\n\n
Key developments: Chinese technology firms have purchased over 1 million Nvidia gaming and workstation GPUs this year, many of which are unsanctioned and lower‑performance parts that are nonetheless plentiful. At the same time, the cost to rent Nvidia H100 AI compute has surged 60% over the past 12 months, reflecting sky‑high demand across the sector. Industry insiders estimate that total global demand for AI‑grade chips this year will top 5 million units, while domestic supply – even with an expected 3.5 million AI chips, including roughly 1 million made abroad – falls far short.
Why it matters: The influx of Nvidia GPUs into China highlights how the country is turning to foreign‑sourced hardware to keep pace with its aggressive AI rollout, despite sanctions and limited domestic alternatives. The sharp rise in rental pricing signals tighter supply for data‑center‑grade GPUs, which could drive up operating costs for AI‑heavy firms worldwide and accelerate a shift toward alternative architectures or more efficient chip designs.
What to watch next: 1) Whether Nvidia reallocates production from consumer‑grade GPUs (such as the RTX 5090) toward AI‑focused dies, further constraining the gaming market. 2) The evolution of China’s domestic chip programs and any policy changes that might curb or expand imports of Nvidia hardware. 3) Price trends for GPU rentals and the emergence of new suppliers or second‑hand market dynamics that could ease the current shortage. [[1]]