A recent Financial Times report says Chinese AI companies have bought more than 1 million Nvidia GPUs in a single year. The chips are the same gaming and workstation cards that many gamers use, but they are being repurposed for AI workloads. Insiders say the buying spree shows that Chinese firms still depend heavily on U.S. hardware, even as domestic alternatives grow.
Nvidia did not comment on the total figure, and the companies involved were not named. The report points out that the GPUs are likely being used to fill a short‑term gap in compute capacity while Chinese manufacturers scale up their own AI chips. Export‑control rules that restrict some high‑end AI accelerators appear to be a factor, pushing buyers toward more widely available gaming GPUs.
What the numbers mean
A million GPUs is a sizable chunk of Nvidia’s yearly shipments. If each card costs roughly $500 – $1,000, the total spend could be in the $0.5 – $1 billion range. For Chinese AI startups and larger firms alike, that amount of hardware can power dozens of large language model trainings or run inference for many services.
The influx of U.S. GPUs into China also raises questions about supply chain resilience. While Nvidia is expanding its own production, the demand from abroad could tighten availability for other markets, especially if export controls tighten further.
Why it matters
For Chinese AI developers, the GPUs provide a quick way to scale up compute without waiting for domestic chips. For the global market, the large order could affect Nvidia’s inventory and pricing, and it shows how export rules can shift demand toward more widely available hardware.
