While Chinese companies are banned from acquiring Nvidia Blackwell and other AI GPUs that offer the same or better performance containing U.S. technologies, a loophole has so far allowed them to gain access to compute by renting them from data centers abroad. The U.S. House is attempting to patch it and passed the Remote Access Security Act , but it must still go through the Senate and the White House before it can turn into law. In the meantime, Chinese AI labs are exploiting the loophole and training their models on foreign data centers that have acquired Nvidia chips.
The U.S. and China have been locked in an AI arms race, with the latter’s open-weight models behind frontier U.S. models by just a matter of months . More importantly, these models, like Kimi K3, are significantly easier to run and could even be much cheaper to use . These advancements happened despite Washington’s efforts to make it harder for Chinese AI labs to acquire hardware that uses American technologies through export bans. But even though President Donald Trump eventually made a 180-degree turn and allowed Nvidia H200 chips for export to China , Beijing has taken steps to protect its domestic AI chip industry by blocking H200 imports at the border, meaning these companies must look elsewhere if they want to access the latest, most powerful AI GPUs to train their models.
Moonshot also has a limited window if it wants to secure compute capacity outside of China — Nvidia has reportedly told its data center clients to limit their business and avoid serving more than 49% of their capacity to Chinese outfits . This could be a harbinger of things to come, especially as reports like these could spotlight the bipartisan bill and increase pressure to advance the legislation sooner rather than later.
Follow Tom's Hardware on Google News , or add us as a preferred source , to get our latest news, analysis, & reviews in your feeds.
Chinese AI companies buy 1 million Nvidia GPUs in a single year, report claims
Key considerations
- Investor positioning can change fast
- Volatility remains possible near catalysts
- Macro rates and liquidity can dominate flows
Reference reading
- https://www.tomshardware.com/tech-industry/policy/SPONSORED_LINK_URL
- https://www.tomshardware.com/tech-industry/policy/chinese-ai-lab-moonshot-reportedly-looked-at-european-data-centers-to-access-nvidia-chips-report-claims-kimi-maker-primarily-interested-in-inferencing-capacity-but-also-asked-about-training-clusters#main
- https://www.tomshardware.com/membership
- UPERFECT GR14BU portable monitor review: A 120 Hz 3K OLED touch display at a killer price
- Vera Delivers the Best of Both Worlds: Highest Throughput and Fastest Agents
- AI Is Changing How Semiconductor Fabs Work — and Engineers Are Moving Faster Because of It
- Gelsinger-backed Nagare Labs’ Z2RAM promises SRAM speed at DRAM density
- Bambu Lab H3C 3D printer review: A nine-tool monster
Informational only. No financial advice. Do your own research.