China Constructs 'Stargate' Counterpart to Compete in AI Race

China is mobilizing all its resources to build its own artificial intelligence (AI) data center in response to the U.S.’s ‘Stargate’ project...

China is mobilizing all its resources to build its own artificial intelligence (AI) data center in response to the U.S.’s ‘Stargate’ project. A data center, which houses tens of thousands of graphics processing units (GPUs), acts as the ‘heart’ that stores and transmits data to enable the ‘brain’ of AI models to function. As the U.S. pushes forward with the ‘Stargate’ project—aiming to construct the world’s largest AI data center complex by 2029 with Oracle, OpenAI, and Japan’s SoftBank investing $500 billion (approximately 700 trillion Korean won)—China has begun catching up in AI infrastructure by building a large-scale data center in an agricultural city. According to Bank of America (BoA), China’s AI capital expenditure (CAPEX) is expected to reach $98 billion this year.

◇Creating a Chinese Version of Stargate

The Financial Times (FT) reported on the 21st (local time) that a large-scale data center is under construction on an island farmland of approximately 3 million square meters in Wuhu City, an agricultural city in Anhui Province, China. A source from a company participating in the project stated, “This building is part of efforts to construct a Chinese version of Stargate.” The data center is located near the Yangtze River Delta economic zone, which includes Shanghai, Nanjing, and Hangzhou—key hubs of the Chinese economy. The data center being built here is expected to handle AI inference demands in the region. Four new AI data centers operated by Huawei, China Telecom, China Unicom, and China Mobile have been established, with $37 billion invested so far. The FT noted, “Local governments are subsidizing 30% of the cost of AI semiconductors for data centers,” adding, “This is part of a broader plan by the Chinese government to effectively handle surging AI demand.”

China’s large-scale expansion of data centers stems from the significant gap in AI computing capabilities compared to the U.S. According to the U.S. non-profit research organization Epoch AI, as of May of this year, the U.S. accounts for approximately 75% of the world’s GPU computing capabilities. China, in second place, holds a 15% share, lagging far behind the U.S. Despite demonstrating AI capabilities earlier this year with the ‘DeepSeek shock,’ China still faces difficulties in securing the GPUs needed to train and infer AI models. While U.S. big tech companies like Meta and Google have secured NVIDIA’s latest semiconductors, China cannot obtain advanced AI semiconductors due to U.S. export controls.

Since 2022, China has pursued the ‘East Data West Compute’ (Dong Shu Xi Suan) project, concentrating data centers in remote western regions. The initiative aims to process data from the economically developed eastern regions in the power-rich western areas. The goal is to build 10 large-scale data center hubs across eight regions. By August of last year, $6.12 billion had been invested in this plan. However, numerous data centers built in the sparsely populated western regions have been criticized for being ‘idle’ due to insufficient demand. To address this, the Chinese government plans to connect data centers nationwide and allocate surplus computing resources to where they are needed.

◇Companies Rush to Invest

Chinese companies are also making large-scale investments to build their own data centers. Alibaba announced it would invest 380 billion yuan (approximately 74.4 trillion Korean won) in AI infrastructure over the next three years. Tencent, which stated it would invest about 10% of this year’s revenue in AI infrastructure, decided on the 17th to issue corporate bonds worth a total of 9 billion yuan. Chinese media pointed out that funds will be concentrated on expanding AI infrastructure. A tech industry insider said, “AI supremacy ultimately depends on who secures more computing resources,” adding, “While the U.S. is preparing the world’s largest AI computing hub through the Stargate project, even overinvestment shows the determination not to fall behind in the AI race.”