Quantitative Private Equity Firms Race to Capture Market Opportunities in Ulanqab - 36Kr

Quantitative private equity firms are relocating and expanding computing infrastructure to Ulanqab, a city in central Inner Mongolia. According to a Goldman Sachs report cited in the article, Ulanqab has become one of the largest and fastest-growing AI computing clusters in the Asia-Pacific region, hosting data centers for Alibaba, ByteDance and Apple. Bloomberg reported that DeepSeek plans an ultra-large AI data center of about 1GW there, deploying at least 160,000 Huawei Ascend 950DT chips. Alongside DeepSeek/High-Flyer, quantitative firms Jiukun and Lingjun have established operations in the city. Lingjun, one of the first private equity firms to enter Inner Mongolia, began building its Ulanqab supercomputing center in 2020; it now has more than 100,000 CPU cores, over 1,000 GPUs and over 10PB of storage, and told Yuancuan that commercial electricity and operating costs there are far lower than in Beijing.

The article reports that the sector's scale has expanded sharply since June, with six quantitative private equity firms now managing over 80 billion yuan and 13 managing over 50 billion yuan. Computing build-outs have become an arms race, and GPU costs are a major pressure point. A quantitative fund manager told Yuancuan that graphics card prices have risen about threefold since last year, and that prices for high-end switches, CPUs and memory have roughly doubled. Firms without established GPU inventories have turned to resellers, dismantled complete machines from manufacturers such as Dell and Lenovo, and in some cases sourced hardware from Hong Kong. Some newer entrants are reportedly stockpiling GPUs as fixed assets and renting out surplus capacity, with monthly rents for an 8-A100 machine reported to have risen from 30,000-40,000 yuan to 80,000 yuan this year. The article notes that quantitative excess returns have pulled back since July, which has increased the strategic importance of computing capacity.

Cost reduction is driving siting decisions. Established firms face high migration costs and constrained capacity and energy-consumption approvals in Shanghai, while Ulanqab offers wind and thermal power, a mean temperature of 4.3 degrees Celsius that allows natural cooling for nearly 10 months a year, and, according to one fund manager, direct negotiated electricity prices below 0.1 yuan per kWh. Envision Energy's Xinghe Base in Ulanqab reportedly runs at 80% green power, with self-supplied wind power more than 20% cheaper than the local tariff, and all its air-cooled computer rooms were reported as fully reserved as of October 2026. Building self-owned facilities there involves land acquisition hurdles, approval requirements for capacity above 7.5 megawatts, and construction periods generally exceeding 18 months, so leasing is viewed as the more practical option for most firms. The article concludes that smaller firms may need to maximize utilization of limited hardware, citing a small firm that deployed large language models locally on low-cost hardware.