Data Center
A physical facility used for the centralized housing of computer systems and associated equipment. Data centers serve as core infrastructure for the digital economy and cloud computing, consuming large amounts of electrical power and requiring continuous cooling.
Core Facts
Hyperscale data centers typically house more than 100,000 servers, and a single campus can consume hundreds of megawatts of electricity.
A typical large data center can consume millions of liters of water per day, equivalent to the daily water consumption of a small town.
Training a frontier AI model routinely requires hundreds of millions of dollars in computing investment.
Timeline (last 90 days)
Power supply involves long-cycle steps such as grid connection approval, substation construction, and clean energy procurement, and is often more time-consuming than purchasing GPUs.
Data centers have some load-adjustment potential: by shifting non-urgent computing tasks to periods when clean electricity is abundant or electricity prices are lower, dependence on new fossil-fuel peaker units can be reduced.
According to the latest analysis by the Sierra Club, by 2030 the US power sector's carbon dioxide emissions from meeting data center electricity demand could surge 42% compared with 2025.
New hyperscale data centers use less water than TSMC's chip manufacturing.
Possible mitigation paths include stricter water disclosure requirements, incentives to site data centers in non-arid regions, and research and development of low-water or even waterless cooling technologies.
A large data center complex has been built in the Arizona desert to support AI computing demand.
As AI models continue to grow in scale, the water demand of data centers will only increase.
Inconsistent water disclosure standards across companies and the lack of third-party verification make it difficult for outsiders to accurately assess the true water footprint.
Training and running large-scale AI models requires massive computing power, and the cooling systems of data centers consume large amounts of water.
The water use data of many data centers is not transparent, and communities often learn the true water consumption of nearby facilities only after problems emerge.
40 more timeline events
All Facts (20)
Hyperscale data centers typically house more than 100,000 servers, and a single campus can consume hundreds of megawatts of electricity.
90%VerifiedA typical large data center can consume millions of liters of water per day, equivalent to the daily water consumption of a small town.
85%VerifiedTraining a frontier AI model routinely requires hundreds of millions of dollars in computing investment.
75%VerifiedWith the proliferation of large language models and generative AI, power consumption in data centers is rising rapidly, making energy use a key bottleneck in the expansion of the AI industry.
75%VerifiedThe core metric for measuring data center energy efficiency is PUE (Power Usage Effectiveness), which is the ratio of total facility energy consumption to actual IT equipment energy consumption, with an ideal value of 1.0 and typically ranging from 1.2 to 1.6 in practice.
75%VerifiedA 100-megawatt-class data center may consume millions of gallons of water per day, equivalent to the daily water usage of tens of thousands of households.
75%VerifiedNorthern Virginia, home to the world's largest data center cluster, has seen its local utility Dominion Energy file multiple requests to raise electricity rates to cover grid expansion costs.
75%VerifiedThe cooling systems of data centers consume large amounts of water resources, which may exacerbate water scarcity in some regions.
70%VerifiedData center construction in arid regions or water-stressed watersheds directly competes with agricultural and municipal water use.
65%VerifiedThe training and inference of large AI models place enormous demands on computing power, and the electricity consumption and carbon emissions of data centers have become topics of industry discussion
65%VerifiedQueue times for grid access by data centers are growing increasingly long, and transmission capacity in many regions is approaching saturation.
65%VerifiedThe long-term local jobs created by data centers are relatively limited and disproportionate to the resources they consume.
65%VerifiedA large supercomputing data center can consume hundreds of megawatts of electricity annually, equivalent to the total electricity consumption of hundreds of thousands of households.
65%VerifiedTech giants building or expanding data centers to support AI operations have become a significant source of new load on the power grid.
65%VerifiedThe annual electricity consumption of a GW-scale data center campus can reach billions of kilowatt-hours, equivalent to the total electricity consumption of approximately 1 million households.
65%VerifiedThe inference process of large language models consumes as much computational resources as the training process.
65%VerifiedData centers are highly automated with limited local employment contribution; a large data center may require only dozens of technical staff to maintain operations.
65%VerifiedA hyperscale data center with an investment of billions of dollars may require only 50–200 people for daily operations.
65%Verified国际能源署（IEA）2024年报告预测全球数据中心用电量将在2026年前翻倍，其中AI工作负载是最主要的增量来源
90%UnverifiedWith the explosive growth in demand for AI training and inference, global data center construction is entering an unprecedented peak period.
70%