All-In PodcastAll-In Podcast

Elon’s Anthropic Deal, The Next AI Monopoly?, “FDA for AI” Panic, Trading the AI Boom

David Sacks on elon’s compute pivot, AI safety politics, and trading boom narratives.

David SackshostJason CalacanishostChamath PalihapitiyahostJason CalacanishostChamath PalihapitiyahostJason Calacanishost
May 8, 20261h 22mWatch on YouTube ↗
SpaceX–Anthropic compute lease (Colossus)Compute/power constraints and data-center protests“Elon Web Services” hyperscaler thesisAnthropic growth rates and monopoly/duopoly dynamicsAI safety regulation vs regulatory captureCyber-capable models, KYC, and API monitoringTrading/positioning around AI infrastructure and margins
AI-generated summary based on the episode transcript.

In this episode of All-In Podcast, featuring David Sacks and Jason Calacanis, Elon’s Anthropic Deal, The Next AI Monopoly?, “FDA for AI” Panic, Trading the AI Boom explores elon’s compute pivot, AI safety politics, and trading boom narratives The hosts frame the SpaceX–Anthropic compute lease as a strategic win for both sides, easing Anthropic’s power/compute bottlenecks while turning Elon’s data-center buildout into a revenue-generating hyperscaler-like business that subsidizes xAI’s model training.

At a glance

WHAT IT’S REALLY ABOUT

Elon’s compute pivot, AI safety politics, and trading boom narratives

  1. The hosts frame the SpaceX–Anthropic compute lease as a strategic win for both sides, easing Anthropic’s power/compute bottlenecks while turning Elon’s data-center buildout into a revenue-generating hyperscaler-like business that subsidizes xAI’s model training.
  2. They argue that frontier-model revenue is currently gated more by compute and power availability than by demand, and they predict activism and local politics could delay new data-center capacity—raising the premium value of secured power and GPU clusters.
  3. A sharp disagreement emerges over whether Anthropic’s growth trajectory implies an imminent AI monopoly versus premature monopoly panic, with concerns that “AI safety” rhetoric could become regulatory capture that entrenches early leaders.
  4. On the policy front, the panel downplays reports of an “FDA for AI” as largely media-driven, while still acknowledging real near-term cyber risks from more capable models and discussing lighter-touch mitigations like KYC, monitoring, and faster gov–industry coordination.
  5. They connect the AI infrastructure buildout to public markets, claiming hyperscaler acceleration and AI-driven operating leverage are supporting equities, while also debating whether productivity gains are truly AI-caused and when ROI must show up across the broader economy.

IDEAS WORTH REMEMBERING

5 ideas

AI revenue is portrayed as supply-constrained, not demand-constrained.

Chamath argues Anthropic/OpenAI forecasting “beats or misses” mostly reflect power and data-center limits; once compute is unlocked (e.g., via SpaceX capacity), revenue can continue compounding quickly.

Elon’s data-center buildout is being reframed as a standalone cloud business.

Brad and Sacks describe leasing GPU capacity as a way to monetize CapEx immediately, offset xAI losses, and position Elon as a hyperscaler competitor while preserving upside for future space-based compute.

Power acquisition and permitting are becoming competitive moats.

The panel claims organized activist efforts and local backlash can slow grid expansion and data-center projects, so firms that lock in megawatts and sites early gain disproportionate strategic leverage.

Monopoly worries hinge on trajectory, but policy responses could backfire.

Sacks warns that if Anthropic’s growth persists, it could become an unprecedented monopoly; others counter that early heavy-handed regulation would let Washington “pick winners” and create barriers that entrench incumbents.

“FDA for AI” is criticized as a dangerous approval regime—yet cyber risk is acknowledged as real.

They argue pre-release federal approval would slow innovation and encourage capture, while emphasizing a practical need to harden systems as models rapidly improve at offense/defense cyber tasks.

WORDS WORTH SAVING

5 quotes

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.

Chamath Palihapitiya

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.

David Sacks

Unless something about their current trajectory changes, Anthropic will be the most powerful monopoly ever created in human history.

David Sacks

Imagine if John D. Rockefeller was way better at public relations, and instead of calling his company Standard Oil, he called it Safe Oil.

David Sacks

I'd give the community, the tech leaders a D- trending to an F.

Chamath Palihapitiya

QUESTIONS ANSWERED IN THIS EPISODE

5 questions

What are the concrete terms of the SpaceX–Anthropic lease (duration, pricing per GPU-hour, power pass-through), and how much does it improve Anthropic’s effective capacity versus prior cloud contracts?

The hosts frame the SpaceX–Anthropic compute lease as a strategic win for both sides, easing Anthropic’s power/compute bottlenecks while turning Elon’s data-center buildout into a revenue-generating hyperscaler-like business that subsidizes xAI’s model training.

Brad estimates $4–5B incremental revenue from “EWS” this year—what utilization and pricing assumptions drive that number, and how sensitive is it to GPU generation (H100 vs Blackwell)?

They argue that frontier-model revenue is currently gated more by compute and power availability than by demand, and they predict activism and local politics could delay new data-center capacity—raising the premium value of secured power and GPU clusters.

Sacks cites extraordinary Anthropic ARR figures—what exactly is counted as ARR here (API usage run-rate, contracted commitments, net of credits), and how should listeners reconcile these numbers with typical SaaS ARR definitions?

A sharp disagreement emerges over whether Anthropic’s growth trajectory implies an imminent AI monopoly versus premature monopoly panic, with concerns that “AI safety” rhetoric could become regulatory capture that entrenches early leaders.

If “AI safety” can become regulatory capture, what specific proposed rules (compute thresholds, licensing, export controls, model evaluations) most risk advantaging incumbents versus enabling competition?

On the policy front, the panel downplays reports of an “FDA for AI” as largely media-driven, while still acknowledging real near-term cyber risks from more capable models and discussing lighter-touch mitigations like KYC, monitoring, and faster gov–industry coordination.

On the cyber-capable model issue, what would a workable KYC + monitoring standard look like (who qualifies, auditability, privacy boundaries), and how would it apply to open-source releases?

They connect the AI infrastructure buildout to public markets, claiming hyperscaler acceleration and AI-driven operating leverage are supporting equities, while also debating whether productivity gains are truly AI-caused and when ROI must show up across the broader economy.

EVERY SPOKEN WORD

Install uListen for AI-powered chat & search across the full episode — Get Full Transcript

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome