SpaceXAI Shockwaves Continue After Staggering Deal

Hinzugefügt: 2026-05-10

[00:00:00]Here's the latest all-in podcast recently dropped. Elon Musk posting this and there is some discussion opening the pod about the SpaceX anthropic deal and some of the broad implications.

[00:00:09]>> It's interesting way to put it. Uh don't disagree. All right, first story. Elon just leased all of Colossus 1, his data center.

[00:00:15]>> He did lost what? Yes. Shocking to Daario and Anthropic Chamath on last week's pod. Uh go ahead and give yourself a pat on the back. You said Elon and Dario should do a deal tomorrow. It didn't happen the next day.

[00:00:29]It happened 5 days later, so you came close, Jama. Uh, but no cigar. Because of Anthropic's obvious compute constraints, Anthropic just added over 220,000 Nvidia GPUs, over 300 megawatts of energy. The deal is already having an impact. As we've discussed here, Claude users have been experiencing rate limits. Well, Claude has now doubled the Claude code rate limits, removed peak usage caps for paid users, and increased API volumes for Opus models. XAI is now training their models at Colossus 2. So they have more than enough compute. Elon made a great bet on compute and built up those data centers really fast and that is now paying off. We had the cursor deal we talked about last week. Let's talk about the emergence of Elon Web Services EWS Chim. He is now in the hyperscaler competing against Google Cloud, Amazon Web Services and Azure.

[00:01:22]and uh I don't know if you had inside information or just a brilliant uh epiphany but uh take us behind the call and what do you think about the deal itself?

[00:01:30]>> I think the deal is fantastic. I'll say maybe three quick things. The first is as I mentioned a couple weeks ago anthropic and OpenAI's revenue performance has nothing to do with demand. Zero. It is entirely to do with the supply constraints that exist in data centers and specifically in power.

[00:01:53]>> So credit to Jamath first of all for suggesting this deal should happen only well less than a week before it did.

[00:01:58]Some big brain [ __ ] there or maybe he knew something. And Jamas also right about the first point around constraints. It's not just power although that's a major one. A huge bottleneck. It's also becoming increasingly challenging to physically build the AI computer on Earth. delays with approvals, all sorts of [ __ ] It's not just power. Long-term, obviously, AI data centers in space are where it's at. And as I record this, there's only one company on the planet, SpaceX, with the launch cost and cadence for this to be viable. There are very big implications within. If you're an AI company, option one is to use SpaceX to launch your own AI data centers into space. Thousands, tens of thousands, hundreds of thousands, ultimately a million plus. And then you can have your own version of what I continue to call Starink. A few challenges though. Do you have expertise in manufacturing satellites at scale? Are you aware that SpaceX is not a charity? They will actually charge you a reasonable margin on those launches as they should.

[00:02:54]They're a business, not a charity. Do you have the capital? Are you willing to invest? If not, there is an alternative, which would be get the [ __ ] in line for access to Starink. By the way, can't wait till we find out a better name.

[00:03:06]Make a deal. provide some capital to help accelerate things. This appears to be exactly what Anthropic has recently done. They've seen the writing on the wall, the massive constraints, a huge bottleneck. What's the solution? Data center in space. Are we going to do it ourselves? [ __ ] no. Let's just partner with, collaborate with, get in bed with SpaceX. We'll pay what's necessary for access. We'll contribute capital. Most likely, we won't have to do this ourselves. and we will be prioritized over other AI companies that didn't have their [ __ ] together, didn't see the writing on the wall, and weren't willing to make the same deals. To be clear, when the deal was announced yesterday, Space XAI will provide Anthropic with access to Colossus One, blah blah blah.

[00:03:43]There was a full announcement, but there was more, and it matters. The XAI account specifically called out from the announcement this little tidbit. SpaceX AI and Enthropic have also expressed interest in partnering to develop multiple gigawatts of orbital AI compute capacity. Phase one, a terrestrial partnership, access to Colossus 1. The next step, orbital AI compute. If they had infinite power, I think that their revenues would probably be even more parabolic. And so all the breathlessness about either exceeding or underperforming a forecast, in my opinion, mean nothing. I think the 5-year view for those two companies is quite robust. The thing that they really need is more compute and more power.

[00:04:31]That's the first thing. The second thing is while they need that, we have a very big problem which is we unfortunately have very poor leadership at the head of most of these AI firms. I think they are coming off as untrustworthy or too self-interested. The political reaction now is starting to turn negative. The community reaction is negative. You have about 9 gawatts that are supposed to come online this year. Almost 50% of it now is being protested. More than likely, if if history holds, most of that will get turned off. So, they will get even more supply constraint. So, that's the setup. So, what's the opportunity I think for Elon? If you look inside of how people try to nitpick the SpaceX valuation case or let's not even let's give them sorry let's be more generous when people try to paint the bear case or they tried to red team the valuation the biggest element is the on the comeum value around the orbital data centers and by actually landing a bunch of terrestrial capacity I think you start to blunt that because you can now start to say that even if the orbital data centers get delayed by a few months or a few quarters even if the techn technological d-risking of it takes longer. He now has this structural core business that will effectively subsidize his ability to train Grock which I think is a really important and under reportported theme. So you have all this infrastructure he somehow saw the tea leaves before most people. He built to a level of scale and secured power before most people. It is >> this is all true and just for anyone kind of wait what the hell is he talking about? Jamath is essentially referring to the Colossus 1 deal. Colossus 1, a gigantic AI training computer. This is what Anthropic has just agreed to have exclusive access to for training to obliterate, at least in part, their current constraint on compute. This colossus is the very same AI training cluster that Nvidia's CEO Jensen Hang described as having gone up in record time. And not just that, but suggesting that no other company could have got this done anywhere near as quick.

[00:06:34]Without Musk's leadership, there's no chance. Now, this is something he did not need to say voluntarily out loud, but he did tacidly implying that everyone else who's also buying chips from Edvidia, his customers are too [ __ ] inept, incompetent, or slow to move nearly as fast as XAI. Now, this is important. Imagine if it continues to hold true that XAI are able to get AI training compute massive scale up faster than anyone else. And there's a huge constraint on AI training compute at the moment. This alone puts SpaceX AI in an incredible position. Let me explain.

[00:07:10]Colossus 1 Anthropic now has exclusive access. Space XAI currently using Colossus 2 for their own needs. You know, it's only a matter of time before Colossus 3 and four and five. While this is happening, they're putting up AI training comput faster than anyone else on the planet. They'll leaprog to their next generation training compute, license access to the prior gen, receive an enormous income stream, which will in part help fund the cost of the orbital data centers. And the same thing's going to play out there. Build a gigantic swarm. I mean, they're literally targeting a million AI satellites and then AI companies can license access to that as well. It's very important to understand that if you're an AI company and you understand there's a huge constraint around AI training compute and you know you acknowledge bro we can't build these things as fast as SpaceX AI can. So let's just agree to license access to their prior jet as they're moving on to the next one. This is the best way for us to move forward.

[00:08:03]This is going to be a phenomenal business. the terrestrial based AI compute over the next few years and then orbital data centers >> now become the critical asset and now he's kind of kingmaking and I think that that's a really interesting valuation reinforcement as SpaceX goes through testing the waters and the and the road show.

[00:08:21]>> Brad, your take?

[00:08:21]>> Yeah. No, I think it's well said. I mean, first we we know that there's nobody better on planet Earth than Elon at converting electrons to tokens. It's a critically important evolution to the story. You know, I think our friend Shawn Magcguire, he he he sent out a tweet that summed it up well and he said, "SpaceX has this five layer cake, launch, connectivity, compute, hyperscaler, space data centers, and then applications and models and then other bets, right? The question on the road show has been, but X.AI doesn't isn't on the revenue trajectory of OpenAI and and Anthropic, and yet there are huge commitments." And now we see the ace card that Elon's playing. He said he was building AWS all along or EWS all along. And so I estimate that this is going to generate in this year an incremental 4 to5 billion of revenue on top of what I I've seen analysts estimates in the mid20s. That's a material amount of incremental revenue to offset the cost of the investments that he's made here. And that will subsidize to Chamas's point all that he's investing to build the next generation of Grock. Remember too that he has three facilities, Colossus, Macro hard, and macro harder. 1.2 gawatts in Macro hard and macro harder in Blackwell. So he's given the one that's kind of less connected. H100's great for inference to Anthropic. He's monetizing it in a big way. It's terrific for Anthropic and it solves what I think was the biggest question in the valuation story which is what if he spends ahead of X.AI's revenue. It takes the pressure Chimath off X.AI delivering immediate revenue. Now he becomes uh an immediate competitor in the hyperscaler. I don't think this is the last announcement. I think he's >> there's no way it's the last announcement, bro. We're going to see many more deals like this, presumably with many other companies. As I've mentioned, if you're an AI company at this point, there's no point in you attempting to build out your own AI infrastructure at a fraction the speed you know Space XI can do it and all the risk that comes along with that when you now have a company that you know is going to be scaling massive amounts of AI compute both on Earth and then in space, which you can just license access to and focus on what you do best, which is not moving atoms.

[00:10:39]>> Going to make a lot more, you know, moves in this direction. I think it will be a material part of their story and their revenue projections uh as they come together. And I would just say finally, you know, again, everybody has talked about how we don't have enough power, how we don't have enough compute, how the revenues would not show up this year, you know, but the chaos that is American capitalism somehow finds a way.

[00:11:02]Okay? And there's tremendous demand for anthropic and we find a way. I was so happy to see kind of the dant and the kind exchange between the team of Anthropic and Elon because we need all of this in order to produce uh American frontier models to stay at the frontier.

[00:11:21]And then finally, I just say, you know, Chimath, you referenced these activists that are protesting delaying these data centers in in these localities. One thing I want to dispel this myth. These this is not like organic hyper local protests by people in a community that aren't being spurred on. This is highly organized activists that are moving across the country to stir up trouble in the exact same way they did to stop all fision reactors being built 30 years ago in America. Now we have no nuclear reactors being built. China's got a hundred of them. Who was funding those activists? I think we need to really look into who's funding the activists.

[00:11:58]Now, I'm not saying that there aren't any concerns, but the misinformation about water, the misinformation about electricity bills, electricity bills are going up in the places that are not building data centers, New York and California, because they haven't built any supply on the grid. In Texas, where you're building the most data centers in the country, electricity costs are going down. So, um I'm I I think that's a boogeyman that we got to take on.

[00:12:23]Another of the many reasons why AI data centers in space make sense. I mean, what are the protesters, the haters, the activists going to do? Shake their fist at the sky? Yell up at the clouds? So, let's just sum this up. Anthropic announced yesterday a deal for exclusive access to Space XAI's Colossus 1 AI training computer. At the same time this was announced in the formal announcement there was also a disclosure that the two companies had already discussed the idea of working together on orbital data centers which to be clear will be an extremely capitalintensive project. The intent is to launch a million satellites to do AI training and inference. The good news is anthropic growing at a stupid rate only constrained by AI compute which SpaceX AI is alleviating.

[00:13:10]Anthropic is going to have so much cash to splash which is going to fuel further expansion for SpaceX AI. Here's what Sax had to say about Anthropic. Just think about this. Of all the customers you could have, is there a better than this?

[00:13:24]>> My take is look, let's just honestly and accurately assess where the state of this AI market is at right now and Anthropic's place within it. So for the last 3 years, Anthropic has been growing at a rate of 10x a year. I think going into this year probably the conventional wisdom was that there'd be no way to sustain that kind of rate of growth at this level of scale. And what happened in the first four months of the year?

[00:13:48]First we find out that from January 1st to March 31st they grew from roughly 10 billion of ARR to 30 billion. So it tripled and then in April if anything the rate of increase seemed to accelerate. They went from 30 to 44 billion of ARR. Nobody in Silicon Valley has ever seen anything like it. Forget about the rest of the country. I mean, all we do in Silicon Valley is deal with exponentials and still people have never seen that kind of growth at that level of scale.

[00:14:14]>> Pretty important point. I just want to expand a little bit to the earlier discussion around the constraints. The reason this deal was announced between SpaceX and Enthropic as these guys were mentioning. Anthropic was massively constrained by compute, not demand, not interest from users. Compute. They desperately needed to address this bottleneck in order to increase the number of people, well technically tokens, but let's not get too technical using Claude. They're still constrained.

[00:14:40]The whole industry is constrained. The desperation from AI companies to get their hands on more AI compute is off the charts. These companies find themselves in a situation anthropic, especially with its crazy rate of growth. Their revenue floodgates will open if only they can address the compute bottleneck. This creates an incredible feedback loop. If SpaceX AI can massively scale AI compute, they will immediately have customers willing to pay whatever is necessary to get access. Those customers by gaining access will then be able to offer services to more customers and or more tokens causing an instant surge in revenue which can then be redirected toward more AI training compute, more revenue. You understand, right? This is going to be a crazy crazy feedback loop.

[00:15:27]The only thing holding them back in the future was compute. Now they've made this deal. They've made other deals as well to get that compute. I think it's pretty much a foregone conclusion that they will hit that forecast of 10x this year exiting the year call it roughly 100 billion of ARR. And now the only question is whether they hit a trillion in 2027. And we can debate whether >> okay I got to stop here. The expectation here anthropic end of year 100 billion in revenue for the year an annualized rate we heard earlier they've 10xed this number for 3 years in a row 110 100 another 10x which is what Sax is describing here now if they 10x in 2027 from 100 they'd exit 27 at the truly absurd IR of $1 trillion by the way that's a lot of cash to splash on AI compute >> getting on board we can we can debate whether that's true or not but look if they do that. I think they'll easily be the most valuable tech company in history. In fact, they might even be more valuable than the rest of the Mag 7 put together. Just to give people some basis for comparison here, you know, the biggest tech companies, Apple, >> Nvidia, >> Nvidia, Google. I think they they kind of do around 4 to 500 billion a year right now of of revenue. I guess Nvidia is in a little bit of a different category, but you look at you look at >> hyperscalers. The three hyperscalers.

[00:16:44]Yeah. I mean, Google is doing what, like 120 billion a quarter, something like that. 100 billion a quarter.

[00:16:50]>> Correct.

[00:16:51]>> But growing at what 20% year-over-year, not 100% and certainly not a thousand%.

[00:16:56]So the fact that Anthropic could be on track, in fact, let me correct that.

[00:17:01]>> You see them going to it'll be a Maggie.

[00:17:04]>> I'm saying something else which is that unless something about their current trajectory changes, Anthropic will be the most powerful monopoly ever created in human history. Oh, >> again it will be, you know, a trillion dollars of AR growing at some exponential.

[00:17:21]>> Interesting.

[00:17:22]>> Dario calls it AGI. I call it the biggest monopoly in human history.

[00:17:25]>> Interesting to hear that word monopoly sachs. Very interesting placement.

[00:17:30]Chimath, go ahead and then just in 2025 was 420 billion. Microsoft was 300 billion. Alphabet was 390 billion.

[00:17:40]Amazon 700 billion. Nvidia 190 billion.

[00:17:43]Meta 185 billion. Tesla 110 billion total about 2.3 to 2.35 trillion. So if Sax is right and Anthropic, you know, can tack on a trillion, it won't be the mag 7, it'll be the mag one.

[00:17:58]>> Wait a second. Well, listen isn't this interesting. A post roughly a week ago from some guy with three first names who may have been thinking about AI when he posted, "You've heard of the Mag 7, but you are not ready for the big one. Want more content, early access, bunch of perks? Click the links in the pin comment. AG1 has been part of my daily health protocol for almost half a decade. It's packed full of vitamins and minerals and helps me fill in nutritional gaps while supporting energy, digestion, and immune function.

[00:18:28]It also has prebiotics and probiotics to promote gut health. I take my health very seriously, and so should you. Try AG1 today by visiting drinkag1.com/smr or click the link in the pin comment and enjoy a free welcome kit with vitamin D3 plus K2 and AG1 travel packs. That's drinkagg.com/smr.