Nvidia CEO says company went from 95% China market share to 0% — Palantir's Shyam Sankar fires back in the WSJ
Oct 20, 2025
Key Points
- Nvidia CEO Jensen Huang claims the company lost all China market share due to US export controls, but continued sales through Singapore and Malaysia suggest transshipment to Chinese customers.
- Palantir CTO Shyam Sankar challenges the 'China dove' narrative in the Wall Street Journal, arguing Beijing views competition as zero-sum and citing recent export controls on rare earths as proof economic conflict is already underway.
- JPMorgan estimates Nvidia's projected $3-4 trillion AI capex by 2030 is financially feasible with $1 trillion in new debt, but the real constraint is generating enough revenue to justify the spending.
Summary
Jensen Huang claimed at a Citadel event that Nvidia has gone from 95% market share in China to zero. Shyam Sankar, Palantir's CTO, pushed back in a Wall Street Journal op-ed titled "Why the China Doves Are Wrong."
Huang's claim drew scrutiny because Nvidia continues operating in Singapore and Malaysia, jurisdictions observers say lack clear business rationale unless they serve as transshipment points for Chinese demand. Huang has also argued publicly that the "China hawk" label carries shame rather than honor and that US-China relations need not be adversarial.
Sankar argues that American business leaders refusing to acknowledge the Communist Party's intent to weaken the US are deluding themselves. He cites China's recent export controls on lithium batteries and rare earth materials as evidence of economic competition and references President Trump's characterization of the move as "sinister and hostile." Sankar frames the dynamic as economic war already underway, not a future risk. The CCP views great-power competition as zero-sum, he argues, with China rising and the US falling. The first step to ending dependence on China is admitting the problem exists.
Semiconductor control
The chip control lever appears narrower than it initially seems. China has shown willingness to reject Nvidia's constrained offerings like H20 chips in favor of Huawei alternatives and internal development. This signals a multi-year bet on semiconductor independence rather than immediate need. China's strategy appears to be building a domestic stack by the time the next major AI scaling run arrives, betting it can compete on frontier AI without US chips.
Rare earth materials loom larger as a leverage point. China's recent export controls on lithium batteries and rare earth products give Beijing supply-chain leverage in ways semiconductor restrictions alone do not.
Capital requirements
JPMorgan estimates that Huang's projection of AI capital expenditure growing from $600 billion today to $3–4 trillion by 2030 is financially feasible but ambitious. The bank expects roughly $1.2–1.6 trillion in annual funding gaps. Private markets could cover about $500 billion annually by 2030, leaving the remainder to leverage. JPMorgan projects $430 billion in bank loans and $600 billion in bond issuance, roughly $1 trillion in new debt entering the system. Even with this increase, the tech industry's net debt-to-cash-flow ratio would rise from 0.7x to 1.2x, below the global average.
The real constraint is not balance-sheet capacity but revenue. The $3–4 trillion capex figure requires corresponding revenue growth to justify the spend. That growth depends on monetizing agent-based AI products like developer tools and GitHub Copilot at scale.