7 min read
7 min read

China is pushing ahead with an ambitious plan to build between 36 and 39 AI-focused data centers in its western regions.
These facilities will be powered by more than 115,000 high-performance NVIDIA GPUs, many of which are subject to US export restrictions.
This bold move signals China’s determination to advance its AI infrastructure, even as Washington tries to slow it down. Whether viewed as strategic innovation or geopolitical defiance, the scale of this buildout is undeniably massive.

Reports suggest that a large portion of the GPUs, possibly the majority, will be concentrated in a single data center in Yiwu, Xinjiang Province.
This state-owned site is strategically located to capitalize on abundant solar, wind, and coal energy and cool high-altitude air to manage heat.
Choosing Yiwu as the hub reflects China’s focus on optimizing energy resources and operational efficiency while supporting an enormous computing load.

China’s plan revolves around NVIDIA’s cutting-edge H100 and H200 GPUs, which are officially banned from exporting to Chinese buyers.
These high-end chips are critical for AI workloads like training large language models and running large-scale inference operations.
In defiance of US restrictions, China’s use of these chips shows how indispensable these processors remain, even for countries developing their own AI ecosystems.

While US bans are designed to block direct sales, reports suggest China could be obtaining NVIDIA chips via third-party suppliers or black market channels.
Some speculate Southeast Asian intermediaries might be involved. However, US officials remain skeptical that such unofficial routes could supply China with over 115,000 GPUs without detection.
Still, China’s unwavering confidence in the project suggests they anticipate filling these data centers by whatever means necessary.

Even without clear evidence of how the chips will be sourced, Chinese tech firms and government entities are pouring billions into the infrastructure. Construction is visibly underway, particularly around Yiwu.
It’s a calculated risk: building large-scale AI data centers assumes future access to the GPUs needed. This indicates China’s belief that, whether through procurement or domestic production, it will have the hardware soon enough.

Xinjiang and Qinghai offer China vast solar and wind energy reserves, making them ideal locations for power-hungry facilities.
Coupled with cooler temperatures in these high-altitude regions, this setup helps reduce operating costs and manage thermal challenges associated with GPU clusters.
By leveraging its geography, China aims to power its AI ambitions sustainably, positioning itself as a future green tech leader.

The US initially restricted NVIDIA’s high-end chip sales to China in 2022, aiming to stunt Beijing’s AI advancement. However, China’s ambitious infrastructure plans highlight how ineffective those controls have become.
The scale of this project suggests that trade restrictions alone might not be sufficient to slow down China’s AI industry. This development could push US policymakers to reconsider their export control strategies.

With 115,000 high-performance NVIDIA GPUs planned, China’s new data centers could rival the compute capacity of the most powerful AI facilities worldwide.
For comparison, Elon Musk’s xAI reportedly used 100,000 H100 GPUs to train its Grok 3 AI model. This indicates that once operational, China’s forthcoming data centers will be capable of supporting top-tier AI models and applications.

This massive expansion isn’t just a private-sector effort. State-owned companies and prominent Chinese AI firms are working under government guidance to realize the vision.
This blend of corporate resources and national strategy suggests that China views AI infrastructure as an economic and strategic priority. The goal: develop AI capabilities that are independent of Western technology constraints.

While plans are moving, US investigators reportedly doubt China can source over 100,000 restricted GPUs. Estimates from US officials suggest only around 25,000 banned chips are currently in China, far short of the 115,000 needed.
This skepticism raises a critical question: Are Chinese firms bluffing, or is there an undiscovered supply network feeding this buildout?

If completed as planned, the Yiwu-based data center could be one of China’s most powerful AI hubs. Current documents suggest computing power equivalent to 24,000 petaflops is already operational in the region.
This is comparable to 12,000 NVIDIA H100s. Up to 80,500 GPUs will vastly exceed existing capabilities, potentially supporting a national AI infrastructure backbone.

As global tech giants upgrade to NVIDIA’s new Blackwell architecture, older H100 GPUs will be decommissioned and replaced. Industry watchers speculate that China could target these used chips for acquisition via unofficial channels.
While unproven, this scenario is plausible, given the high volumes of aging hardware circulating globally, a potential grey-market windfall for China’s AI expansion.

Although China invests heavily in local AI chip manufacturing, domestic alternatives like Huawei’s CloudMatrix 384 can’t yet compete with NVIDIA’s H100 and H200 GPUs.
This technological gap explains why China is still focused on sourcing foreign chips. Domestic manufacturing may fill some gaps in the future, but for now, NVIDIA’s CUDA-optimized hardware remains critical for China’s AI ambitions.

Without official access to NVIDIA’s hardware support and optimization services, China faces risks in running these large-scale data centers. NVIDIA has stated that operating clusters built from smuggled or unofficially sourced chips is risky and inefficient.
However, given geopolitical tensions, Chinese operators may accept lower performance and reliability in exchange for progress and independence from US-controlled technology.

One of the most striking details is that construction in places like Yiwu is progressing despite no public confirmation that China has secured the GPUs needed.
This shows how China is willing to “build first, solve later,” betting that chips will be sourced eventually. It’s a risky strategy that demonstrates national resolve in advancing AI leadership.

China’s colossal data center initiative underscores a larger reality: the AI race between the US and China is accelerating. While the US maintains an overall lead, China’s defiant investment in restricted technology shows it is determined to close the gap.
As this situation unfolds, technological innovation and geopolitical maneuvering will shape the future of global AI dominance.
What do you think about China’s bold move to enhance and fuel AI chips despite a ban from the US on Nvidia? Please share your thoughts and drop a comment.
Read More From This Brand:
Don’t forget to follow us for more exclusive content on MSN.
This slideshow was made with AI assistance and human editing.
This content is exclusive for our subscribers.
Get instant FREE access to ALL of our articles.
Dan Mitchell has been in the computer industry for more than 25 years, getting started with computers at age 7 on an Apple II.
We appreciate you taking the time to share your feedback about this page with us.
Whether it's praise for something good, or ideas to improve something that
isn't quite right, we're excited to hear from you.
Stay up to date on all the latest tech, computing and smarter living. 100% FREE
Unsubscribe at any time. We hate spam too, don't worry.

Lucky you! This thread is empty,
which means you've got dibs on the first comment.
Go for it!