SemiAnalysis released a model tracking China’s datacenter infrastructure on September 25, the research firm announced. The bottom-up model, built from building-level data on more than 1,000 facilities across over 60 operators, aims to resolve a measurement problem: published estimates of China’s datacenter capacity have differed by as much as 15 times.
China’s total delivered fleet exceeds 24 gigawatts, according to the model—larger than the EMEA region. ByteDance, which remains private, leases roughly one-fifth of that capacity, making it the single most important tenant for wholesale colocation providers in the country.
The figures on capacity, tenant allocations, and delivery curves come from SemiAnalysis’s proprietary model and have not been independently verified.
The model arrives as hyperscaler spending in China accelerates sharply. Combined capital expenditure at Alibaba, Tencent, and Baidu reached $20 billion in the second quarter of 2026, more than doubling year-on-year, SemiAnalysis said. For the first time on record, all three posted negative free cash flow in the quarter.
High vacancy rates have long confounded outside observers, but the report argues those numbers reflect a legacy retail market that cannot serve AI workloads. Wholesale buildings are filling fast, while many older, low-density retail racks sit idle. GDS and VNET, the only two Chinese datacenter landlords listed in the US, signed 1.3 GW of wholesale orders in the first half of 2026 but captured barely a third of ByteDance and Alibaba orders between 2024 and mid-2026, according to the model.
The release includes quarterly delivery curves through 2032, tenant identification per building, lease signing and delivery trackers, and analytics for the government-backed “Eastern Data, Western Compute” hubs. SemiAnalysis said it plans to add AI capacity and power analysis in subsequent updates.