- Deepseek is only one of the many reasons why China’s AI growth simply wasn’t realized
- Up to 80% of the new data center capacity has not been used according to local sources
- If this capacity hits the wider market it can cause a larger headache to data center developers
China’s AI -Infrastructure Boom Wobbles, as according to a report in My Technology ReviewThe country built hundreds of data centers to support its AI ambitions, but many are now unused.
Billions were invested by both state and private units by 2023 and 2024, with the expectation that the demand for GPU houses would continue to grow, but the recording has actually fallen off and as a result many operators are now fighting to survive.
Much of the early momentum was driven by hype. The government, who was eager for China to become a global leader in AI, called on local officials for quick-track data center construction, with the result that more than 500 projects were announced nationwide and at least 150 were completed by the end of 2024, according to state-related sources. However, My Technology Review Says local publications report that up to 80% of this new computer capacity remains inactive.
Sales of GPUs
Location is also a problem, My Technology Review Notes. Facilities built into central and western China, where electricity is cheap, now face to meet latency requirements. In cities like Zhengzhou, operators allegedly even give free computer coupons in an attempt to lure users.
In some regions, developers began selling GPUs not to secure long -term clients.
Xiao Li, a data center project manager who was talking to My Technology Reviewsaid that many WeChat groups that once boasted by Nvidia -chip offers have gone quiet. “It seems that everyone sells, but get the buyer,” he noted.
If this capacity hits the wider market, it can cause a larger headache to data center developers that flood an already soft sector with even more supply and push prices down further.
One of the reasons for the fall in demand is the increase in Deepseek, which emerged the global tech economy when it was launched in January 2025.
Its open source resonance model, R1, matches the performance of Chatgpt O1, but at a lower price that shifts interest away from model training and against inference -real -time use of AI models that require different infrastructures.
Many of the data centers built during urgency were designed for large -scale training, not the requirements for low latency in real time.
Despite the overshadowing, the Chinese government is allegedly engaged. Central authorities held an AI Symposium by early 2025, and companies such as Alibaba and Tiktok own bytance have announced large investments.
However, too many early data center investors, expectations have collapsed. The infrastructure was built, but the demand it was intended to deliver has simply not arrived.
