The news: China’s local governments, state-backed firms, and real estate speculators launched over 500 AI data centers in 2023 and 2024, with over 150 now operational. But reports indicate up to 80% of these newly built computing resources remain unused, per the MIT Technology Review. 

This market collapse stems from hasty construction, unexpected AI shifts from compute to inference, and weakening market demand. The misfire serves as a cautionary tale for US hyperscalers rushing to meet AI demand. 

Why China’s data center push failed: Many centers were built in remote areas for cheaper power, but they’re too far from AI hubs where low-latency inference matters—making them inefficient despite lower energy costs.

Most of the projects were driven by a mix of hype and politics. Local officials prioritized short-term, high-profile AI projects to boost political careers amid real estate and tech industry slowdowns.

• Builders lacked AI experience and data centers were poorly optimized for real-world AI workloads. 
• The MIT Technology Review found that some operators never intended to use facilities for AI and instead exploited them for subsidized electricity or state-backed loans.

GPU rental prices have plummeted—Nvidia H100 servers now rent for RMB 75,000 ($10,407) monthly, down from earlier RMB 180,000 ($24,978) peaks, reflecting cooling demand.

DeepSeek changed the game: DeepSeek’s efficient open-source models shifted AI economics. It achieved performance comparable to OpenAI’s latest models with older hardware and presumably at just 3% to 5% of the total cost.

DeepSeek’s example, and the subsequent rise of copycats, rendered many data centers technologically obsolete, as they were optimized for large-scale training workloads rather than low-latency inference needs.

Key takeaway: China’s data center bust reveals that building AI infrastructure without aligning it to real-world demand, workloads, and location kills value. US companies should exercise caution and anticipate industry shifts, which could be difficult as AI adoption and investments continue to rise.

This content is part of EMARKETER’s subscription Briefings, where we pair daily updates with data and analysis from forecasts and research reports. Our Briefings prepare you to start your day informed, to provide critical insights in an important meeting, and to understand the context of what’s happening in your industry. Non-clients can click here to get a demo of our full platform and coverage.