AI Data Center Competition Shifts to Power: Envision's Green Direct-Supply Model Sets New Benchmark for GW-Scale Compute

Goldman Sachs recently released a report on China's data center industry, positioning Ulanqab as one of the largest and fastest-growing AI computing clusters in China and the Asia-Pacific region. According to the report, as of June 2026, the total commissioned and planned capacity commitments for data centers in Ulanqab reached approximately 12.5 GW, compared with roughly 1.2 GW of commissioned capacity in 2025 — implying a potential expansion of more than tenfold.

Among GW-scale project operators, Envision, GDS Holdings, and 21Vianet were ranked by Goldman Sachs as the leading tier. Unlike the latter two, which are traditional third-party data center operators, Envision is the only new energy technology company in this group. This distinct identity enables it to supply a high proportion of green electricity to computing centers, earning it special mention in the report.

Shortly before the report's release, on August 6, Envision announced the official commissioning of the 'Envision Ulanqab Xinghe Base,' launching an AI computing super monolith with a floor area of 120,000 square meters. The base has a total planned capacity of 2 GW.

Ulanqab's geographic conditions form the foundational support for the computing industry: low electricity prices, a cold climate, network latency of just 4.2 milliseconds to Beijing, and abundant wind and solar resources, with green power accounting for 67% across the region. Leveraging local wind energy, Envision's super monolith adopts a model of self-built wind farms plus dedicated-line direct supply to deliver a high proportion of green power, with a compute density target set at more than ten times that of traditional data centers.

Hu Xingzan, CTO of Envision AI Computing Center, attributes the improved compute density to three layers of synergy. First, high-proportion green power direct supply — relying on self-built wind farms, self-built transmission lines, and Envision's self-developed AI power system — provides stable and clean electricity for high-power-density AI racks, breaking through the power capacity bottleneck of traditional data centers. Second, the 120,000-square-meter single building offers centralized deployment space for AI clusters, reducing the space loss and interconnection latency associated with distributed layouts. Third, compact power equipment such as solid-state transformers and BESS (Battery Energy Storage Systems) optimize the power distribution architecture, compressing the non-computing footprint of traditional power distribution and backup facilities and freeing up more space for AI racks.

Public records show that in June 2026, Envision unveiled the 'Gobi Mission' at the VivaTech summit in France, planning to build 5 GW of green AI computing centers in Gobi desert regions worldwide by 2030. The Ulanqab Xinghe Base is the first project under this initiative.

Jiayida (Suzhou) Fluid Power Technology Co., Ltd. focuses on providing one-stop liquid cooling fluid system solutions and services for data center scenarios. The product portfolio covers OCP-standard UQD, UQDB and BMQC quick disconnect couplings, water manifold assemblies, Continental EPDM hoses and hose assemblies, FF9 full-flow double interlock quick disconnect ball valves as an alternative to FD83 series products, API heat exchangers, Gates data center-dedicated EPDM hoses, as well as WIKA pressure gauges and sensors from Germany.