SpaceX's $2T Case, Nvidia's Shock Selloff…

SpaceX's $2T Case, Nvidia's Shock Selloff, America Turns on AI, Trump Pulls AI Order, Bond Crisis?

1. Key Themes

The AI Revenue Inflection Point Has Arrived

The conversation repeatedly returns to a central thesis: AI is no longer theoretical—it's generating real, profitable revenue at unprecedented scale. Gavin Baker quantified it directly:

"If OpenAI and Anthropic are at, call it $100 billion of ARR now, with 80%-ish gross margins on inference, like the returns are there. And then if we add in Gemini, we add in Cursor, we add in XAI, we add in open source, you know, it's not hard to see $200, $300, $400 billion of ARR at the end of this year at high margin." 00:04:56

Anthropic's profitability was flagged as a narrative-changing moment: "The fact that they are now, they were EBIT positive per the Wall Street Journal in the most recent quarter is a really important fact for kind of the whole AI narrative." 00:04:34

SpaceX as the World's Most Important Infrastructure Company

SpaceX's S1 revealed a business transformation that most investors haven't priced in. The emergence of "Elon Web Services" (EWS) as a compute infrastructure provider—with Anthropic paying $1.25B/month for Colossus—signals SpaceX is becoming the backbone of AI compute. Gavin Baker explained why this is structurally different:

"Their first data center was 122 days. The second one, it took them 91 days. The third one was, I think, 66 days. They build data centers dramatically faster than anyone else at a lower cost. And now that you have a clear offtake partner, and I would expect partner to become partners, there is no reason they can't start stamping these data centers out really fast." 00:48:02

Chamath put a simple underwriting framework to it: "Terrestrial data centers alone are $100 or $200 billion of revenue by 2030, 2032. And that means just building it. So already you're buying it at 20 times revenue just for that business. Everything else is great." 01:03:47

The Anti-AI Sentiment is Not Organic—And Has a Dangerous Source

Multiple speakers argued that the anti-AI backlash in America is at least partially manufactured or amplified by foreign state actors. Gavin Baker was direct:

"It is starting to feel or seem like there may be a CCP funded campaign against AI and data centers in America. And that's very logical for China, but it is not good for America." 00:12:52

Friedberg extended this historically: "I don't think it's just China with NGOs today. I think that there is a long history of state actors intervening in media activities in foreign nations to try and create the sentiment and fuel a sentiment that reduces progress in that competitive state. I think this goes all the way back to KGB design during the Cold War." 00:21:18

2. Contrarian Perspectives

Recursive Self-Improvement May Make Moore's Law Look Conservative

Most people think AI progress is impressive but linear. Chamath and Gavin Baker suggest that once recursive self-improvement is unlocked—the focus of Karpathy's new role at Anthropic—the trajectory could become truly parabolic.

"This idea of recursive self-learning puts these models on a combination of overdrive and autopilot. And if you put those two things together, I think that you start to, you could potentially live out this idea that there's an order of magnitude improvement on a yearly basis." 00:03:48 — Chamath

Gavin Baker backed this up: "Chamath's statistics of 10X-ing every year might seem conservative if that comes to pass. And then, of course, continual learning is the holy grail, where the model learns from experiences the way humans do. And that's something we haven't unlocked yet. And those two combined, I think, they might pull the future forward in a very real way." 00:06:39

The AI PR Crisis Is Caused By CEOs Who Are Good at One Thing and Bad at Everything Else

While mainstream narrative blames anti-AI sentiment on fear of job loss, Chamath argued the real problem is that tech CEOs are catastrophically bad communicators who are actively creating the backlash with tone-deaf messaging.

"I thought the Matthew Prince note was horrible... This was like from the PR school of retards... You could not have written a worse memo. It's like you reduce humans to a label called the measurer. And then you're like, I'm going to lay off all the measurers." 00:40:23 — Chamath

His prescription: "Shut the f*** up. Get behind the keyboard. Just do your job. And if you need to manage something, just manage it. But don't write these missives. You're terrible at it. All of you." 00:41:50

GPU Useful Life Bears Are Wrong—And This Saved an Entire Industry Segment

Michael Burry and others argued GPU amortization schedules were too long (4-6 years vs. a "true" 2-year lifespan), implying neoclouds like CoreWeave were massively overstating profits. Gavin Baker argued the opposite, and that NVIDIA's quarter proved it:

"Now that we've disaggregated in front of us, we have these different domain-specific accelerators. You can mix and match them... you can put, whether it's a Grok accelerator, whether it's a Cerebris accelerator, in front of old GPUs... And those older GPUs, they have a useful life for 10 or 15 years. And this means that you can finance GPUs." 01:20:47

Chamath added: "That quarter single-handedly saved the neoclouds this quarter. I mean, they single-handedly saved them all." 01:21:31

The Strait of Hormuz Closure Is Relatively Good for America

While universally framed as a global crisis, Gavin Baker made the counterintuitive case that the closure is net positive for U.S. competitive positioning:

"Every day the Strait of Hormuz is closed, I think is relatively good for the re-industrialization of America... Electricity is a base input to every manufacturing or industrial process, essentially all of them. And what we make electricity with in America, overwhelmingly, is natural gas. And NG1 is down this year. The input cost for electricity in the rest of the world, LNG is a very important one, and it's up 100%, 200%." 01:30:38

3. Companies Identified

Anthropic AI lab, currently EBIT positive and growing faster than any company in history at scale. Hired Andrej Karpathy to lead pre-training focused on recursive self-improvement.

"They were EBIT positive per the Wall Street Journal in the most recent quarter... they're now the size of... they're bigger than 100 different countries for sure. And come out really fast and now profitable." 01:28:23 — Gavin Baker

SpaceX Aerospace and now AI compute infrastructure company. S1 revealed $11.4B Starlink revenue (50% growth), $4.4B operating income, and a new $15B/year EWS compute business anchored by Anthropic.

"The fact that they are now—they were EBIT positive per the Wall Street Journal—$15 billion from Anthropic is extraordinary." 00:49:22 — Gavin Baker

Cursor AI coding tool with more proprietary coding tokens than exist on the public internet. Composer 2.5 just became Pareto dominant after only 3-4 weeks of reinforcement learning on Colossus 2.

"Cursor allegedly has more tokens of coding data than exist on the public internet. And that is a stat from, I think, more than a year ago. So I'd imagine it's grown significantly." 00:50:01 — Gavin Baker

CoreWeave AI-focused cloud/neocloud, notable for GPU financing at ~6% rates and long-term contract structures proving GPU useful life theses.

"CoreWeave's lowest financing—I can't forget if it's 6% or 7%. 6%. It's going to come down. And if you can get an asset-backed loan, an asset-backed financing for GPUs at a lower rate than other chips, that's a profound advantage." 01:21:16 — Gavin Baker

NVIDIA Semiconductor company. $81.6B in Q1 revenue, 85% YoY growth, 75% gross margins. Announced $20B CPU business and $80B additional buybacks.

"The one thing I would just say... they said if they thought their CPU business was going to be $20 billion this year... That's extraordinary. It means overnight [they became] one of the world's largest CPU manufacturers." 01:18:31 — Gavin Baker

Flock Safety AI-powered crime prevention platform using cameras and license plate recognition. Being deployed bottom-up by municipalities.

"Crime is now a choice... A16Z had a great essay on flock. We can really, really solve crime. And it's just a choice." 01:34:30 — Gavin Baker

Exite Labs Semiconductor company, Atreides portfolio company, chips going into essentially every Starlink satellite despite not being radiation-hardened—a testament to SpaceX's engineering.

"My firm, Atreides, is an investor in a company called Exite Labs. It's a matter of public record. It's going to be in essentially every Starlink. And the chips were not designed to go to space. They're not radiation hardened... they just happened to pass [rad testing]." 01:10:19 — Gavin Baker

Abacus AI AI company building verticalized, small language model solutions for corporations.

"These very small models, small language models, and then verticalized ones are the future. We've got a company, the abacus, that's doing it for corporations, crushing it." 00:09:32 — Jason Calacanis

4. People Identified

Andrej Karpathy AI researcher, coined "vibe coding," previously led FSD at Tesla and co-founded OpenAI, now leading pre-training at Anthropic focused on recursive self-improvement.

"He's been at the wave upon wave of AI. He was probably the first person that really commercialized the Richard Sutton bitter lesson essay when he was leading FSD at Tesla... And what he's done as a kind of a free agent is also quite impressive." 00:02:51 — Chamath

Gavin Baker Managing Partner at Atreides Management. Deep expertise across semiconductors, AI infrastructure, and space tech. Investor in Exite Labs, among others.

"I manage more than 100 positions at my firm. And I do that with a team. We're over 30 people now." 01:27:54

Leopold Aschenbrenner AI investor and researcher, reportedly a Rhodes Scholar at 19, running a fund that went from zero to $5B. Known for work at OpenAI and thesis on AI scaling.

"He's clearly a brilliant man. I think he's a Rhodes Scholar at like 19. And I think my understanding is he's putting up pretty extraordinary numbers." 01:13:30 — Gavin Baker

Shyam Sankar CTO of Palantir. Advocate for listening to frontline AI users rather than AI lab CEOs.

"Stop breathlessly asking these model makers what they think. Go to the end user and ask the person in the factory that's using the model and ask him what he or she thinks, ask what the doctor thinks, asks what the scientists think, and start to tell those stories." 00:41:00 — Shyam Sankar (via clip)

Felicia Horowitz Wife of Ben Horowitz; credited with transforming the Las Vegas Police Department's use of AI and drone technology for crime prevention.

"She has done this incredible job with the Las Vegas Police Department. It is one of the most impressive things I've ever seen... If you gave the Las Vegas Police Department 30, $40 million a year, it would be the safest city in America." 00:36:08 — Chamath

5. Operating Insights

Concentrate Ruthlessly—Five or Fewer Public Positions

Chamath gave a clear framework for how he manages public market investing that has direct applicability for any operator-investor:

"I have a few companies that I really believe in. I have extremely concentrated large holdings... How many public stocks can you keep in your brain and still sleep at night holding for the long-term? It's five or less." 01:26:56 — Chamath

The implication: tracking 30 positions means tracking none of them deeply. Concentration forces genuine conviction and reduces noise-driven decision-making.

AI Market Is Cross-Sectionally Inefficient—Use It

Gavin Baker identified a specific pricing anomaly that can be exploited. Two contradictory things are being priced simultaneously:

"Cross-sectionally, if you look at the valuations for all these AI names, they just, they can't all be accurate. You have memory makers that are three to five times PE. You have NVIDIA at a really low PE... The AI market is cross-sectionally inefficient right now." 01:14:22 — Gavin Baker

For an investor, this is a specific actionable framework: figure out which sector's valuation is correct, and position accordingly—the power/cooling/optical names vs. NVIDIA and memory cannot both be right simultaneously.

Tell Customer Stories, Not Founder Stories

Shyam Sankar's clip and the broader discussion produced a specific communications/marketing insight for anyone building or investing in AI companies. Stop letting lab CEOs and product founders speak for the technology:

"Stop breathlessly asking these model makers what they think. Go to the end user and ask the person in the factory that's using the model and ask him what he or she thinks." 00:41:00 — Shyam Sankar

For operators: the ROI of switching your marketing/PR from founder-led narratives to customer-outcome stories is exceptionally high right now, especially given the backlash environment.

6. Overlooked Insights

DC-to-DC Power Architecture Is the Next Great Data Center Moat

This was mentioned briefly in passing but is potentially one of the most significant infrastructure insights in the episode. Chamath noted that Jensen Huang needs a design partner to eliminate power conversion losses in data centers—and that Elon is the most likely candidate:

"There's a great push that Jensen's making, which he needs a partner. And I think Elon becomes a natural partner to do DC to DC. Forget all this DC to AC to DC nonsense that goes inside of a data center. All the lossiness... If we could just do DC to DC, like it comes in as DC, direct current, it goes right to the rack as DC. But it requires a fundamental re-architecture." 01:04:17 — Chamath

This is barely discussed in the AI infrastructure world but could have enormous implications: whoever solves DC-to-DC architecture in data centers will meaningfully cut power costs, improve compute density, and reduce cooling requirements. This is likely to define the next generation of hyperscale data center design and represents an early-stage investment theme that has not yet attracted mainstream attention.

AI Seasonality May Disappear—Creating a Year-Round Revenue Step Change

Gavin Baker briefly floated this observation almost as an afterthought, but the implications are significant for forecasting AI revenue growth:

"AI has been seasonal. In the past, that's been because college students, they use a lot less ChatGPT and Claude in the summer. And generally people maybe work a little less hard when the weather's nice. Now with the genetic AI, will the fundamentals still be seasonal? We will see." 01:32:07 — Gavin Baker

If agentic AI eliminates the seasonal trough (because agents run 24/7 without summer breaks), then all current revenue models for AI companies are structurally understating annualized run rates. This is an underappreciated upside case for AI ARR estimates that could catch analysts flat-footed in Q3 2025 reporting.