon Point facility with the same tenant for phase one. As for phase two. That for our second story following that, we have another deal coming out from iron. Multiple deals, actually, according to this press release for 2.8 billion billion in contracted value for its AI rollout. And at the end of the show, our boy Charlie over here is going to give us the latest on Kimmy K3 and whether or not Uncle Sam will put the ban hammer on the model that shook the AI stock market.
- [01:11]
A
That's right, Block Space goes live every weekday at 1pm Eastern, featuring quick hits on AI data centers, markets and emerging tech. We also cover crypto and bitcoin occasionally. Right now that market is slow and your AI stocks are ripping. If you like reading and listening about the landscape of PowerShells, Neo Clouds and AI data centers, go to BlockSpace Media. That's our website, to check out all of our written and video and podcast content. BlockSpace Media, you can also get our newsletter@newsletter.blogspacemedia media. That newsletter goes out every single day, recapping what happened in the markets and what happened on the live stream. This show is brought to you by CleanSpark. Nasdaq listed ticker CLSK. We're on CleanSpark later on in the show. Colin, let's talk markets.
- [02:09]
B
It's time, Charlie.
- [02:11]
A
Yeah.
- [02:12]
B
You still have time to trade your way out of the future underclass.
- [02:17]
A
With just a little bit of leverage and, and the right positioning and just
- [02:20]
B
a little bit of leverage in a Robinhood account, you too can lose all of your money. No kidding. But yeah. So a green morning for. For most of the stocks that we cover this morning.
- [02:33]
A
Yeah, most of the stocks we cover. I mean, overall right now it's mixed across the S and P, but the majority of the gains have been in the sector that we cover on this show. That's PowerShells, NeoClouds, data centers, yeah, it
- [02:48]
B
looks like the S and P kind of having a day today. It's barely up 20 basis points, but it surged on market open and then it tanked and it's kind of like chopping sideways. NASDAQ is up about half a percentage point and Dow Jones is down. Kind of makes sense because oil is surging given the fact that the umpteenth peace deal in Iran didn't work out. Shocker, no one saw that coming. But overall, like you said, Charlie, our segment is doing okay. I'm going to pull up diametrics here. I think you just had it up, but I just want to highlight a few names, or rather sectors. Alrighty. So PowerShells are up 9% according to the index here on Diametrics. NeoClouds up 6%, data center real Estate Investment Trust just up half a percent. Hyperscalers up 0.2%, and a few other smatterings of negligible gains from AI, grid and power equipment and construction up 1%. So a big rebound after, like I said, last week's bloodletting.
- [04:06]
A
Can I jump in and look at some of the specific stocks within the PowerShells here? If you expand here, I've actually got it here.
- [04:14]
B
Yeah, go ahead.
- [04:15]
A
Look at this. Mara up 10. Clean Spark up 13.7%. Cipher 17, Hut 11, Galaxy 10, Saluna with the coming over the 23 and a half Greenridge, I guess 13 and a half. But, like, it's just big gains across the whole sector. So, yeah, this is a fantastic day. If you were getting a little antsy about the compute. The compute crunch.
- [04:47]
B
Yeah. Especially if you bought the dip on Friday.
- [04:50]
A
Yeah, you should have bought the dip. And this is the thing is, like, if you look back at our content and our coverage on Friday, it was pretty much like the markets overreacted. You know, Kimmy is a deep Seek moment. But look what happened after Deep Seek. The neoclouds and the compute continue to crunch and soar. So I think the market is acting like this is another Deep Seek moment where it's a temporary blip on the timeline.
- [05:21]
B
And Deep Seek was like a year and a half ago, too.
- [05:24]
A
Yeah.
- [05:26]
B
Kind of wild. And to see that boomerang come back in the form of Kimmy and also the market kind of shrug it off. I was kind of scratching my head this morning, Charlie, wondering what was the actual catalyst for this. And it could be just the fact that things were oversold on that news. There's also earnings coming up and a lot of analysts expect, as we've Seen with prior quarters, just phenomenal beats across revenue and earnings. I'll cover some of those expectations here in a second. But I was wondering, okay, do we get some developments in the Iran war? Is that actually coming to a close?
- [06:01]
A
My opinion, you know, my opinion is that the market is just very reactionary and there's a whole lot of traders who, who can't zoom out and take their hands off the keyboard. So if they see a blip or they see some negative news, they do away with the entire compute crunch thesis and they sell off. The market becomes really, really reflexive. However, I think it's funny because if Twitter is any indication, Twitter kind of started recognizing this towards Thursday and Friday as people realized that you can't just run Kimmy like on a small little cluster or home cluster. These are giant. Giant. It just doesn't work on the existing architecture of these data centers. And when Moonshot, the company who owns Kimi, said, we are oversubscribed. We have to turn off new subscriptions, which we'll get into in segment three here or the last segment. I think people realized, oh, wait, there is huge demand. So Jayvon's paradox will continue to play out as it looks like.
- [07:12]
B
And also potentially quelling the fears of this reducing cost to run the compute. Right. As I understand, that was largely what led to a freak out with the neoclouds, because most people, including myself, don't fully understand the complete ins and outs of running those compute clusters and what's actually needed on the hardware side to service demand. I mean, people get paid a lot more money than we do to figure this out at these companies. But I think within the fog of Kimmy's release, and within the fog of all of the Chinese AI stuff, there's this assumption that somehow their models are cheaper because they were trained more cheaply because they're piggybacking on frontier models in the us. But that doesn't mean that the inference necessarily is cheaper or it's not as cheap as being marketed by certain bulls for Chinese AI or for doom pointers. But going back to the Iran conflict, the AI stocks are just seemingly shrugging a lot of this off. I think that the market has kind of become inured to the war in Iran right now. Notwithstanding an inflation shock from elevated oil prices, oil got down to its lowest level since the start of the war in early July when the peace deals were going on. Now I believe Brent crude got back up to like 90. WTI is close as well. What is WTI at right now?
- [08:40]
A
It's back up depending on 83. Yeah, 82, 83. It did hit 90 previously, but yeah.
- [08:50]
B
And you know, at the end of the day, that obviously could have incredible shockwave effects for the global economy if it stays too elevated for too long. But as far as I can tell, a lot of the markets kind of just shrugging off escalation, at least within the high growth AI stock sector. And so the only thing I can really look at, Charlie, as to why this bounce is happening now is a combination of what you said, people realizing that Kimi doesn't have the teeth that they thought it did. But also, as I said, earnings season is kicking off and FactSet has the following expectations for earnings growth. They're expecting 24.7%, which is up from 22.5% estimates in the week prior and 23.2% at quarter end. So June 30th. So these are again, expectations. These aren't the actual blended earnings rate for S&P 500 companies. If this holds though, it would be the second straight quarter of earnings growth above 20%, which is crazy. And the seventh consecutive quarter of double digit earnings growth, which is also nuts. Revenue growth for the blended average for The S&P 500 is estimated at 12.8%, which would be. If this holds, it would be the highest revenue growth since Q2 2022 and the second straight quarter of double digit revenue growth. So you can almost look at this bounce as looking forward to earnings season, seeing that stocks were sold off and recognizing that if these companies do beat to the extent that people are expecting, then what we saw at the end of last week would be fire sale prices for the near term at least. It's kind of hard to make the argument that it's fire sale, at least for the indices. Right. Because the S and P. The S and P didn't take that much of a hit last week. It fell 1.6%. Nasdaq fell 3%, roughly. So that's a pretty big hit. But for these individual names that people were trading with an AI, a lot of them are down quite significantly year, year to date. Like iron Drew down, I think as much as 50%.
- [10:48]
A
Yeah, drew down 50%. And, and to point out that, you know, while I am bullishly, you know, bullion and talking about the market recovery, things soaring right now, I believe like iron and core weave are down on the year, year to date and the sector is still actually a little bit mixed year to date. Nebus up Almost, you know, 97%. WI fiber up 49.4%. So the sector is pretty mixed year to date. However, it is kind of starting to across the power sales and neoclouds, they are kind of trading together right now. So the sector overall is recovering.
- [11:26]
B
Look at that volatility. Nibia is down 35% over the last month. Yeah, this is token price action, man.
- [11:35]
A
Yeah, Wild.
- [11:38]
B
All right. I mean we can wrap that unless you have anything else to say and we can.
- [11:41]
A
No, I mean if, if you're, if you're in crypto, pivot to AI stocks.
- [11:46]
B
I mean if you're an AI stocks, pivot to bullets. Depending on your, your take on things,
- [11:52]
A
pivot to self defense. Heck of. Fine. Okay. We are gonna keep going. We have a packed deal ticket today. Hot eight, then iron and then we're going to talk about is Trump gonna ban Chinese models? Before we go to our next story, a word from our sponsor, CleanSpark.
- [12:11]
B
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. This is our proof of work. We are setting the standard for what's next. Learn more about the intersection of energy and bitcoin@cleanspark.com. 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'd lose zero sleep about whether or not that's gonna happen. I just ask the question of when. It's literally matrix map that you're running
- [12:58]
A
on large pieces of data so the bitcoin miners can absorb that energy. And in many ways this feels like
- [13:04]
B
a second bite at the apple to
- [13:05]
A
build a new Internet.
- [13:15]
B
All right, Charlie, first big story of the day. Hut 8 commercializes 1 gigawatt Beacon Point campus with second AI lease. This is coming hot off the wire this morning. This is coming to y' all from block space and the TLDR is this. This is a 352 megawatt 15 year triple net lease valued at 9.8 billion over its base term and includes a 3% annual base rent escalator. It's the same tenant as phase one. Unnamed tenant. We'll get to that in a second because there's a little bit of process elimination we can do here to narrow it down to a handful of names, but it's an unnamed tenant, the same tenant as phase one. So that tenant will now occupy 740 megawatts of critical IT within Beacon Point. This Beacon Point data center is 1 gigawatt so the rest of that will be made up with not critical IT load, but supporting load for all the other infrastructure in the site. And according to this Release as well, Hut 8 has said that it has revamped its architecture for Beacon Point. Specifically it's designed to Nvidia's DSX reference architecture. And the phase one hall is also going to be reworked for this, if I understand the press release correctly, for what they advertise as a 57% capacity boost. So more compute power, same space. A few other things to Note here. Now Hut 8's total contracted IT capacity across its entire data in their portfolio is 949 megawatts of critical IT. 704 of that is at Beacon Point and 245 is at River Bend. With 1330 megawatts of utility capacity. The aggregate base term contract value across this portfolio now is 26.6 billion with an average annual NOI. With average annual NOI expected above 1.75 billion. Charlie, any second takes before I do a little guesswork at who our tenant could be?
- [15:23]
A
No, I don't have a take. I think the, the big question is let's do some guesswork. We're gonna, for the listeners, we're putting on our speculation hat here and we're due process of elimination because there's only a handful of potential tenants. So let's think through who could it be.
- [15:40]
B
Colin, we're gonna do a little divination. So break out your tarot cards or your turban or your crystal ball. Whatever suits you in this moment.
- [15:47]
A
So
- [15:49]
B
in financing docs for the first phase of Beacon Point, Hut 8 says the data center facility will be leased to a tenant that is a high investment grade company. In other words rated AAA minus or higher as of the date here of the date being when when they release this SEC filing. So that leaves more or less the Mag 7, right? And so it could be any of those companies, but probably definitely not Apple because Apple's not doing infrastructure spend and it's probably or almost certainly not Google because during an earnings call over the last few years or during an earnings call over I think in May 2026. So the Q1 earnings call, Hut 8's Piper Sandler noted that management had confirmed the tenant at Beacon Point is not anthropic Google or Fluid Stack specifically making that point to say hey, we've got new demand coming in from new customers so we can rule out Google. Because Sandler said that it's not Google that really leaves Amazon Nvidia Microsoft or Meta for this site. And Meta rating is the weakest of the three, I believe. Moody's doesn't quite rate them at the threshold that they need. But they didn't say that, you know, their aggregate credit rating is AAA minus. They just said it's what they have a credit rating for that. So it could come from Fitch, it could come from Moody's or S P. So I mean, I'm, I'm guessing that it's probably that it's one of these for Amazon, Nvidia, Microsoft or Meta. And our boy Rittenhouse Research also popped up on the timeline today, kind of venturing the same. Although he put Google in there. I don't think he had seen the earnings call transcript. So, yeah, he says here in his tweet quote, financing Doc suggests either Nvidia, Amazon, Microsoft, Google or Meta. So we don't know who it is, obviously. We actually asked Asher Ganut this on one of his recent forays onto the live stream and we were like, why are you not disclosing this? And he had a pretty good answer, Charlie.
- [18:07]
A
Yeah, I'm going to misquote him. But it was basically, you know, who it is. But we want to protect our deals and our existing clients and customers.
- [18:18]
B
Yeah. The rationale being when these Bitcoin miners were making these announcements for the tenants at their sites, I mean, there's a kind of a chicken and egg thing where you kind of have to be like, hey, this is actually real. Here's an investment grade company, one of the most heavily capitalized in the world, backstopping this, like with Terra Wolf and Cypher on the fluid stack deals, things like that, there's obviously an ethos that they're evoking with that and a prestige. But the fact of the matter is, as Asher pointed out, the counterparties don't always like that because depending on the deal structure, they might have other Neo clouds or other powershells that they're working with who say, hang on, your terms with them are a lot better than your terms with us. So they don't want to give away any sort of competitive advantage with announcing these things, or they don't want to rock the boat with other clients, so to speak.
- [19:13]
A
Yeah, I'll add a little color here. So we have Amazon on the short list here, but if we look at the actual architecture of what Amazon likes to run on, they have their own trainium silicon, and that's not what this build is. In fact, if I'm understanding the press release right, and the kind of refactor of the build. They're using the Nvidia DSX reference architecture. So I wouldn't be surprised if it's Nvidia because they've been acting a little bit more like a tenant these days and less of the, like just the chip vendor. So if I were a betting man, I'd.
- [19:55]
B
You would bet Nvidia. That's where you would put your prediction market bet for this one.
- [19:58]
A
Again, you know, it's what, like maybe 40 Nvidia 25 meta. Who knows?
- [20:06]
B
I mean Meta is also a strong candidate just in the moving.
- [20:10]
A
Yeah, yeah.
- [20:12]
B
So that, that's good, good color there, Charlie. I'm glad you looked that up.
- [20:16]
A
Yeah. So I, if you're just going to look at it's. If it's. If we consider the toss up between, you know, Nvidia Meta, Amazon probably on Amazon favorite Nvidia right now. So we'll see.
- [20:26]
B
We'll see. Tune in for updates.
- [20:29]
A
Tune in for updates. We have another big deal announced. Iron with 2.8 billion AI cloud deals. We'll go to that story after a word from our sponsor, Luxor.
- [20:51]
B
This episode of Blockspace Live is brought to you by Luxor's Commander Bitcoin miner management software for enterprise operations. Luxor's Commander gives you real time fleet monitoring bulk remote commands across your fleet. And intelligent miner, that's an automated profitability engine that runs every five minutes and adjusts your fleet's power settings to live energy and hash rate markets. Ercot backtest show 10% improved profitability with intelligent mining versus binary mining. Commander Pro is $100 per megawatt or a 25 basis point pool fee adder. But you can also try it for free for 60 days. So if you'd like to learn more, go to Luxor tech forward slash Commander to get started. All right, Charlie, let's hop on over to Iron. Second deal of the day. Iron signs 2.8 billion in AI cloud deals. Raises 2026 ARR target. This is coming at y' all from block space, the TLDR. For this deal they signed 2.8 billion in new AI contracts with multi year cloud service contracts. But they didn't give any specificity regarding the duration of these contracts. They also didn't explicitly name the tenants, but we can very easily by process of elimination point out who they are because they list all of their current tenants under current deals and then we just back that out from the ones that they've already announced before this. But this lifts its annualized run rate revenue for ITS Cloud, its AI cloud business from 3.7 billion to more than 4 billion. And they highlight in this press release that we covered that they have received 45% of that they've received prepayments for these customers covering roughly 45% of the associated GPU CapEx. So that will ease the massive CapEx requirement iron is going to have for these sites. They've got 7.6 billion in cash and cash equivalents as of 6-30-2026 but that includes 1.7 billion at restricted cash tied to GPU financing for its Microsoft contract at the Horizon site. It's going to be really expensive. It's going to be massively expensive because Iron is not only building the facilities, they're also buying the GPUs. They're going fully integrated NEO cloud.
- [23:23]
A
Yeah and if we look at the, I think the it's we if we want to compare it versus the Hut 8 story. Hut 8 has one customer for the whole site whereas Iron is announcing kind of a who's who of like the VC startups including like Perplexity Figure, AI Hume AI. I haven't heard of VAL AI but it's interesting that you know these are all well capitalized VC funded startups probably I would assume top of the pack here versus a single tenant but the single tenant conventional, better investment grade rating, kind of interesting business model. Iron has other tenants they can, you know that they that are investment grade but it's interesting to see like a new crop of maybe smaller venture, private, private venture tenants. So that's kind of interesting if you want to think about it from like, like a growth story or as these companies maybe go public like Figure is going to go public sometime big robotics companies. So like when is that, does that affect you know their deal? Do they get assigned an investment rating later down the road after they've, after Ivan's already locked them in for as a tenant. You know these are interesting questions.
- [24:48]
B
100% and it looks like the average contract duration for its portfolio is four years according to our coverage here. And that's drawn from investing.com I need to look into the docs to see if that's just for this specific announcement or for all of their AI cloud services contracts. But with regards to the customers, Iron on the press release says the following. They say Internet conspiring against me today. Man, here we go. Iron's customer base now includes Microsoft, Nvidia, Perplexity Figure, AI Together AI Fluid Stack, Firework AI fal, AI Hume AI and a Partridge in a pear tree. So I was wondering. Okay, well then that should be pretty easy to back this out right. Considering their current contracts for who this new round of tenants is. And it was pretty easy. So the customers they've already signed before today are Microsoft, Nvidia and together AI Fluid Stack and Fireworks AI. So it seems that the current cohort that is in this most recent 2.8 billion dollar aggregate deal is our perplexity figure AI foul AI, Hume AI and some one unnamed quote new leading AI developer, end quote. So a little secrecy from that one. Yeah, I actually didn't know. The only company I knew on this Charlie was actually Perplexity. So they're like what like an AI search engine? They're not. You can't really.
- [26:25]
A
Perplexity. Really until this time last year I would say Perplexity was in the same was on the tip of everyone's tongues when you would say OpenAI anthropic and perplexity in terms of like consumer AI products perplexity has fallen out as their not really considered a frontier lab. Although they've been really like great consumer product focused. These other companies are kind of the who's who of private venture figure is like the other robotics company right now the leading robotics company probably besides Tesla
- [27:02]
B
and it's private Foul AI is Generative Media Inference Hume AI is Voice and Emotion AI. You know, I guess making AI therapists.
- [27:14]
A
Yeah. Which we will all need after this is all said and done. But if you want to compare to strategies it's like if you can get these big deep pocketed venture backed companies to prepay like what do you consider the new risk model to be like with Hut 8 you have the single investment grade tenant or if you have a bunch of prepaid smaller pre private tenants where you can get half, near half of the deal up front. These are two interesting strategies. So you could almost say that the private tenants you have locked in some money and maybe you can lock in and reduce some downside risk but they are dependent upon the AI funding cycle which. What do you think that looks like?
- [28:00]
B
Yeah, it's interesting. I think that if iron had its choice it would rather have the investment grade tenant to get the credit backing so that it could go out and finance something cheap. Cheaply. But if you don't have that option or you have that option elsewhere but you need to fill in the gaps in some of your other data centers. To your point Charlie, if you can get them to help you finance the GPUs with prepayments, you own those GPUs so that company could go belly up, but you could still find another client to use that rack space, right? Yeah. So definitely a different, I mean I would say, you know, riskier, but in terms of neutralizing the risk where you can. It makes sense. But overall, two pretty interesting deals in terms of where the puck is going. This is really the first time, I think Charlie, that we've seen. This is the first time we've seen a hyperscaler or excuse me, a Neo cloud or powered shell within our universe of coverage announce multiple deals lumped together without really giving too much information outside of the dollar amount for the entire consideration. And also the kind of fragmenting of the revenue across all these different companies with different verticals within AI is at least somewhat interesting. Obviously a lot of these are startups that carry much greater risk than having a Microsoft or Nvidia per their prior contracts, but kind of spreading out the chips a little bit. Right. And it shows that there's more than just the Neo clouds and the hyperscalers with regards to demand for capacity.
- [29:38]
A
Yep, really interesting deals to kick off the week. Fun fodder to chat about. If you are an investor and you sold off because you're afraid COMPUTE was way too overpriced, you're probably wrong. We're going to come roaring back and more more deals will be announced until morale and market improves. We have another story we're going to talk about Trump Kimmy and the back half of the Kimmy Deepseek Compute repricing story here after a word from our sponsor. Lygos.
- [30:18]
B
Bitcoin is back above 65,000. But don't get complacent. Lending desks that are playing funny games with your Bitcoin could still lead you to lose your entire position. They are built like Celsius or BlockFi and they're rehypothecating or their balance sheet is a mess. You're leaving risk out there that you can't afford to take with your precious coins. Do yourself a favor, check out Lygos Finance. They are our preferred Bitcoin back lender here at Block Space they use Bitcoin native smart contracts to make sure you are always in control of your Bitcoin and no one else. You heard that right with Lygos Finance. When you're using them for Bitcoin backed lending you always hold the keys to your kingdom. Nobody else. With Lygos there's no bridging, no rehypothecation, no rapping and no funny business. Just competitive Rates as low as 8.5 to 10% APR. Go to Lygos Finance to learn more.
- [31:17]
A
All right, so pulling up the tweet here. This is it. This is the tweet Kimmy by moonshot tweets yesterday morning, Sunday morning quote, Kimmy, K3 has received far more love than we expected and our GPUs are feeling it. Demand has pushed close to the limits of our current capacity to protect the experience of existing subscribers, we're temporarily pausing new subscriptions and prioritizing COMPUTE for current members. And they added a little bit of a more, you know, a little change to their membership plans. But basically Kimmy said there's too much demand, it's sold out. So they've sold out of inference, basically. And this flies in the face of what everyone worried about, which is, oh no, compute's way too cheap. Everybody who's providing COMPUTE is vastly overvalued. The neoclouds, the powershells, the entire cycle is over. The capex cycle, it's over. This basically says the demand is insatiable and the game is now probably, how do you execute upon that insatiable demand? So I have this other tweet from Wayne at Orn Exchange who reminds us that the last time we saw this, we called it the Deep Seek moment, when Deep Seek came out with their near frontier AI and caused the entire AI data center sector to pull back. Wayne says, I can't believe it's been 1.5 years since the original deep seq moment. K3, Kimmy, K3 is the live test of whether Jayvon's paradox ideas hold for intelligence. Any friction on A on AI adoption in theory roughly went to zero. He says that on their compute index, H100 rental prices are up 52% and B2 hundreds are up 66% over the past three months. Even though the models are getting cheaper, Wayne says, I think better models are still getting framed as relief for the compute market. But every efficiency gain so far has come back as demand and the demand doesn't stop. You know this. If, if you thought that Jayvon's paradox was not applying to this, then you were probably sorely mistaken. And for those of you who've been under a rock, Avon's paradox basically says as something becomes cheaper to produce, it does not result in net less demand. In fact, the opposite, it tends to result in net more demand. We see this with energy specifically, and we are seeing this with compute. So if we see that tokens get cheaper, COMPUTE gets cheaper, and AI inference gets cheaper to Serve to customers. The customers don't just flatten demand. They actually want more of it because it's cheaper and even more useful and more reasonable for them to use that. I'll toss it to you, Colin, before getting into the geopolitical side of this.
- [34:29]
B
Yeah, I think it's just another validation that everything's moving so quickly. Fact of the matter is, very few people can confidently say what is actually happening within the market for compute. And you're going to see more of this volatility with whatever's left of this massive bull run. Right. People are scrambling to figure out what valuations make sense. They're also scrambling to figure out what business models actually make sense for this whole thing. And especially when it's coming from China. It's really easy to get scared on the headline and maybe easier still to get scared once you've dug into it a little bit, but it's never what it seems. We know this from bitcoin mining in China. Right. Even after the ban, roughly 15 to 20% of the hash rate for the entire network is still active in China. There's not just a cultural and language barrier, there's the geographic barrier and the great firewall. So it's really hard to say what exactly is happening with some of these companies and find your gurus that have the right calls online, I guess. But a lot of people have gotten this one wrong and will continue to get this one wrong.
- [35:44]
A
Yes. And if you want more context, I think we do live in a bubble here because according to a report from Goldman Sachs. I wish I had the chart here. Only 2% of households in the United States are paying for AI, and that's just paid customers. It doesn't necessarily even describe how the usage of those customers. So if you compare this to adoption trends of the Internet or any kind of technology, two, this means we're just very early on the adoption curve. Now, is AI something that people pay materially for? Is it gonna, you know, is it a considerable part of the average person's budget and expense to be determined. But you know, if you want a baseline, everybody pays for Internet service, so that might be a good comparison. So this gets into part two, because Kimmy came out and it was hitting all of these milestones. I have some benchmarks here for you. Right here. There's a bunch of different benchmarks. I won't go into each one, but if you want like just a ballpark one, Kimmy is right up there with GPT and Fable, the frontier models, and in some cases surpasses them just in terms of all round like speed, power and capabilities. Then as far as pricing is priced roughly the same as cloudsonnet, which is a bit of a mid tier pricing, but mid tier pricing for frontier intelligence. So in light of that, and in light of the US Government having increased attention to and often banning or restriction of frontier intelligence to Americans, we have this story from Axios. Headline reads the Secret Trump Administration Battle to Fight Chinese AI now there's I'll get ahead of it. There's not an announcement from the Trump administration, it's just inferring a trend that the Trump administration has talked about. The article says the Trump administration is showing signs it could ban cutting edge Chinese AI models, a momentous move that could lock in dominance by OpenAI and anthropic. As we know, parts of the administration have tried to implement de facto bans on foreign open source models before and US companies are increasingly using open source models from China because they're cheaper. Behind the scenes, the Commerce Department considered adding some Chinese AI labs to its entity list, which would cut off US access without a license, says a source close to the administration to Axios. Again, key voices from the White House, such as White House advisor Sriram Krishnan, have left the premises and national security hawks have grown louder. So this would kind of be a great irony as we framed last week, where the closed source, the closed weight, the private frontier labs which are US based, would be enshrined as the ones we could use, whereas the comparatively open models of China would be restricted. So in kind of an inverse, you know, United States has spent the past 300, 250 years branding ourselves as the ones of like open free information. And yet here we have a hilarious inversion of narrative. This led David Sachs, the co chair and President's Council Advisors, the aizar, to tweet this. I'm not sure whether Dean Ball is confessing to a regular regulatory capture strategy or simply predicting predicting this will happen. The weaponization of regulatory uncertainty as a competitive tool should be completely unacceptable, the subtweet person says. He argues there's no need to ban Chinese open source models, just direct agencies to issue soft law warnings that create enough FUD so regulated enterprises. Back off David. Tax says wrong. Regulatory decisions should always be well justified and grounded in facts, logic and evidence, not the deliberate exploitation of fear and uncertainty. Implementing a surreptitious policy through manufactured doubt rather than strong and explicit justification corrodes the rule of law and invites future abuse against anyone. So you have here the AIs are coming out and being really against banning anything. So he's right.
- [40:57]
B
I think he's, I think he's right in the sense that if you look at the export restrictions that they're or import restrictions for the models that they're debating, I'm not going to simp for Pooh Bear Ping and communist China, but this is what the Chinese do with American intellectual property. This is how they protect their domestic production and we typically revile that. Right now the one thing I will say is like I can see a kind of real politic there in terms of the. Makes somewhat sense for the survival of the nation state and the national economy if you really want to be a hawk on these things and restrict tooling and technology that otherwise is better than what you have. But the fact of the matter is if you read deeper into David Sacks comments there, I think he's basically saying, look, if OpenAI and Anthropic want to be competitive, then let them compete. Let them do the things that they need to move towards doing the things that Kimmy is doing currently. I mean I saw, I think it was Kali or something, a developer tweet, Bitcoin developer that he couldn't get open a. He couldn't get ChatGPT or Claude to debug some of his code. I don't really understand why. But then he went to Kimmy and it did it like that. So if you really want to protect your products and protect your positioning as the frontier models, you need to create a better model. And if we just ban the Chinese models, then the ones that get enshrined under the new regulatory apparatus in the US they're not going to have any impetus to compete except against each other. And the question there of degrees is really worth asking because I mean, how much difference is there in terms of potential pitfalls between OpenAI and Anthropic with regards to what they're creating?
- [42:45]
A
Right.
- [42:46]
B
I mean I know there's the whole like Anthropic's more woke or OpenAI is woke and GROK isn't whatever, all those, you know, the culture war aspects of it. But there are other legitimate questions too with regards to efficacy and with regards to the tooling that they're providing.
- [43:02]
A
So yeah, I view this. I split this into kind of two big camps here. One which is that of the U.S. government trying pretty ham fistedly and effectively to protect Americans from the implications and of this intelligence which we've seen is almost impossible to do. I respect the attempt like it is very scary, but we're going to get this intelligence from somewhere or another. And the whole banning of it is, has been very ineffective. See information theory, see historical precedent with fable still being able to be exploited. And others see the challenge of like restricting users from using Chinese or other open models. These things are coming to your computers whether you like them or not and eventually there'll be something you can run locally. That said, there is actually probably a good argument for maybe some type of economic protectionism here because if we understand correctly, Kimi and other Chinese models have basically let OpenAI and Anthropic and other frontier models in the US front the money and the effort and R and D to produce the intelligence and then they kind of through distillation attacks and through basically siphon off and like a parasite pull some of this intelligence for themselves. There is a pretty good strong case the US needs to figure out how to help American companies like protect against that happening. I agree with that.
- [44:40]
B
That's a good counterpoint.
- [44:41]
A
Yeah, but I don't see anybody saying that. I don't see anybody making that argument. That's not the argument from the US Government because we try to insofar as we can like enforce copyright law. We try to, you know, say counterfeit goods aren't allowed in the United States especially. And so we use the, the forces at our disposal to, you know, clamp
- [45:05]
B
down on fake Gucci bags.
- [45:08]
A
Exactly. You're not allowed to do it because Gucci is a brand that somebody owns and you can't sell legal stuff, copyright, yada yada yada. Maybe there's something to be said here in that vein for Chinese models. The problem is how do you prove this? How do you stop it? And I don't know. I will say I don't see anybody making an argument. It's all like a we got to protect, we got to protect people. And as much as I get that, it's terrifying to think about what this level of intelligence in the hands of average people will do. I don't think the solution is this a ham fisted ban because we're running at a breakneck speed in this direction. I don't see anybody being able to stop it. So with that in mind, I think we could see this is, this story is clearly not over. We'll have to see what Kimmy does. Behind Kimmy, I assume there are more of these labs like this Deep seq, Pokemon probably. Grok 5 is imminent. And then we have the next iteration of the frontier models Fable and GPT anticipated for next month. These are Rumors right now. But everybody who knows, you know, probably going to get a Fable 5.1 or other or a GPT6 very soon.
- [46:34]
B
So it's pretty thrilling. It's like the nuclear arms race, except private companies are building the nukes.
- [46:39]
A
Yeah.
- [46:39]
B
That's how it feels. And I think that's part of the reason why trying to approach this from a regulatory front or even just walk the line in terms of figuring out what to do about this technological advancement is going to be really, really true. It's going to require adeptness and precision that quite frankly, I don't think any of the recent administrations would be up to the task for, you know, like the Biden administration. Definitely not the Trump administration. I don't think so. Probably not the Obama administration, let alone Bush. So, you know, I expect ham fisted responses all the way down.
- [47:18]
A
Yeah.
- [47:19]
B
Is kind of what I'm bracing for.
- [47:22]
A
Yeah. Not a lot of, not a lot of confidence in the high level strategic decision making of the executive branch right now, regardless of which side of the aisle you sit on. I will point back, maybe we'll bring it up again. But Alex Karp of Palantir had a whole soapbox about this soapbox ramp on a rant on like CNBC a couple weeks ago. I think it was kind of a harbinger of what, what is was coming. Talking about protectionism and like open access and these, these cloak and dagger games of private companies kind of vying for both moral and economic superiority. Yeah.
- [48:03]
B
And at the heart of it, he was talking about why would you as a com. Why would you as a business use a model that you knew was training other or newer versions of that model on your proprietary business information when that will end up being de facto in the software stack for your competitors? You know. And it really is an interesting problem. This gets to the heart of one of the earliest worries about AI from the creative standpoint is like copyright law is not ready for this at all. At what point when information becomes segmented and then reanimated again? It's kind of like the ship of Theseus was. Is it the original information if it's, you know, pieced apart and then put back together is like this Frankenstein's output of knowledge that is that that goes into these models? I don't know.
- [48:58]
A
Yeah, I do think people will be talking about information theory a lot. Um, it's just very difficult to, to restrict the flow of information. You can just kind of create a more when we're talking about things which are almost, which are almost like purely algorithms. It's very difficult to restrict that. So more on that as it develops. Thank you for listening to Blackspace Live, a show every single weekday at 1pm Eastern featuring quick hits on AI data centers, March markets and emerging technology. Get all things block space, news, written content, rss, more content on our website at Blockspace Media. This show is brought to you by CleanSpark. NASDAQ listed ticker clsk I'm Charlie. I'm Colin and we will see you tomorrow.
- [49:58]
B
Sam.