새 보고서에 따르면 중국 AI 기업들이 올해 최대 100만 개의 Nvidia GPU를 구매했을 가능성이 있다.
Chinese technology companies have bought more than 1 million Nvidia gaming and workstation GPUs to fill gaps in demand for AI computing power, according to the Financial Times. The report cites industry insiders who say the buying spree shows a continuing and significant reliance on U.S. AI hardware, despite a growing domestic chip market.
The report notes that China's national strategy for developing AI has led to a doubling of domestic AI chip production this year, but the country still cannot keep up with demand from its largest technology companies. It cites a shortfall of nearly 1.5 million chips: only 3.5 million are expected to be manufactured domestically, while demand "far exceeds" 5 million. To compensate, China is reportedly turning to Nvidia's gaming and workstation processors, which comply with U.S. export controls, and repurposing them for AI workloads.
According to the FT, the chips are "mostly" Nvidia's RTX Pro 6000D, RTX Pro 5000, and RTX 5090, the last of which is among the best graphics cards on the market for enthusiast gaming. The report notes that these Blackwell-based chips cannot compete with Huawei's Ascend 910C and other homegrown flagship products, but they are useful for "simpler AI workloads," such as prefilling.
The move adds pressure to a GPU market where prices for enthusiast cards have already climbed sharply. The RTX 5090 has a list price (MSRP) of $1,999, but it can now be found at only three third-party resellers, at no less than $7,199, and has nearly disappeared from the market.
The report also highlights that demand for Nvidia's GPUs is driven in part by its CUDA software, which remains more attractive to developers than Huawei's CANN alternative.
In the enthusiast market, it was recently reported that Nvidia has halted RTX 5090 production and will replace it with a 24GB RTX 5080 flagship. That would be a significant step down from the 32GB the RTX 5090 currently offers, a troubling sign that the enthusiast gaming GPU market may be receding rather than advancing.