May 22, 2026 · 1h 41m · allin
SpaceX’s $2T Case, Nvidia’s Shock Selloff, America Turns on AI, Trump Pulls AI Order, Bond Crisis?
Jason Calacanis · 28m spoken Gavin Baker · 28m spoken Chamath Palihapitiya · 19m spoken David Friedberg · 11m spoken Mark Zuckerberg · 35s spoken Shyam Sankar · 19s spoken
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In episode 274 of the All-In Podcast, guest host Gavin Baker joins Jason Calacanis, Chamath Palihapitiya, and David Friedberg to analyze major developments in artificial intelligence, technology regulation, macroeconomics, and geopolitics. The panel evaluates Andrej Karpathy's move to Anthropic, SpaceX's $2 trillion valuation case, Nvidia's financial performance, and societal backlash against AI alongside U.S.-China strategic competition.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 66.8% of the talking time here. How this is scored →
The hosts as informed peer 6.2 Guest teaching 5.0 Guest disagreement 1.9 The hosts pushing back 2.5
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 40:27 Chamath blasts tech CEOs for horrific PR memos
Chamath aggressively attacks corporate leadership and PR teams over tone-deaf layoff communications, telling CEOs to shut up and get behind the keyboard.
Hardest push from the hosts ▶ 9:46 Chamath refutes Jason's 'covert' deployment claimChamath explicitly interrupts and rejects Jason's characterization of Google's Chrome Gemini Nano rollout as covert, warning that such framing drives Luddite backlash.
Biggest teaching moment ▶ 1:19:53 Gavin dismantles the short case on GPU depreciationGavin explains how domain-specific accelerators paired with GPUs extend hardware lifespans to 10-15 years, disproving short seller theses and enabling 6% asset-backed debt for NeoClouds.
The host holds their own ▶ 59:50 Chamath presents a detailed $2T SpaceX valuation modelChamath demonstrates deep financial and operational expertise by detailing revenue compounding, gigawatt data center execution, and DC-to-DC power transformations.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Andrej Karpathy Joins Anthropic & Recursive AI Learning | 6 | 4 | 1 | 1 | Chamath and Jason analyze Andrej Karpathy moving to Anthropic and the financial scale of AI labs reaching 100B ARR. Gavin complements the discussion by highlighting gross margins and recursive self-improvement dynamics. | |
| Local AI & Debate over Google's Chrome Gemini Nano Deployment | 6 | 3 | 2 | 7 | Chamath immediately interrupts Jason for using the word 'covert' regarding Google's Gemini Nano Chrome rollout. Chamath argues that breathless fear-mongering turns AI into a dirty word and urges focusing on end-user utility instead. | |
| CEO Incentives, Regulatory Moats & Positive AI Use Cases | 6 | 6 | 1 | 2 | Chamath breaks down how AI CEOs like Dario Amodei build regulatory moats to protect their businesses. Gavin educates the hosts with a poignant real-world story of a father using LLMs to find a targeted treatment for his daughter's rare genetic mutation. | |
| Philosophical & Geopolitical Roots of the Anti-Technology Backlash | 8 | 4 | 1 | 2 | Friedberg delivers an intellectual monologue tracing public backlash against technology back to nuclear weapons, foreign disinformation tactics, and non-human-centric psychological fears. Chamath probes whether AI progress should be slowed down. | |
| Proposed Interventions: Phased Autonomous Rollouts & Robot Taxes | 7 | 3 | 2 | 6 | Chamath rejects Jason's premise on job protection and robot taxes, pointing out high warehouse turnover and asking if critics have actually surveyed workers. He also notes the pulled presidential executive order and advocates for AI detente with China. | |
| Frontier Model Regulation, Tort Liability & Autonomous Vehicle Safety | 6 | 5 | 4 | 5 | Jason presses Gavin on whether frontier models require pre-release safety testing. Gavin resists top-down government mandates, emphasizing tort liability and how state competition drives better outcomes. | |
| AI in Public Safety: ShotSpotter vs. Las Vegas AI Police Command | 6 | 5 | 1 | 1 | Gavin highlights Cambridge turning off gunshot detection, which the hosts mock. Chamath recounts visiting the Las Vegas real-time AI command center, explaining how de minimis costs yield massive public safety returns. | |
| Corporate Layoffs & PR Blunders (Cloudflare & Meta Leaked Audio) | 7 | 2 | 3 | 4 | Jason reviews Cloudflare's memo on firing 'measurers' and Meta's employee tracking. Chamath strongly denounces Cloudflare's PR approach, calling it tone-deaf and telling tech executives to stop writing destructive missives. | |
| Memes & Banter: Anthropic Hires & Chamath's Leg-Maxing Photo | 2 | 1 | 1 | 1 | The hosts engage in lighthearted banter, poking fun at satirical news of Sam Altman and Tucker Carlson joining Anthropic, as well as teasing Chamath about his workout photos. | |
| SpaceX S-1 Teardown and Elon Web Services | 5 | 2 | 1 | 1 | Jason summarizes the key figures from SpaceX's S-1 filing, detailing revenue splits across Starlink, space operations, and Elon Web Services hosting Anthropic. | |
| Gavin Baker Analyzes Compute Clusters, Cursor, and Grok | 5 | 8 | 2 | 1 | Gavin provides an in-depth breakdown of SpaceX's rapid data center construction, Cursor's Composer 2.5 Pareto dominance, proprietary coding data advantages, and Grok Build's agentic harness. | |
| Space-Based Compute as a Shield Against Tyranny | 7 | 5 | 1 | 1 | Friedberg argues that orbital compute and space-based communication networks serve as a critical uncensorable backup for civilization against government regulation and data center restrictions. | |
| Underwriting SpaceX Valuation and Launch Dominance | 8 | 4 | 2 | 2 | Chamath lays out the financial and operational case for underwriting SpaceX at a two trillion dollar valuation, detailing gigawatt data center execution and DC-to-DC power efficiency. | |
| Starship Rapid Reusability and Orbital Compute Feasibility | 5 | 8 | 2 | 1 | Gavin clarifies the difference between reusability and rapid reusability for Starship. He notes that NVIDIA H100 GPUs and off-the-shelf commercial chips are already functioning in orbit. | |
| Nvidia Record Quarter and Semiconductor Disconnect | 6 | 8 | 3 | 2 | Jason presents NVIDIA's record quarter. Gavin addresses cross-sectional market inefficiencies, refuting narratives that NVIDIA is losing market share to Broadcom or custom ASICs. | |
| Domain Architectures, GPU Lifespan, and NeoCloud Financing | 7 | 8 | 3 | 2 | Chamath notes domain-specific architectures within NVIDIA, while Gavin explains how combining domain accelerators extends GPU lifespans to 10-15 years, debunking short seller arguments and securing low-cost debt for NeoClouds. | |
| Spiking Inflation, Bond Yields, and Macro Headwinds | 7 | 5 | 3 | 4 | Jason outlines rising bond yields and inflation fears. Friedberg warns of severe debt-driven macro collapse, while Chamath advocates ignoring market noise to hold a concentrated portfolio of top companies. | |
| Portfolio Positions, AI Seasonality, and US Energy | 6 | 8 | 2 | 2 | Gavin details his portfolio approach and explains why high energy prices harm Europe and Asia far more than the self-sufficient United States, while introducing the concept of AI usage seasonality. | |
| Evaluating the US-China Presidential & Tech CEO Visit | 6 | 4 | 2 | 3 | Jason and Friedberg question whether the US-China presidential summit produced meaningful structural changes. Chamath contends that behind closed doors, critical geopolitical alignment was achieved. | |
| Nvidia Chip Exports and Global Energy Geopolitics | 7 | 7 | 2 | 2 | Gavin argues that selling deprecated GPUs to China deters them from developing competing ecosystems, while US control over global energy bottlenecks provides significant leverage against conflict. |
Statements from this episode (45)
Chamath: Karpathy spent 25% of his time hand-labeling Tesla video data
“He spent a portion of his time. I want to say a quarter of his time labeling data. Could you imagine like 20, 1617, like hand labeling video data from Tesla's.”
Baker: Anthropic was EBIT positive in the most recent quarter
“They were EBIT positive per the Wall Street Journal in the most recent quarter”
Baker: Total LLM ARR could reach $200B-$400B by end of year
“It's not hard to see 203 104 hundred billion dollars of ARR at the end of this year at a high margin.”
Calacanis: Google included 4GB Gemini Nano in Chrome without telling users
“Chrome included Gemini or Google included in their Chrome browser, the Gemini Nano model without telling anybody four gigabytes on your computer.”
Chamath: Breathlessly discussing every AI model update is a waste of time
“We're in the phase now where I think breathlessly talking about every model improvement is a waste of time.”
Chamath: OpenAI used AI and human collaboration to solve decades-old math problem
“So for example, we were able to solve, I'm just collectively saying we, in this case, it was specifically open AI. By the use of a human, and this is important, you know, a math problem that stood outstanding, not been solved for decades and decades.”
Chamath: AI is bringing previously discarded drug candidates into clinical trials
“There are drug candidates that are about to enter clinical trials and INDs that were sitting on the shelf and people didn't think were very viable at all.”
Baker: A CCP-funded campaign may be targeting US AI and data centers
“It is starting to feel or seem like there may be a CCP funded campaign against AI and data centers in America.”
Chamath: Dario Amodei warns of AI risks to build regulatory moats
“It makes enormous sense for Dario to try to create the boundary conditions for a regulatory moat, because he will be inside of the tent pissing out.”
Friedberg: Foreign state actors historically fuel anti-tech sentiment to slow progress
“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 durin…”
Friedberg: Slowing US AI development risks disrupting global geopolitical balance
“If we do try and slow down AI, we kind of lose this moment of balance that's necessary when you have a technology proliferation, like we saw with the arms race after World War II, that, you know, we're gonna see again here.”
Chamath Palihapitiya: Amazon has 35% to 40% warehouse employee churn
“Amazon has 35 or 40% churn inside of their warehouses.”
Chamath Palihapitiya: Presidential AI executive order was scrubbed at last minute
“There was supposed to be an EO, a presidential executive order that was announced today, and then it was pulled, it was scrubbed at the last minute.”
Chamath Palihapitiya: China is less than nine months behind US in AI
“I think it's actually good that China is less than nine months behind us.”
Chamath: Cities banning autonomous vehicles will face wrongful death suits
“They're going to get sued for wrongful death. When somebody runs over somebody else, and you could have implemented a solution that has a zero death rate.”
Baker: Visiting cities without autonomous vehicles will feel barbaric and unsafe
“I think going to a city where you can't get in a Waymo or a cyber cab is going to feel barbaric and unsafe.”
Baker: Cambridge City Council voted to disable gunshot detectors
“I think that the Cambridge city council voted to turn off gunshot detectors two days ago.”
Chamath: $30M-$40M annually would make Las Vegas America's safest city
“If you gave the Las Vegas Police Department 30, forty million dollars a year, it would be the safest city in America, and that's all it would take.”
Zuckerberg: Meta's internal engineers will accelerate AI coding progress beyond rivals
“Having people internally build tools that, or solve tasks that that help teach the model how to code, We think is going to dramatically increase our model's coding ability faster than what others in the industry have the capability to do, who don't have thousa…”
Sankar: Public dialogue should focus on frontline AI users, not AI inventors
“We're listening too much to the inventors of AI. I know that's appealing. They're geniuses. They're smart. We need to be listening to the frontline factory workers who are using AI saying, wow, I was able to add a third shift. I was able to hire more workers. …”
Chamath: Big tech doing AI layoffs will announce buybacks and dividend increases
“Wait till the next Regulatory filing comes out from these companies and they authorize a massive share buyback and an increase in their dividend.”
Calacanis: Starlink generated $11.4B in revenue and $4.4B in operating income
“Starlink did 11.4 billion in revenue last year on 50% growth with 4.4 billion in operating income.”
Calacanis: Anthropic signed $45B compute rental deal with SpaceX
“Anthropic is paying SpaceX, wait for it, 1.25 billion a month. To rent out Colossus One and parts of Colossus Two. It's a forty five billion dollar deal over three years, fifteen billion a year.”
Gavin Baker: Cursor allegedly has more coding tokens than public internet
“Cursor has, we will never know, but Cursor allegedly has more tokens of coding data than exist on the public internet.”
Gavin Baker: Major tech firms are convinced orbital compute will become reality
“Google, Anthropic, Amazon, NVIDIA. They are all very convinced that orbital compute is going to be reality, and obviously SpaceX is extraordinarily well positioned for that.”
Gavin Baker: Only xAI, Google, OpenAI, and Anthropic are AI frontier companies
“And the companies on the frontier are XAI with one build of Grok 4.3, which is a five hundred billion parameter model. Google, 3.1 pro, and then OpenAI and Anthropic, and that's it. Those are the companies on the frontier.”
Palihapitiya: SpaceX revenue will probably reach $25B-$30B this year
“Well, last year it did 18, nineteen billion dollars. It'll probably do 25 to 30 this year.”
Palihapitiya: SpaceX revenue will reach $40B-$45B next year and double after
“So I suspect what happens is next year, it's probably 40, forty five billion. And then the year after that, it probably doubles again.”
Palihapitiya: Tesla will eventually merge into Elon Musk's wider ecosystem
“I still don't like the fact that Tesla's over here. And as I've told you, that will get merged in.”
Palihapitiya: Terrestrial data centers will generate $100B-$200B revenue by 2030-2032
“So if you had to pick an underwriting case, Jason, I would flex the revenue and realize that terrestrial data centers alone are a hundred or two hundred billion dollars of revenue by 2030, 2032.”
Baker estimates SpaceX will achieve rapid Starship reusability within 1-2 years
“I don't know when. I don't want to make a prediction. I would guess a year or two. Maybe sooner.”
Baker: An Nvidia H100 GPU is currently operational in space
“There is a working H-one-hundred, an NVIDIA H-one-hundred GPU in space today. Andre Karpathy both trained a model on and used for inference.”
Baker predicts orbital compute will become a reality between 2028 and 2030
“Second half of 28 to First half of 20 30 would be my point prediction.”
Baker: AI hardware sector valuations are cross-sectionally inefficient and inconsistent
“Cross-sectionally, if you look at the valuations for all these AI names, they just, they can't all be accurate... The AI market is cross-sectionally inefficient right now, which is what I was trying to say.”
Baker: Nvidia's Western AI business grows faster than Broadcom's
“Within the Western AI world, Within data centers that need being built, whether they're being built by CoreWeave, XAI, Amazon, Google, NVIDIA's AI business is growing faster than Broadcoms.”
Baker: Competitors withhold custom AI ASICs from benchmarks because they would lose
“These other ASICs. . Are not being submitted for benchmarks. They're not in the CB analysis inference max. They're not in MLperf. I think the reason they're not being submitted is they will lose.”
Chamath: Domain-specific architecture evolution is happening inside Nvidia
“The reality is that that DSA Market evolution is actually happening inside of Nvidia.”
Gavin Baker: Older GPUs have a useful life of 10 to 15 years
“All GPUs use Grok or Cerebrus for decode, and then those older GPUs, they have a useful life for 10 or 15 years, and this means that you can finance GPUs.”
Calacanis: Polymarket gives 99% odds May inflation exceeds 4.2%
“Polymarket says, 99% chance May inflation comes in at 4.2% or higher.”
Friedberg: Japanese bond yield spikes could trigger a credit crisis
“I would think that this is one of those things that could be a catalyst for a credit crisis because there's a lot of people that are in this carry trade and we'll see, you know, this is.”
Gavin Baker: Nvidia trades at low to mid-teens multiple of real earnings
“And, you know, these, the Nvidia of the tech bubble is Cisco and it traded a hundred times forward earnings. And, you know, I think Nvidia is probably at a low teens, multiple of kind of low to mid teens, multiple of real earnings.”
Friedberg: US-China summit failed to produce grand de-escalation deal
“And, you know, there were some purchases of aircraft and some agricultural product commitments, but fundamentally the grand deal, the big deal that I would say reduces, deescalates tension probably didn't manifest as some had hoped it would.”
Gavin Baker: Selling deprecated Nvidia GPUs to China delays Chinese semiconductor independence
“I think selling in deprecated Nvidia GPUs to China. Lowers the odds of them developing their own alternative ecosystem, which would be a lot power hungrier because you use optical, you bring in optical a lot earlier for scale up fabrics.”