The U.S. Department of Energy (DOE), in collaboration with data center industry experts, has released a report on data center energy use, primarily to address the growing energy needs of artificial intelligence and digital infrastructure. Data center electricity demand is growing rapidly due to the popularity of AI applications, and it is expected that by 2028, data center electricity consumption will increase from 176 terawatt-hours (TWh) in 2023 to 325 to 580TWh, accounting for 6.7% to 12% of total electricity consumption in the United States.
The report points to AI servers as the main driver of this growth, with their power demand expected to grow four to eight times between 2023 and 2028. To address this challenge, DOE has proposed a "portfolio approach" that includes innovative measures to improve energy efficiency, develop clean energy, optimize the grid, and advance liquid cooling technology. In addition, the report calls for increased cooperation with utilities to promote the development of scalable nuclear and real-time renewable energy technologies to balance energy demand with environmental sustainability.
After consultation with several personal leave industry stakeholders, including Amazon, Google, Meta, and OpenAI, the US Department of Energy made its findings and recommendations. Create a data center AI testbed within the U.S. Department of Energy to facilitate collaboration among national laboratories, academia, and industry in developing energy-efficient AI algorithms.
Energy utilities, data center developers and operators, and other key stakeholders will collaborate to discuss how to address current power supply bottlenecks. Evaluate and support regional data center expansion as power generation, storage and grid technologies face cost, performance and supply chain adjustments. Encourage the development and deployment of emerging clean energy technologies such as advanced nuclear energy, enhanced geothermal energy, and long-term energy storage.
Ai will bring huge potential growth scale to the power and information technology industry, which is expected to be at the forefront of electricity demand growth. Even with its many recommendations, the report falls short in key areas. Even as the report encourages public-private partnerships to find ways to make LLMS more efficient, there is no discussion of trying to make data centers themselves more efficient. Also, many other countries are concerned about data center energy efficiency and the PUE is not even mentioned in this report. Know that PUE can help the industry focus on sustainability.
A lot of the global investment in data centers is related to artificial intelligence, but large-scale data center development should be happening largely thanks to cloud computing, which will include AI but not necessarily be limited to AI.
The U.S. Department of Energy (DOE) is driving the sustainable expansion of the data center industry through several initiatives to address the growing energy demands of artificial intelligence (AI) and digital infrastructure. Among them, the COOLERCHIPS program is a key initiative to improve cooling in data centers by developing silicon wafers that operate at temperatures lower than existing technology. Including single-phase and dual-phase immersion cooling, direct chip cooling and other technologies, the goal is to complete the proof of concept in the first half of 2026. By 2030, countries with the most efficient and powerful data centers with the lowest total cost of ownership (TCO) will have a huge advantage.
Liquid cooling technology is considered key to solving the cooling and power challenges of data centers and is expected to surpass air cooling as the primary method of cooling IT infrastructure by the end of the century. Liquid cooling technology can effectively cope with the high power density and heat dissipation requirements brought by AI servers through more efficient cooling efficiency and lower energy consumption.
