Ulanqab is rapidly becoming a leading AI computing power cluster in Asia-Pacific

Overview

The report focuses on the data center industry cluster in Ulanqab, Inner Mongolia, analyzing its capacity expansion, key participants, actual demand, utilization, power and water resources, green power procurement models, and industry implications. Leveraging low temperatures, proximity to Beijing, relatively low electricity tariffs, and abundant wind and solar resources, Ulanqab has become one of the fastest-growing AI computing power hubs in China and even Asia-Pacific. The report believes China’s AI demand and capital expenditure still have upside potential, but the local planned supply scale is very large, and industry value creation will depend more on customer lock-in, commissioning pace, utilization improvement, and energy management capability rather than simply relying on rental increases.

Core views

First, Ulanqab’s 12.5GW capacity commitment shows that AI infrastructure investment is rapidly concentrating in core computing power hubs with low costs and abundant green energy. Second, demand is not supported solely by long-term planning: utilized capacity in 2025 increased by more than 70% year on year, operating computing power reached 165 EFLOPS in June 2026, and VNET’s project also showed a relatively fast utilization ramp-up. Third, leading third-party data centers, with larger campus scale, capital deployment capability, and power efficiency, remain key participants in absorbing AI demand. Fourth, the operating model is evolving from cabinet leasing to data center operations, green power asset management, and computing power or Token revenue sharing. Fifth, the large supply reserve may weaken rental pricing power, and project returns will be jointly affected by utilization, all-in green power costs, energy storage and dedicated line investment, and backup power requirements.

Analysis framework

The report combines bottom-up project mapping with top-down industry supply-demand analysis: it tallies Ulanqab’s operating, under-construction, and planned capacity; identifies major operators and internet or AI companies; compares local capacity with China’s total supply and Asia-Pacific hubs such as Johor; uses utilized capacity, computing power, electricity consumption, and VNET’s phased commissioning projects to assess demand and ramp-up speed; and breaks down the impact of green power procurement, transmission and distribution, energy storage, dedicated lines, and backup power on project economics.

Methodology notes

• Supply-demand analysisData center capacity and utilization framework

  Examine operating, under-construction, planned, and utilized capacity simultaneously to avoid judging real demand solely from project announcements.

  The report combines the 12.5GW capacity commitment with approximately 1.2GW of operating capacity, around 800MW of utilized capacity, and 66% overall utilization, and notes that rapid new supply will temporarily dilute utilization.

• Regional comparisonAsia-Pacific computing power hub benchmark comparison

  Compare different data center hubs by capacity scale, latency, electricity tariffs, climate, and green power conditions.

  Ulanqab’s total capacity commitment is about three times Johor’s future data center capacity, and it has advantages including proximity to Beijing, low latency, low temperatures, and relatively low electricity tariffs.

• Operational analysisPhased commissioning utilization ramp-up analysis

  Observe changes in data center capacity utilization across quarters after delivery by commissioning batch.

  The average utilization of VNET’s Ulanqab project increased to over 40% and over 80% around two quarters and four quarters after capacity delivery, respectively, which is used to validate conversion of customer commitments and the pace of demand fulfillment.

• Cost analysisGreen power comprehensive economics framework

  Green power costs depend not only on generation prices, but also include certificate, transmission and distribution, energy storage, dedicated line, and backup power costs.

  The report compares two main models, green power direct supply and green power trading plus certificates, and emphasizes that green power mainly satisfies policy and energy-efficiency requirements and is not necessarily inherently cheaper than conventional grid electricity tariffs.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• VNET

  Early mover in Ulanqab and potential major beneficiary

  Strengths

  As of the first quarter of 2026, it had 398MW operating, 388MW under construction, and approximately 1GW of future development reserve; around 300MW of utilized capacity and relatively fast post-commissioning utilization ramp-up indicate strong customer acquisition capability.

  Weaknesses

  Large-scale capacity under construction and in reserve brings ongoing capex, construction execution, and utilization realization pressure.

  Comparison

  Compared with other local projects, VNET has disclosed more complete data on operating, under-construction, reserve capacity, and utilization, making it a representative case for validating demand strength in Ulanqab.

  Risks

  Rapid industry supply growth, pressure on rents, customer concentration, energy cost fluctuations, and slower-than-expected ramp-up of new capacity.

• GDS

  Gigawatt-scale third-party data center investor and potential beneficiary

  Strengths

  Plans to build an approximately 2GW data center campus, with investment exceeding Rmb30bn, and aims to achieve over 80% green power coverage through direct supply and trading.

  Weaknesses

  The project is large in scale, and capital intensity and the quality of customer pre-commitments have a significant impact on investment returns.

  Comparison

  Compared with internet companies’ self-built campuses, third-party operators such as GDS generally have larger campuses and more capacity reserves.

  Risks

  Construction delays, demand conversion slower than supply rollout, financing costs, and all-in green power costs exceeding expectations.

• Chindata (Dongyangguang, 600673.SS, not covered)

  Participant in a gigawatt-scale Ulanqab project

  Strengths

  Plans a gigawatt-scale campus and adopts an integrated source-grid-load-storage green power direct supply model.

  Weaknesses

  The report does not provide detailed commissioning schedules, customer commitments, or utilization data for the project.

  Comparison

  Compared with VNET, there is less verifiable information on operating capacity and utilization, so project value depends more on subsequent execution.

  Risks

  Project approval, construction progress, capital investment, customer implementation, and green power supporting execution risks.

• Envision Group

  Participant in coordinated data center and green energy deployment

  Strengths

  Planned capacity is up to around 2GW, combined with its own wind farms and green power direct supply, targeting over 80% direct green power supply.

  Weaknesses

  The report does not disclose specific investment amounts, commissioning plans, or customer mix.

  Comparison

  Its energy asset synergy capability is relatively prominent, reflecting the trend of data center operations extending into green power asset management.

  Risks

  Uncertainty exists in planned capacity realization, energy infrastructure construction, customer demand, and project returns.

• China’s leading third-party data center operators

  Core absorbers of AI demand growth

  Strengths

  They have capital deployment capability, power efficiency, customer service capability, and capacity reserves in core computing power hubs, with planned capacity generally larger than internet companies’ self-built campuses.

  Weaknesses

  The business is capital-intensive, and cash flow and returns are susceptible to utilization, financing costs, and electricity tariff changes.

  Comparison

  Compared with enterprise self-built data centers, third-party operators cover more customers and have larger campuses, but also bear higher vacancy and price competition risks.

  Risks

  Oversupply, limited rental growth, substitution by customer self-built data centers, technology iteration, and prolonged capex payback periods.

• Green power, energy storage, and power equipment value chain

  Indirect beneficiary assets from data center expansion

  Strengths

  Green power share and PUE compliance requirements drive demand for direct supply, trading, microgrids, energy storage, power semiconductors, energy routers, and solid-state transformers.

  Weaknesses

  Green power is not necessarily low-cost; certificates, transmission and distribution, dedicated lines, energy storage, and backup power will increase all-in costs.

  Comparison

  Compared with traditional grid power supply, green power solutions have stronger compliance, but economics depend on project structure and supporting investment.

  Risks

  Policy changes, power market price volatility, insufficient energy storage utilization, grid connection restrictions, and additional backup capacity costs.

Key data

• Ulanqab capacity commitment12.5GWAs of June 2026, including operating and planned capacity; significantly higher than 3.3GW in July 2025.
• Ulanqab 2025 operating capacityApproximately 1.2GWAbout 4% of China’s operating data center capacity in 2025.
• Ulanqab 2025 utilized capacityApproximately 800MWUp more than 70% year on year, with overall utilization of about 66%.
• Ulanqab operating computing power165 EFLOPSAs of June 2026, more than doubled year on year, accounting for about 7% of China’s intelligent computing power.
• VNET Ulanqab capacity398MW operating, 388MW under construction, approximately 1GW reserveAs of the first quarter of 2026; estimated utilized capacity of around 300MW, about one-third of the local market.
• VNET utilization ramp-upOver 40% after two quarters, over 80% after four quartersBased on phased commissioning performance at the Ulanqab campus, with project commitment and pre-commitment rates at 100%.
• China data center operating capacity forecast28GW in 2025; expected 53GW in 2030The report expects China’s operating capacity to nearly double by 2030.
• Data center electricity consumption3.3TWh in the first half of 2026Up 90% year on year, accounting for about 7% of Ulanqab’s citywide electricity consumption.
• Monthly data center electricity consumption in June 2026649GWhUp 75% year on year.
• Data center end-user electricity tariffRmb0.33-0.36/kWhLower than approximately Rmb0.6-0.7/kWh in Beijing and approximately Rmb0.37-0.42/kWh in Zhangjiakou.
• Renewable energy installed capacityOver 20GW, expected 25GW in 2026Green power accounts for about 67% of Ulanqab’s power mix.
• Source-grid-load-storage projects4 projects, 1.6GW of green power capacityAnnual power generation is expected to be 4.5TWh after full completion.

Impact & implications

The rise of Ulanqab reinforces the trend of China’s AI computing power migrating toward regions with lower energy costs and abundant renewable resources. Leading third-party data center operators are expected to benefit from campus scale, financing capability, customer resources, and energy efficiency, and gradually take on asset management functions such as green power procurement, microgrids, energy storage, and power trading. At the same time, the 12.5GW planned supply may create phased supply pressure, limiting rental growth, and investment returns will depend more on customer pre-commitments, commissioning discipline, and rapid utilization ramp-up. Power semiconductors, energy routers, solid-state transformers, energy storage, liquid cooling, and other segments may see incremental demand, but water scarcity will affect cooling technology routes and project costs.

Risks

• Ulanqab’s 12.5GW planned capacity is far above current demand, and new supply may cause utilization declines and price competition.
• Data center rental growth may be constrained by the rapid expansion of China’s overall supply.
• Large-scale projects have high capex, and construction progress, financing costs, and customer pre-commitment conversion rates may affect returns.
• The costs of green power certificates, transmission and distribution, dedicated lines, energy storage, and backup power may weaken the advantage of low electricity tariffs.
• Ulanqab has relatively tight water resources, which may limit some cooling solutions and increase the design and operating difficulty of liquid cooling systems.
• If AI demand or capital expenditure falls short of expectations, it may lengthen the utilization ramp-up cycle for new capacity.
• New business models such as Token revenue sharing carry risks related to revenue volatility, customer credit, and immature pricing mechanisms.
• Customer self-built data centers may substitute for some third-party colocation demand.

What to watch

• The progress of converting the 12.5GW capacity commitment into actual construction starts, commissioning, and customer contracts.
• Subsequent growth in Ulanqab’s utilized capacity, overall utilization, and 165 EFLOPS of computing power.
• Construction pace, pre-commitment rates, and utilization ramp-up of VNET, GDS, Chindata, and Envision Group projects.
• The supply-demand gap during the growth of China’s operating data center capacity from 28GW in 2025 to an expected 53GW in 2030.
• Changes in local data center rents, electricity tariffs per kWh, and AI computing power service prices.
• The all-in costs of green power direct supply, trading, and certificates, as well as implementation of the requirement for over 80% green power coverage.
• The impact of requirements such as PUE below 1.25 for new data centers and below 1.2 for national hubs on equipment and energy solutions.
• Adoption progress of direct-to-chip liquid cooling and other water-saving cooling technologies under water resource constraints.
• Profitability performance as third-party data centers transition from cabinet leasing to green power asset management and computing power revenue-sharing models.