SPCX, Anthropic, OpenAI: Three Companies to Lead Next Tech Frontier

also lower on the session a a quarter of a percent for RTX. I do want to shift our gears here, though, bring Sam backto n our attention to three companies that our next guest says arethe frontier of techno Joining us now, Ryan Air, the founder and director of diabetic. Ryan, thank you for with us today. Let's startith SpaceX. Because of the timeliness of its successful land finally reaching earth orbit, placing some satellites out into orbit. When you look at eX,s SpaceX right now? Is it a rocket company? Is it an AI company? Is it a compute company? I had a guest a few weeks ago tell me they expect SpaceX to become the big neo cloud in the world. So where do you stand on SpaceX? Yeah. So if you if you take a look at revenue, either the guided 100 billion IRR from the company or our calculated inr i, 56 billion in, in 2026 revenue that is split about one third in the AI compute. So outside of cursor and and grok. But the actual compute Starlink is one quarter. And then launch is a little bit less than than 20%. So what that tells us is that today the company is primarily an AI compute business.t's important to remember that the, the core resource that that SpaceX is actually selling is their engineering talent. They have the Seal team of engineering, and they can point that Seal team six at different problems, whether it's AI compute, rocket launch or or connectivity, the important implication to away from this is you know, less than a year ago, there was a lot of conversation about whether the AI acquisition was a bailout. And combined with the cursor one two pun it may go down as one the best acquisitions in the history of technology, right up there with Instagram, followed by WhatsApp. And obviously, we need a lot of compute with what the anthropic's in the open AI isar, I'm just wondering, I mean, it's been a fascinating 48 hours what we've heard from both of these companies. I mean, this alarming leaked prospectus from anthropic, whichyou know, generating a lot of debaten the market yesterday on the other side of the country, you had Sam Altman saying, guys, we've got to put the brakes on this stuff. We can't release models because we've got to get the safety right. How are y thinking about these two companiesht now and their endeavor to go public with all of this noise and this juxtapositionthe market of the dangers versus what to do about it? Yeah, I think the two companies are on a different trajectory in in this moment, obviously with with s one filed, anthropic is looking too public this year, you know, $2 trillion valuatio that underwritten against trailin losses of 42bile a very price insensitive or who's underwriting that that purchase. It looks like OpenAI will delay well into next year. I would guess that that is probably d to the fact that they have not had the favorable quarters that that aic has in these recent two quarters. It has surprised me that anthropic has not followed up the the leak of some of the trailing numbers of 2025, with the more favorable 2026 numbers. But the net net here is that it does appear that both are a going out the door as quickly as they possibly can. That may be due heg margin e, according to David Freiberg, as of 13 weeks ago, split between compute tokens between open source and frontier models was 80 for the frontier and 20 for the open source. That has shifted to 2080 over the last 13 weeks. So we're seeing a huge shift in in the market towards on prem open source. You know, we we see this on the ground and in our own company as, as we move on prem and open source and a lot of CEOs and peers that I speak with are saying what our spring Claude Bill was, is 5 t X what our what our fall Claude bill will be. That's really important to look at moving forward. So if we've seen that dramatic shift then toward this open away from frontier, and you say that that has us looking at anthropic rushing to market, perhaps because of some of this potential erosion, what does that mean for the economics then, and the compe mode of these companies bui frontier models who are expected toe these unicorn IPOs and arguably be massively successful companies once they, at least prior to the release of that data from the last 13 weeks. Yeah, for sure. You know, frontier mare what we often call the sport of kings is the largest prizee history of capitalism. Current gross margins are expected to be somewhere between 70 and 80% for for the frontier models. So there's lots of space to still give. And ultimately, there a series of edge use cases that will always lean towards frontier and payeally significant premium for that. What that means is tha again, the, the, the underlying investor is not a rate sensitiv investor. They're not a price sensitive investor.Whato is the of delivering super intelligence to the world in the f which again, is the largest prize in the history of capitalism. And y say, I mean, sort of getting back to space that there's a new class of company that's emerging within the mag seven. You would call the space conglomerate. Just walk us throughhat. Yeah. So we see in most partic with with SpaceX, where today the compute business and the connectivity business funds the space exploration and, you know, and the rest of the, the R&D build out that's evolving into the future, into orbital space, comput connected through Starlink and then serving super intelligence to the world on any spot on the planet. So that is the stack. SpaceX is the only one that owns the full stack. You know, of that new superintelli space corate. But we do see other companies who are sma getting excited about it. Google in particular just announced that they're they're launching on SpaceX. You know, we see the work from Amazon having terrestrial compute, having a space program you know, being able to puthe these together. You know, these companies clearly see where the future is going, that data centers are moving off planet. And, and then we'll be, we'll be connected via satellites to the world. And so withceX having the full stack, , they have the rockets, they have the connectivity and have the compute infrastructure. f a gigawatt of energy and several terabytes per second of connectivity into orbit, how quickly can orbital compute become commerciallyngful? Yeah. So in our models, for the sake of the recognition that that, you know, SpaceX pulls off the impossible slightl late, we don't have orbital compute being a meaningful contribution to to the margin until 2029. And terre compute will be the workhorse for the next several years as they build out towards, you know, 20GW of, of terrestrial computethe next 24 months. You know, Jensen, just this just today mentioned that each gigawatt of terrestrialpute could produce up to 50 billion pretty incredible. As we move into the the 20s, when you consider the Nimby issues, when you consider the, you know, the advantage of solar and high capacity factor solar in space, it's very obvious that we're moving off planet with with compute and, you know, and that's going to make for a really exc 2030s and the superintelligenct it serves. Ryan, really appreciate you joiningto take a closer look at these three companies and what their futures may look