Most important take away

Elon’s lease of the entire Colossus 1 cluster to Anthropic effectively turns SpaceX/xAI into a hyperscaler (“Elon Web Services”), solving Anthropic’s compute constraint while subsidizing Grok’s training and reinforcing SpaceX’s pre-IPO valuation. Meanwhile Anthropic’s ARR jumped from ~$10B to $44B in four months, prompting Sacks to warn it could become the largest monopoly in history if the 10x trajectory holds — a backdrop that makes the New York Times’ “FDA for AI” report (largely shot down by the White House) the most consequential policy fight of the moment.

Summary

Key themes and actionable insights:

• Compute, not demand, is the binding constraint. Anthropic and OpenAI revenue growth is gated by power and data-center supply. Roughly half of the ~9 GW slated to come online this year faces protests from organized activists; expect supply tightening to persist.
• Elon Web Services is real. By leasing Colossus 1 (220K+ Nvidia GPUs, 300+ MW) to Anthropic, Elon adds an estimated $4–5B incremental revenue this year, blunts the “xAI burns cash without revenue” critique ahead of a SpaceX IPO that could trade at 40–50x revenue, and uses the cash flow to fund Grok training. Watch for follow-on moves: distributed compute via Powerwall/Tesla fleet, then orbital data centers via Starlink.
• Anthropic’s trajectory is unprecedented. ARR went $10B → $30B (Q1) → $44B (April). If 10x/year holds, it lands near $100B exiting 2026 and a possible $1T in 2027. Sacks frames this as a potential monopoly larger than the entire Mag 7 combined and warns that “AI safety” rhetoric could function as regulatory capture protecting that monopoly (his Rockefeller/“safe oil” thought experiment).
• Coding is THE wedge. The TAM for software development is ~$1T and likely doubles. Anthropic was the porcupine focused on code while competitors fragmented (Sora, image gen, character chatbots). OpenAI is now refocusing via Codex/GPT-5.5 (Spud base model); xAI tied up with Cursor. Expect competitive pressure on Anthropic’s lead within 6–12 months.
• “FDA for AI” is mostly fake news, but the vibe shift is real. Hassett and Bessent do NOT want a pre-approval regime; Susie Wiles publicly shot it down. The actionable concern is cyber-capable models (Mythos, soon OpenAI’s equivalent and open source within 3–6 months). Solution path: KYC for API access during preview periods, get models into hands of CrowdStrike/Palo Alto/cyber startups for defensive pen-testing, harden systems now.
• Tech leaders are losing the PR war. Chamath grades the industry “D-minus trending to F” on broad-based benefit messaging. Actionable proposals raised: IPO-K (give 1–5% of frontier-AI IPOs into Invest America accounts for every American), free electricity for households near data centers (Abilene model), and reframing AI’s deflationary impact on cost-of-living (the #1 voter issue; AI ranks 29/39 in salience).
• Trade the boom but watch the fork. Multiples are not bubble-territory: Meta 17x, Nvidia 19x, Microsoft 20x, Google 24x, memory names (SK Hynix 5x, Samsung 6x, Micron 7x). Hyperscaler growth re-accelerated: AWS +28%, Azure +39%, Google Cloud +63%. Brad went heavily long in January when AI revenue showed up; ~80%+ exposure to compute/AI/memory.
• Chamath’s bear-case timeline: ~500 days of “be net long,” then a reckoning when enterprises must show that AI tokens lifted operating margins. S&P operating margins are up ~200 bps to 13%, but it’s unclear how much is AI vs. post-COVID fitness. Two divergent paths: (a) opex shrinks (workforce contraction) or (b) revenue grows faster than opex. The political and market response diverges sharply between those outcomes.
• Macro setup is Goldilocks: 10-year at 4.3, inflation contained, accelerating GDP (AI was 75% of Q1 GDP growth), construction wages up 25–30%, and recent college graduate employment improving. Policy credit goes to Trump rescinding Biden’s flops-based model approval regime week one and unleashing energy buildout.

Chapter Summaries

• Spencer Pratt and California politics — Cold open praising Pratt’s social-media-driven LA mayoral campaign against Karen Bass; Chamath bullish on a CA constitutional amendment banning a wealth tax; Ken Griffin moves political spend from NY to FL after Mamdani targets his home.
• Elon’s Anthropic deal — Elon leases all of Colossus 1 to Anthropic (220K+ GPUs, 300+ MW). Discussion of “Elon Web Services,” the five-layer SpaceX cake (launch, connectivity, compute, hyperscaler, space), and how this de-risks the SpaceX IPO while subsidizing Grok.
• Anthropic’s growth and the monopoly question — ARR $10B → $44B in four months. Sacks argues if 10x/year holds, Anthropic becomes the largest monopoly in human history, dwarfing the Mag 7 combined; Brad pushes back that it’s too early to invoke antitrust.
• The Rockefeller “safe oil” thought experiment — Sacks’s satirical framing that AI-safety advocates may be (intentionally or not) running the same regulatory-capture playbook a savvier Rockefeller could have used.
• “FDA for AI” panic — NYT reports Trump considering an executive order; Hassett’s FDA analogy muddied waters but Bessent and the White House Chief of Staff push back. Mythos/cyber capability is the real catalyst.
• Cyber, KYC, and self-policing — Anthropic and OpenAI both self-restrained on releasing cyber-capable models. Consensus on KYC for API access during preview periods, government-industry coordination, and getting tools to defenders fast.
• The vibe-shift and tech’s PR failure — Chamath argues tech leaders failed to deliver/communicate broad-based benefit; antibodies are forming. Jason proposes IPO-K, minimum-wage reform, universal healthcare debate; Brad proposes free electricity in data-center towns.
• Trading the AI boom — Hyperscaler revenue acceleration (AWS, Azure, GCP), Mag 5 multiples, memory stocks, Brad’s portfolio positioning, market at all-time highs but not bubble territory.
• The fork in the road — Chamath’s 500-day thesis: enterprises must eventually prove ROI on AI tokens via margin expansion, with two divergent paths (workforce shrinkage vs. revenue growth) carrying very different political consequences.
• Labor and macro wrap — Unemployment, college-grad hiring, labor-force participation debate; closing salute to American innovators.