Summary of Key Points
Ulanqab is an extremely arid grassland pastoral area with per capita water resources only one-quarter of the national average, experiencing droughts nine out of ten years. However, due to its three major advantages—cold climate, low electricity costs, and abundant land—it has become a “computing power hub” sought after by tech giants such as Huawei, Alibaba, and Tencent, with 89 projects already established and a total investment exceeding 500 billion yuan. Data centers are heavy water consumers; for example, a large center’s annual water consumption is equivalent to the water usage of a county with a population of thirty to forty thousand people. This has led to conflicts with local herders over groundwater resources: two thousand wells have been shut down to reduce extraction, and the government has initiated measures to address illegal water use. Yet, the water consumption of data centers is poorly documented in their ESG (Environmental, Social, and Governance) reports, and the legal principle of “priority for residential water use” is not being fully enforced. This struggle for water resources between AI computing power and herders’ livelihoods requires transparent rules to be established.
Why Has the Arid Ulanqab Become a Target for Data Centers?
Data centers are essentially giant heat generators; tens of thousands of GPUs operating like steel furnaces must be cooled to prevent downtime. The most cost-effective cooling method is “evaporative cooling,” which requires two conditions: a cold climate and dry air (a cold environment reduces the number of cooling cycles, and dry air accelerates water evaporation). Ulanqab meets both criteria:
- Cold Climate for Energy Savings: The average annual temperature is 4.3°C, and the low wet-bulb temperature in summer allows the evaporative cooling system’s compressors to stop running for 70% of the year, cutting cooling energy costs by 30%.
- Low Electricity Costs: The electricity price per kilowatt-hour is between 0.25 and 0.35 yuan, half the cost in the east. The network latency to Beijing is only 4.2 milliseconds, making it suitable for providing computing power to first-tier cities.
- Abundant Land: With a land area of 54,500 square kilometers and a permanent population of less than 1.7 million (30 people per square kilometer), the cost of acquiring land is very low (a GW-level data center requires several thousand acres).
These natural advantages have turned Ulanqab into a “computing power paradise” for investors, attracting numerous data centers.
Do Data Centers Really Compete with Herders for Water Resources?
The answer is yes, and the conflict is already intense:
- Data Center Water Consumption: A large center using evaporative cooling can consume millions of liters of water per day, equivalent to the annual water usage of a county with thirty to forty thousand people.
- Depleted Local Water Resources: The per capita water resources in Ulanqab are only 484 cubic meters (national average is over 2000 cubic meters). Chayouzhongqi is a region with overexploited groundwater, and Shangdu County shut down 2008 wells in 2020, reducing annual water extraction by 20 million cubic meters.
- Confrontation Leading to Action: In 2026, Inner Mongolia launched a campaign to address illegal water use, with Ulanqab focusing on industrial enterprises’ unauthorized water withdrawals. The first cases of violations have been reported, highlighting the stark conflict between capital-driven data center development and local water shortages.
Is Data Center Water Consumption a “Business Secret”?
The specific amount of groundwater used by data centers is largely hidden in their ESG reports:
- Tencent: Reported total annual water consumption of over 12 million tons but did not specify the proportion used in arid areas.
- KwaiShou: Mentioned the construction of rainwater collection systems in Ulanqab but did not provide actual evaporative cooling water usage figures (which is negligible compared to overall water consumption).
- Alibaba: Its WUE (Water Use Efficiency) for evaporative cooling is 1.799 liters per kilowatt-hour (below the national standard), but no specific data for Ulanqab centers were provided.
- Huawei: As a private company, it does not release ESG reports, leaving its water consumption completely unknown.
Although companies are aware of the risks (Alibaba considers water pressure a chronic issue, and KwaiShou acknowledges that droughts increase costs), they prefer not to disclose their regional water usage publicly.
Why Is the Legal Principle of “Priority for Residential Water Use” Not Being Enforced?
The Water Law clearly states that water resources should first meet residential needs in arid areas, with ecological considerations also taken into account. However, the economic benefits brought by data centers—such as billions in investment and millions in tax revenue—outweigh those of herders’ livelihoods:
- A GW-level data center can generate significant economic value but provides limited income for herders.
- International experiences show that water restrictions in hot regions have led to protests; for example, 88% of voters in Monterey Park, California, supported a permanent ban on data centers (with more opposition than nuclear power plants).
If Ulanqab does not address this disparity between legal priorities and reality, similar conflicts are inevitable.
The Way Forward: Balancing Development and Conservation
AI infrastructure is the future trend, and Ulanqab cannot be stopped. However, rules must be established:
1. Transparency in Water Usage: Each data center should publicly disclose the type of water used (groundwater/surface water), annual consumption, and peak daily usage, so herders can understand the impact on local water resources.
2. Strict Approval Processes: Water resource assessments must consider extreme drought conditions and the long-term decline of groundwater levels, as well as the pressure on local water supplies during dry periods.
3. Cost Sharing: Tech companies should bear the cost of expanding water and electricity infrastructure and provide tangible benefits to communities (e.g., subsidizing water-saving measures for herders or restoring ecosystems).
Herders may not understand technical terms like WUE, but they know that deeper wells mean weaker livestock, indicating a real ecological threat. Balancing the development of computing power with herders’ needs is key for the sustainability of Ulanqab’s status as a “computing power hub.”