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

Summary

Theses (episode spine)

• SpaceX filed S-1 targeting $75B raise at $1.75T valuation (largest IPO in history by more than 2x Saudi Aramco). Ticker SPCX; Nasdaq listing expected June 12. Anthropic paying $1.25B/month ($15B/year, $45B total) for Colossus 1 and 2.
• Chimath’s SpaceX underwriting thesis: buy at 20x revenue on Starlink alone (scalable to hundreds of millions of subs, GDP+10-15% grower) plus Colossus/AI compute optionality. Tesla merger is “inevitable.” “That thing will look very cheap in a few years.”
• Gavin Baker: SpaceX builds data centers dramatically faster than anyone — 122 → 91 → 66 days per center — and now has a clear offtake partner in Anthropic. Cursor Composer 2.5 reached pareto-dominant performance after just 3-4 weeks of reinforcement learning on Colossus 2 using Cursor’s proprietary coding data (allegedly more tokens than the entire public internet).
• Gavin Baker: Nvidia Q1 FY2027 — $81.6B revenue (+85% YoY, +20% QoQ), $58B net income, $48B FCF, 75% gross margins, $5.3T market cap. CPU business to reach $20B this year. Low-to-mid-teens forward PE on real earnings is cheap. The ASIC share-loss narrative is wrong: Nvidia is growing faster than Broadcom’s AI revenue even ex-China.
• Gavin Baker: Nvidia’s Q1 result “single-handedly saved” CoreWeave and neo-clouds by proving GPUs have 10-15 year useful lives with disaggregated inference. CoreWeave’s 6-year financing at 6% is validated; Michael Burr’s “2-year GPU lifespan” thesis is wrong.
• Gavin Baker: The AI market is cross-sectionally inefficient — power, cooling, and optical companies are discounting a reality inconsistent with what Nvidia and memory makers are discounting. They can’t both be correct.
• Andre Karpathy joined Anthropic to lead pre-training on recursive self-improvement. Gavin: this puts models “on a combination of overdrive and autopilot” — potentially a new Moore’s Law for model quality with 10x/year improvement becoming possible.
• America Turns on AI: commencement speeches booed; Cloudflare CEO Matthew Prince’s “measurers” layoff memo; Zuckerberg’s simultaneous layoff and employee-monitoring announcement. Chimath says this is partially a CCP-funded campaign.
• Trump pulled a planned AI executive order at the last minute — leaked version would have required federal oversight of frontier models. Gavin and Chimath favor self-regulation plus courts; agree only on KYC for bioweapons/terrorism as reasonable US-China common ground.
• Freeberg: Bond crisis signals — 10-year at 4.6%, Japan 30-year at 5.1% (all-time high), UK at GFC highs, Germany at 2011 highs. CPI forecast to hit 6% in Q2. Global debt/GDP at 310%. “The roller coaster is at the top.”
• Gavin Baker: Three things are simultaneously true — rates going up is bad; AI fundamentals (Anthropic profitable, growing faster than any company in history) are accelerating; US is self-sufficient in energy and food and benefits relatively from Hormuz closure. “America is the best house in a globally bad neighborhood.”
• Chimath: Stay concentrated in 5 or fewer high-conviction public stocks. “There are no 30 things I’m tracking.” Top two positions represent 20-35% of his net worth.
• Chip sales to China: Gavin Baker says selling deprecated Nvidia GPUs to China lowers the probability they build a rival Huawei ecosystem; economic cooperation is stabilizing. Chimath aligns.

$SPCX (SpaceX — upcoming IPO)

Convergence / divergence: Full convergence on bull case direction. Chimath’s analysis is financial (Starlink compounder + Colossus optionality). Gavin’s is operational (data center build speed, Composer 2.5 proof of Colossus value). Freeberg is the idealist (space-based internet governments cannot control). All three see the Anthropic $45B contract as transformative validation.

Speaker calls:

• Chamath Palihapitiya (bullish, long-term): “I’m buying probably the most important internet infrastructure project since the internet itself. Starlink is going to scale to hundreds of millions of users. Getting a gigawatt nameplate working is freaking hard, man. They are the closest. And then the X factor — there’s only one guy who steps on stage where you’re always curious, ‘okay, what has he got up his sleeve?’ and that’s Elon. So if you had to pick an underwriting case, I would flex the revenue and realize terrestrial data centers alone are $100-200B of revenue by 2030.”
• Gavin Baker (bullish, long-term): “15 billion from Anthropic is extraordinary. I think this business can grow dramatically faster than anyone could have contemplated three months ago. And I think there’s no reason they can’t start stamping these data centers out really fast. GPUs will be allocated to who can plug them in, turn them on, and start converting electrons into tokens.” On orbital compute: predicts “second half of ‘28 to first half of 2030.”
• David Friedberg (bullish on concept, long-term): “I think having a space-based communication network, space-based data centers, space-based communication back down to Earth wireless — it’s generally a good thing. It’s good to have a backup. I like that.”

Cross-check:

• Price: IPO at $1.75T valuation; Nasdaq listing targeted June 12 (SPCX). Not yet trading. Polymarket: 71% chance it closes day one above $2T.
• Recent headlines: S-1 confirmed $1.25B/month Anthropic deal through May 2029. Q1 2026 net loss of $4.28B; accumulated deficit $41.3B. Musk holds 85.1% of voting power (dual-class). Data center build time confirmed: 122 → 91 → 66 days. Cursor Composer 2.5 pareto-dominant after 3-4 weeks on Colossus 2.
• Inconsistencies: S-1 shows a net loss of $4.28B in Q1 2026 alone and accumulated deficit of $41.3B — the podcast bull case is entirely forward-looking. Anthropic contract is cancellable by either party with 90 days notice, which was noted on the show but may be underappreciated by retail investors.

$NVDA (Nvidia)

Convergence / divergence: Gavin Baker is the clear bull with substantive analysis. Freeberg offered no view. No bear case stated on this episode.

Speaker calls:

• Gavin Baker (bullish, long-term): “The ASIC share-gain narrative — I think Nvidia is growing faster than Broadcom’s AI semiconductor revenue in the quarter just reported, even stripping out China… I think the GPU stays at the center of this constellation for a while. Older GPUs now have a useful life for 10 or 15 years with disaggregated inference. The $20 billion CPU business is extraordinary — overnight one of the world’s largest CPU manufacturers. Jensen is justifiably frustrated that these other AS6 chips are not being submitted for ML Perf benchmarks — they’d lose. Nvidia single-handedly saved the neo-clouds with this quarter.”
• David Friedberg (neutral): “Uhhuh. Yep. Don’t see it.” (No elaboration provided.)

Cross-check:

• Price: $214.28 (P/E 33 trailing by some sources, 45-46 by others; Gavin’s “low-to-mid-teens” is a buy-side forward/non-GAAP consensus estimate, mkt cap $5.29T). Next earnings: ~August 2026.
• Recent headlines: Q1 FY2027: $81.6B revenue (+85% YoY, +20% QoQ). Data Center $75.2B (+92%). Net income $58B. FCF $48B. Gross margin 74.9%. $80B additional buyback. Dividend raised 25x to $0.25/quarter. CPU business guided to $20B this year. Stock +16% YTD.
• Inconsistencies: Gavin said “low-to-mid-teens PE” — trailing PE is 33-46 depending on source. He is referring to a buy-side consensus estimate of forward/non-GAAP earnings, not reported GAAP trailing PE. This distinction matters for retail investors.

$TSLA (Tesla)

Convergence / divergence: Chimath is the sole bull, framing it as a merger-driven thesis. Freeberg described the combined entity math but offered no sentiment.

Speaker calls:

• Chamath Palihapitiya (bullish on merger thesis, long-term): “I still don’t like the fact that Tesla’s over here. As I’ve told you, that will get merged in. And now you have this incredible corpus of physical capability — movement of all kinds — learning capability, infrastructure, all the connectivity. That thing will look very cheap in a few years.”

Cross-check:

• Price: $423.67 (P/E ~370 trailing, mkt cap very large at ~$1.36T implied). Next earnings: ~late July 2026.
• Recent headlines: Tesla owns ~19M SpaceX shares. SpaceX merged with xAI earlier in 2026; Tesla also owns xAI equity. Analyst estimate: only 1.4% probability of Tesla-SpaceX merger announcement by June 30, 2026. Terafab chip collaboration announced March 2026.
• Inconsistencies: Chimath calls merger “inevitable” — analyst market-implied probability is 1.4% by end of June. The interlocking equity structure (Tesla owns SpaceX, both own xAI) creates optionality but is not a confirmed merger path. Tesla’s 370x trailing PE reflects extreme expectations already priced in.

$NET (Cloudflare) and $META (Meta) — AI PR crisis context

These were discussed primarily as examples of poor AI messaging, not as investment theses.

$NET (Cloudflare)

Speaker calls:

• Chamath Palihapitiya: “You could not have written a worse memo. You reduce humans to a label called ‘the measurer’ and then you’re like, I’m going to lay off all the measurers. How does that help anybody? It didn’t need to be done this way. Shut up, get behind the keyboard, just do your job.”

Cross-check:

• Price: $212.65 (mkt cap $74.9B). Q1 2026 revenue $639.8M (+34% YoY). Layoffs: 1,100 people (20% of workforce), cost $140-150M. Stock dropped ~23-24% on announcement despite record revenue. CEO sold $32.6M of stock separately.

$META (Meta Platforms)

Speaker calls:

• Jason Calacanis: Zuckerberg simultaneously announced 8,000-person layoff and mandatory recording software on employee computers to train AI models. “The most you can hope for is you keep this job for some amount of time and train your way out of it, and hopefully there’s some more work for you.”

Cross-check:

• Price: $607.88 (P/E 21.8 trailing / 22.6 forward, mkt cap $1.549T). Next earnings: 2026-07-28.

$CRWV (CoreWeave)

Speaker calls:

• Gavin Baker (bullish, medium-term): “Nvidia single-handedly saved the neo-clouds with this quarter. The true lifespan of a GPU is more like 10 to 15 years with disaggregated inference. CoreWeave’s 6-year contracts at 6% financing — if customers didn’t believe in the lifespan, they wouldn’t be signing. They can’t get enough of them.”

Cross-check:

• Price: $105.49 (unprofitable — Q1 2026 EPS -$1.40, missed estimate; mkt cap $57.5B). 52-week range $63.80-$187.00. Analyst avg target $138.90 (Buy). Q1 2026: revenue $2.08B (+112% YoY, beat estimates).
• Inconsistencies: Stock has fallen from 52-week high of $187 to $105 — the market has not fully repriced based on Gavin’s GPU longevity thesis. Q1 EPS missed by ~54% despite revenue beat, suggesting the path to profitability is slower than estimates.

$GOOGL (Alphabet — SpaceX orbital compute context)

Speaker calls:

• Gavin Baker: “I think it’s much easier to see [SpaceX] selling compute to Google. There have already been posts about that. For sure Google is going to want to be part of orbital compute. Google, Anthropic, Amazon, Nvidia — they are all very convinced that orbital compute is going to be reality.”

Cross-check:

• Price: $382.26 (P/E 29.2, mkt cap $4.64T). Next earnings: ~late July 2026. Q1 2026: $109.9B revenue (+22% YoY), Google Cloud +63% to $20B. Gemini Nano silently installed in Chrome. One dot on the pareto frontier for frontier AI models.
• Inconsistencies: None flagged.

Topics discussed

Andre Karpathy joins Anthropic — recursive self-improvement

Summary: Karpathy (age 39, OpenAI co-founder, Tesla FSD lead, coined “vibe coding”) joins Anthropic to lead pre-training focused on recursive self-improvement — having Claude improve itself. His weekend project “auto research” got 82,000 GitHub stars.

Speaker views:

• Gavin Baker: “Andre has been at the wave upon wave of AI… recursive self-learning puts these models on a combination of overdrive and autopilot. You can potentially live out this idea that there’s an order of magnitude improvement on a yearly basis. So like this new form of Moore’s law.” Also flagged continual learning (model learns from experience like humans) as the second “final frontier.”
• David Friedberg: “Recursive self-improvement and continual learning — those are maybe the two final frontiers for AI. Those combined, they might pull the future forward in a very real way.”
• Jason Calacanis: “Anthropic has a 3-6 month lead on competitors, possibly 6-12 months ahead of open source. You put Karpathy in there… at what point do we think the AI is improving the language model more than the humans in the loop?”

Potential impact: Accelerates Anthropic’s lead. Positive for Anthropic customers (SpaceX/Colossus users, Salesforce) and negative for OpenAI, xAI, and Google if recursive self-improvement produces step-function improvements.

America Turns on AI — the PR crisis

Summary: Three commencement speeches booed. Cloudflare CEO Matthew Prince’s “measurer” memo labeling middle management as obsolete. Zuckerberg’s simultaneous layoff + employee-monitoring announcement. Trump pulled a planned AI executive order (leaked version would have required federal oversight of frontier models).

Speaker views:

• David Friedberg: Backlash stems from (1) perceived power asymmetry — AI benefits a small group creating massive leverage, (2) foreign state actors potentially fueling anti-tech sentiment (citing KGB Cold War techniques as historical precedent), (3) AI feels anti-humanist, threatening human ego: “It’s almost anti-humanist and I think that’s a deep psychological current in a lot of people and their disdain for this technology.” Does not think we should or can slow down.
• Chamath Palihapitiya: “I think there is a CCP-funded campaign against AI and data centers in America. That’s very logical for China, but it’s not good for America.” Agreed tech CEOs are terrible at messaging. Solution: “Shut up, get behind the keyboard, just do your job.”
• Gavin Baker: “We should focus on end-user achievements that were heretofore impossible. Stop listening to model makers about risk — listen to the ICU nurse who says she has more time for patients, the factory worker who added a third shift.”

Potential impact: Risk of regulatory backlash or AI taxation legislation. Structural PR risk if anti-AI sentiment translates to November ballot box outcomes.

SpaceX S1 teardown and the $2T case

Summary: SpaceX filed S-1 on May 20. $75B raise at $1.75T valuation. Nasdaq listing June 12 (SPCX). Revenue breakdown: Starlink $11.4B (2025 annual, +50% YoY), space launch $4B (-$650M op loss), AI/xAI $3.2B (-$6.4B op loss). Anthropic: $1.25B/month. Data center build time improving. Cursor Composer 2.5 pareto-dominant after 3-4 weeks on Colossus 2.

Speaker views:

• Chimath Palihapitiya: “You’re buying the most important internet infrastructure project since the internet itself… I’m buying at 20 times revenue just for the terrestrial data center business by 2030. Everything else is the Colossus.” Tesla merger: “I still don’t like the fact that Tesla’s over here. That will get merged in.”
• Gavin Baker: “15 billion from Anthropic is extraordinary. I think this business can grow dramatically faster.” On orbital compute: “second half of ‘28 to first half of ‘30.” Critical difference vs. peers: SpaceX can use non-radiation-hardened commercial chips in space (a key cost advantage). Rapid reusability of Starship — not just reusability but multiple launches per day — is the gating factor.
• David Friedberg: “If you have a communication network that isn’t restricted and controlled by a government on Earth, it’s almost like a backup for civilization.” Does not own SpaceX shares.

Potential impact: Largest IPO ever if it prices. Immediate competitive pressure on hyperscalers for data center construction speed. Anthropic contract cancellable with 90 days notice by either party.

Nvidia Q1 earnings and the AI cross-sectional valuation problem

Summary: Q1 FY2027: $81.6B revenue (+85% YoY), $58B net income, $48B FCF, 75% gross margins, $5.3T market cap (world’s largest). $80B additional buyback; quarterly dividend raised 25x; CFO says 50% FCF returned. CPU business guided to $20B this year. Gavin’s key insight: power, cooling, and optical company valuations are inconsistent with Nvidia/memory valuations — they can’t both be right.

Speaker views:

• Gavin Baker: “Nvidia is growing faster than Broadcom’s AI semiconductor revenue ex-China, disproving the share-loss narrative. The ASIC companies are not submitting to ML Perf — if they knew they’d win, they’d submit. The $20B CPU guidance is extraordinary — overnight they’re one of the world’s largest CPU makers. Older GPUs now have 10-15 year useful lives with disaggregated inference. Nvidia single-handedly saved the neo-clouds.”
• David Friedberg: “Uhhuh. Yep.” No further elaboration.

Potential impact: If Gavin is right on cross-sectional inefficiency, either Nvidia is cheap or power/cooling/optical names are expensive. The $20B CPU projection reshapes Intel and AMD competitive landscapes.

Bond market and macro — crisis or manageable?

Summary: 10-year at 4.6%. Japan 30-year at 5.1% (all-time high). UK yields at GFC highs. Germany at 2011 highs. Polymarket: 99% chance May CPI at 4.2%+. Survey of professional forecasters: CPI hits 6% in Q2. Korea retail investors borrowing record amounts for AI chip stocks. Fed rate cuts now reversed — hikes being discussed.

Speaker views:

• David Friedberg: “Global debt to GDP is 310%. The roller coaster is at the top and beginning its descent. Japan carry trade could be a catalyst for a credit crisis. I’ll just enjoy the ride.” Characteristically bearish but with no actionable call.
• Chamath Palihapitiya: “Stay concentrated in 5 or fewer high-conviction companies representing the future. Ignore the macro. I have four things I stay on top of. That’s it.” Top two positions represent 20-35% of net worth. Does not speculate — “the vicissitudes of the market don’t give me anywhere near the sugar high it used to on the way up, and it makes me feel horrible on the way down.”
• Gavin Baker: “Three things can be true: rates going up is concerning; AI fundamentals are getting a lot stronger (Anthropic profitable, growing faster than any company in history); America is the best house in a globally bad neighborhood — self-sufficient in energy, food, best public markets, and we benefit relatively from Hormuz closure. I don’t think a dollar crisis is around the corner. The tech bubble happened with 10- and 30-year rates much higher than today. Nvidia at low-to-mid-teens PE is not Cisco at 100x forward earnings.”

Potential impact: If Japan carry trade unwinds, risk of August 2024-style or worse volatility. US energy self-sufficiency and AI revenue growth are the key offsetting positives per Gavin’s framework.