Chinese companies have purchased more than 1 million NVIDIA gaming and workstation chips, market research firm TrendForce reported, as domestic AI accelerators fail to cover demand.

The figure covers consumer-grade and workstation graphics processors — hardware not designed or sanctioned for large-scale AI training — repurposed by Chinese buyers to fill capacity gaps left by restrictions on NVIDIA's data-center products.

Why are gaming chips doing AI work?

US export controls block sales of NVIDIA's most powerful data-center accelerators to Chinese customers. Domestic alternatives have not closed the gap. TrendForce reports that supply of Chinese AI silicon falls short of demand, pushing firms toward gaming and workstation GPUs as a workaround.

These chips deliver far lower compute density than restricted data-center parts, so buyers compensate by deploying them at scale — a strategy the reported 1-million-unit volume makes concrete.

What does the number signal?

A seven-figure purchase volume indicates three things:

• Chinese AI compute demand exceeds what domestic suppliers can ship today.
• Workarounds now run through consumer and professional GPU channels, not data-center sales.
• Enforcement gaps persist in non-data-center hardware categories.

TrendForce, a Taipei-based memory and semiconductor research firm, compiled the figure; it did not name the purchasing companies in the report summary.

Who is affected?

The report positions Chinese AI developers as the buyers and NVIDIA as the displaced supplier of last resort. Domestic accelerator makers gain a demand tailwind they cannot yet satisfy, according to TrendForce's assessment that local AI chip supply falls short.

The 1-million-unit mark gives policymakers and suppliers a measurable baseline for tracking how export restrictions redirect, rather than eliminate, chip flows into China.

The takeaway: over 1 million gaming and workstation GPUs have become China's stopgap AI infrastructure.