Fact-first explanation of the immediate and medium-term effects on NVIDIA if a Chinese attack on Taiwan occurred today.

NVIDIA designs GPUs but does not own the fabs that make its advanced dies, it relies primarily on TSMC (Taiwan Semiconductor Manufacturing Co.) for leading-node manufacturing (logic wafers) and on TSMC’s advanced packaging capacity for assembling chips. A kinetic conflict in Taiwan would almost immediately disrupt wafer production and advanced packaging flows that NVIDIA depends on.

NVIDIA had booked a very large share of TSMC’s advanced packaging capacity for 2025 (reports say 60–70 per cent + of certain advanced packaging lines), meaning any stoppage would choke NVIDIA’s ability to finish chips even if wafers existed elsewhere. Advanced packaging is a critical, hard-to-move step; it’s concentrated in Taiwan and nearby supply chains. Losing that capacity would rapidly halt shipments of new GPUs.

China and Hong Kong combined were reported to account for roughly 13 per cent of NVIDIA’s revenue in 2024/25; Taiwan accounted for another 15–16 per cent in some breakdowns, together they are material but not the company’s largest single market. A Taiwan conflict would therefore hit both supply (manufacturing) and a chunk of demand (regional customers), but NVIDIA’s largest revenues come from the US and global cloud/data-centre customers.

Since 2024–25 the US imposed export controls restricting sales of top-end AI GPUs (H100/H200/H20/Blackwell families) to China; China in turn has been trying to source or produce domestic alternatives. Those controls mean NVIDIA had already lost much of its previous China footprint for flagship GPUs, reducing near-term China sales but increasing reliance on non-Chinese customers. In short: demand from mainland China was already curtailed before any military disruption.

Markets treat Taiwan as the world’s semiconductor backbone. News of a conflict would likely trigger a sharp sell-off in NVIDIA stock and in semiconductor indices, and cause volatility in vendors, cloud providers and Asian markets. Analysts and investors would price in months of lost output and severe backlog for data-center customers. Recent market episodes show how geopolitics quickly reroutes valuations.

NVIDIA has secondary suppliers (Samsung, and to a smaller extent other foundries) and has announced cooperation with Samsung on an “AI factory” and other capacity moves, but moving advanced-node production and advanced packaging at scale is not immediate. Building or qualifying alternative capacity (wafer fabs + advanced packaging lines + test/assembly ecosystems) takes many months to years. Ordering extra wafers or using partners can blunt shortfalls, but cannot replace a sudden full stoppage in Taiwan.

Data-center operators and cloud providers would face delayed GPU deliveries, forcing prioritisation of existing inventory (national security customers, high-value contracts). AI training projects would slow; customers might turn to older generations, alternative accelerators (AMD, Habana-class chips, in-house accelerators), or to cloud providers with stockpiles. Supply-chain rationing and re-allocation would be the immediate operational response. (Market reports and NVIDIA guidance in 2025 already foreshadow how constrained supply cascades.)

A real Taiwan conflict would prompt: (a) accelerated diversification of chipmaking to Korea, the U.S., and possibly new Southeast Asian sites; (b) deeper vertical integration and onshore production for critical customers; (c) long lead-times for capacity replacement, all of which would raise costs and likely reduce NVIDIA’s near-term margins and growth. US policy responses (sanctions, export controls, military posture) would also shape which customers and factories could operate — adding regulatory complexity.

If China attacked Taiwan today, NVIDIA’s immediate problem would be a sudden halt or severe slowdown in chip production and assembly because TSMC and Taiwan-based advanced packaging are critical to NVIDIA. That would cause rapid supply shortages, customer re-allocation, stock market shock, and months-to-years of industrial re-tooling to restore global GPU supply at scale. NVIDIA could absorb demand shocks in the medium term using Samsung and other partners, but only after long, expensive capacity shifts and with significant cost and revenue disruption.