Why this is in the vault
Thompson's post-mortem on Jensen Huang's Nov 12 resignation as Nvidia CEO: it frames the episode as several "obituaries" (Huang's tenure, Nvidia's moats, the AI funding cycle, AI as normal technology, tech independence from government, American exceptionalism), so it is the analytical capstone to the vault's running Huang/Nvidia coverage.
The core argument
- Huang's downfall is read as structural, not just the affairs: Nvidia was treated as a defense contractor and he never adjusted, doing the minimum under the letter of chip controls rather than the spirit.
- Nvidia's three moats are all weaker: CUDA is eroded by LLMs writing software and Anthropic running on non-Nvidia chips; networking/systems is challenged by Google; supply-chain mastery was a Huang personal coordination function.
- AI funding has moved from free cash flow to debt to equity. Nvidia's backstopping of neoclouds ("Risky Business") depended on one unconstrained leader and is unlikely to continue; hyperscalers will prioritize their own compute if revenue falls short.
- The doomers (represented by Anthropic) and China hawks have won on policy: AI will not be treated as normal technology. Thompson disagrees on China, arguing demand-side pull (the Great Firewall analogy) builds rival ecosystems.
- Treasury Secretary Bessent wants Nvidia to behave like a bank (know-your-customer, spirit of the law). Thompson calls this a break with tech's long-standing assumption of selling to everyone, and ties it to his earlier Anthropic and Alignment argument about government's say over exceptional technology.
- Closing note is sympathetic to Huang as an American Dream story, critical of ban-and-compliance as a China strategy versus innovation and openness. Thompson discloses alignment with the "AI as normal technology" view.
Source note: written from the email body, which rendered in full; no web reconstruction was needed.
Mapping against Ray Data Co
Strongest link is the vault's open Nvidia/Huang thread (the Nov 10 Washington story and the Nov 13 weekly digest and Sharp Tech note), which this post supplies the thesis for. For the L5 and agents-in-production work, the practical takeaways are the supply-side concentration risk (Anthropic's multi-chip independence is what Thompson credits with removing CUDA dependency) and the policy shift toward treating frontier AI as a regulated, government-accountable asset, which affects how phData client conversations about AI vendors and compliance posture should be framed. The investing frame (capital-cycle, alpha) gets a concrete bear case for neocloud credit and Nvidia's circular financing. Medium-strong; no project decision changes.
Related
- [[2026-11-13-stratechery-this-week-2026-46-the-end-of-jensen]]
- [[2026-11-13-stratechery-sharp-tech-nvidia-after-jensen-huang]]
- [[2026-11-10-stratechery-jensen-huang-story-gets-worse-washington-ny-silicon-valley]]
- [[2026-11-12-innermost-loop-nvidia-ousts-huang-anthropic-ipo-pricing-intelligence-explosion-2027]]
- [[project_investing_markov_capital_cycle]]
⚠️ Sponsorship
Issue carried a "Presented by Stripe" banner (house sponsor relationship, paid placement). It sits outside the article; no bias in the Nvidia analysis is evident. Separately, Thompson's views on Anthropic and China chip sales are a stated, consistent position, not a sponsor effect.