Ulanqab Computing Power Capacity Commitments Rise to 12.5GW, with AI Demand and Green Power Synergies Driving Cluster Expansion

Overview

The report focuses on the data center cluster in Ulanqab, Inner Mongolia, assessing its capacity expansion, major participants, actual demand, power and water consumption, green power supply, and changes in industry business models. Benefiting from low temperatures, proximity to Beijing, low network latency, low electricity tariffs, and abundant wind and solar resources, Ulanqab has become one of the fastest-growing AI computing power hubs in China and Asia-Pacific. As of June 2026, local data center operating and planned capacity commitments totaled 12.5GW, indicating strong investment appetite, but the large supply pipeline also brings utilization, rental, and project execution risks.

Core views

The report’s core views include: first, China’s AI demand and capital expenditure still have upside, and Ulanqab is expected to become an important incremental demand hosting location; second, third-party operators such as GDS, VNET, Chindata, Envision Group, ZDATA, and Centrin Data are the main expansion forces, with large campuses and power management capabilities forming competitive advantages; third, operators are shifting from pure rack leasing toward data center operations and green power asset management, while exploring computing power service revenue-sharing models; fourth, the 12.5GW of planned capacity is equivalent to 45% of China’s roughly 28GW of operating supply in 2025, and the supply surge may limit significant rental increases; fifth, green power mainly serves policy compliance and carbon reduction purposes and is not necessarily cheaper than conventional grid power.

Analysis framework

The report combines company disclosures, local government materials, data from Ulanqab’s development and reform authorities, the Ministry of Industry and Information Technology, the China Academy of Information and Communications Technology, and real estate consulting institutions to cross-compare capacity commitments, operating capacity, utilization rates, computing power, electricity consumption, regional costs, and green power supply. It also uses the phased commissioning and utilization ramp-up of VNET’s campus as a demand validation case, and compares Ulanqab’s capacity with Asia-Pacific data center hubs such as Johor.

Methodology notes

• Industry Supply-Demand AnalysisCapacity Commitment and Actual Utilization Analysis

  Observe operating, under-construction, planned, and utilized capacity simultaneously to distinguish investment commitments from real demand.

  The report analyzes the 12.5GW capacity commitment together with about 1.2GW of operating capacity, about 800MW of utilized capacity, and an overall utilization rate of 66%, avoiding judgments on industry momentum based solely on planned scale.

• Project Ramp-Up AnalysisPhased Commissioning Utilization Analysis

  Observe the pace of utilization improvement after data center commissioning by delivery batch.

  VNET’s project in Ulanqab is fully committed and pre-committed, with average utilization rising to above 40% and above 80% two and four quarters after capacity delivery, respectively, providing case support for local demand strength.

• Regional ComparisonAsia-Pacific Computing Power Hub Capacity Comparison

  Compare different computing power hubs by capacity scale, power costs, climate, latency, and energy mix.

  Ulanqab’s capacity commitments are about three times Johor’s total operating, under-construction, and planned capacity, and it has relative advantages in electricity tariffs, natural cooling, and proximity to Beijing.

• Resource and Cost AnalysisComputing Power and Electricity Coordination Analysis

  Incorporate data center load, green power procurement, grid dispatch, energy storage, and compliance requirements into a unified operating framework.

  The report believes data center operators will gradually take on green power asset management functions, controlling electricity costs and meeting green power ratio and PUE requirements through direct supply, green power trading, microgrids, and smart dispatch.

Asset mapping & comparison

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

• Ulanqab data center industry cluster

  Core incremental hosting location for AI computing power demand

  Strengths

  Large capacity reserves, relatively low electricity tariffs, average annual temperature of about 4.3 degrees Celsius, about 240 kilometers from Beijing, latency to Beijing of about 4.2 milliseconds, and abundant wind and solar resources.

  Weaknesses

  Rapid supply expansion may put phased pressure on overall utilization, and water resources are relatively tight.

  Comparison

  Capacity commitments are about three times Johor’s total operating, under-construction, and planned capacity, and data center end-user electricity tariffs are significantly lower than in Beijing.

  Risks

  Project delays, insufficient realization of customer demand, rental pressure, rising green power costs, and water resource constraints.

• Leading third-party data center operators

  Main investment and operating entities for local planned capacity

  Strengths

  Campuses are typically larger than self-built projects by internet and AI companies, with advantages in capex, power efficiency, customer delivery, and core hub resources.

  Weaknesses

  High upfront capital investment and sensitivity to financing costs, rack-up progress, and utilization ramp-up.

  Comparison

  Compared with self-built data centers, third-party operators have a higher share of planned capacity and larger campus reserves in Ulanqab.

  Risks

  Intensified supply competition may limit rental increases, and if pre-commitments do not convert into actual usage, the return cycle may lengthen.

• Green power and power infrastructure

  Key support for data centers to meet green power ratio and PUE requirements

  Strengths

  Ulanqab has more than 20GW of installed renewable energy capacity, with green power accounting for 67%, and is promoting integrated source-grid-load-storage projects.

  Weaknesses

  Green power is not necessarily cheaper than conventional grid electricity, and direct supply also requires investment in dedicated lines, energy storage, and backup power.

  Comparison

  Direct supply can enhance energy control capabilities, while green power trading and certificate models reduce capex for dedicated lines but entail transaction, certificate, and grid fees.

  Risks

  Power generation or purchase agreement prices, green certificate premiums, transmission and distribution fees, system fees, energy storage investment, and power market volatility.

• 600673.SS

  The report states that its owned Chindata has planned gigawatt-scale data center campuses in Ulanqab, providing indirect industry exposure.

  Strengths

  The project involves large-scale computing power campuses and integrated source-grid-load-storage green power connections.

  Weaknesses

  The report does not cover this stock and does not provide earnings forecasts, valuation, or target price.

  Comparison

  Chindata participates in large-scale local capacity construction alongside GDS, VNET, Envision Group, and others.

  Risks

  Project implementation and capex risks, oversupply, utilization falling short of expectations, and lack of stock-level valuation support in this report.

Key data

• Ulanqab capacity commitments12.5GWAs of June 2026, including operating and planned projects; 3.3GW in July 2025.
• Ulanqab operating capacityabout 1.2GWAs of 2025, accounting for about 4% of China’s operating data center capacity.
• Ulanqab utilized capacityabout 800MWAs of end-2025, up more than 70% year over year.
• Overall utilization rate66%As of 2025, declining due to the rapid increase in operating supply.
• Available computing power165 exaFLOPSAs of June 2026, more than doubled year over year and accounting for about 7% of China’s intelligent computing power.
• VNET Ulanqab utilized capacityabout 300MWAs of the first quarter of 2026, the report estimates it accounted for about one-third of the local market.
• VNET Ulanqab capacity pipeline398MW operating, 388MW under construction, about 1GW long-term reserveAs of the first quarter of 2026.
• Data center electricity consumption in H1 20263.3TWhUp 90% year over year, accounting for 7% of Ulanqab’s citywide electricity consumption.
• Data center electricity consumption in June 2026649GWhUp 75% year over year.
• Data center end-user electricity tariffRMB0.33 to RMB0.36 per kWhLower than about RMB0.60 to RMB0.70 in Beijing and about RMB0.37 to RMB0.42 in Zhangjiakou.
• Renewable energy installed capacityover 20GWExpected to reach 25GW in 2026, with green power accounting for 67% of the city’s electricity mix.
• Forecast operating capacity of China data centers28GW in 2025, expected 53GW in 2030The report expects it to nearly double by 2030.

Impact & implications

Ulanqab’s expansion benefits third-party data center operators with large-scale campus reserves, financing capacity, customer acquisition capabilities, and energy management technologies, and will also drive demand for wind power, photovoltaics, energy storage, transmission and distribution, power semiconductors, energy routers, solid-state transformers, and liquid cooling equipment. Business models may further extend from rack leasing to green power asset management and computing power service revenue sharing. However, large-scale new supply means industry value growth may not directly translate into higher rents, and investors should place greater emphasis on customer pre-commitments, utilization ramp-up, unit power costs, and returns on capital.

Risks

• The 12.5GW capacity commitment is very large, and if AI demand growth falls short of expectations, it may create phased oversupply.
• Data center rental upside may be limited by a surge in supply from Ulanqab and other western computing power hubs.
• Committed or planned capacity does not equal actual delivery, and project construction, grid connection, financing, and customer onboarding may be delayed.
• Overall utilization may decline during the phase of rapid supply commissioning, affecting returns on capital and cash flow.
• Green power is not necessarily cheaper; power purchase agreements, green certificates, transmission and distribution, energy storage, dedicated lines, and backup power costs may erode economics.
• Ulanqab’s water resources are relatively tight, which may constrain cooling system solutions and high-density computing power deployment.
• Tightening policy requirements for green power ratio and PUE may increase construction and operating costs.
• 600673.SS is not covered in the report, and no stock rating or target price can be derived from it.

What to watch

• The progress of converting 12.5GW of capacity commitments into actual construction starts, delivery, and operation.
• Ulanqab’s utilized capacity, overall utilization rate, and rack-up speed for new projects.
• Customer commitments, pre-commitments, and utilization ramp-up two to four quarters after commissioning for operators such as VNET.
• Whether China’s AI capital expenditure and data center demand growth before 2030 match supply expansion.
• Changes in data center rents, unit rack revenue, and returns on capital.
• Local monthly data center electricity consumption and its share of citywide electricity consumption.
• Progress in renewable energy installed capacity, direct supply projects, green power trading prices, and energy storage construction.
• Green power coverage, PUE compliance, and water resource constraints for new data centers.
• The actual earnings contribution from operators’ transition from rack leasing to green power asset management and computing power service revenue-sharing models.