China’s tech landscape is gearing up for a major transformation as its AI companies aim to establish a massive “hyperscale” facility equipped with over 100,000 NVIDIA AI chips. This ambitious project involves constructing 36 data centers powered by NVIDIA’s H100 and H200 AI chips.
As the global race to lead in AI infrastructure intensifies, China is making bold moves to advance, despite facing trade restrictions. A Bloomberg report highlights that, with state backing, Chinese AI firms are set to build a colossal AI facility in Yiwu, China. The plan involves installing upwards of 115,000 high-end NVIDIA AI chips in these centers spread across the country’s expansive western deserts. However, a crucial question remains: How will China source these chips amidst US restrictions?
This development represents one of China’s most significant AI data center initiatives to date, showcasing Beijing’s determination to bolster its AI capabilities. Although not yet on par with US infrastructure, China’s efforts indicate significant progress. The confidential nature of this project is underscored by US sources, who appear largely unaware of its specifics.
With direct access to US chips restricted, China might leverage alternative routes through Southeast Asian countries like Malaysia and Singapore, where Chinese engineers reportedly have some access to AI technology. Moreover, Chinese tech giants possess a robust supply of H20 AI accelerators. While these resources could support the buildout, a consistent supply of NVIDIA chips remains elusive, lacking any direct provision from NVIDIA at present.
Despite the challenges and uncertainty about acquiring the necessary chips, construction is underway in Yiwu. China’s data center sector is expanding rapidly, projected to reach a market size of 300 billion yuan this year. Notably, the country continues to explore creative solutions to bypass limitations, hinting that current US export controls might not fully impede China’s technological ambitions.
