Powering AI: The Energy Demands of Data Centers
Artificial intelligence is one of the hottest technologies right now, but it is also one of the most energy-intensive. Behind the recent surge in AI technology is an immense increase in energy demand to power the data centers that make AI possible.
California, an epicenter of the burgeoning AI industry, is home to 32 of the top 50 AI companies worldwide. Silicon Valley is deploying significant resources to develop AI, largely through new data centers that require significant energy to bring on line and operate.
Data centers currently consume approximately 5,580 gigawatt-hours (GWh) per year of electricity in California, or about 2.6% of California’s 2023 electricity demand, according to the California Energy Commission. While AI energy consumption is already notable, it’s only the beginning. A report from Lawrence Berkeley National Laboratory projects data center load growth will “double or triple by 2028.” Data centers could consume nearly one-tenth of the nation’s electricity generation by the end of the decade.
Server racks make up the core powerhouse of data centers and the artificial intelligence capabilities they offer.
Data Centers in the News
CalCCA Testimony on Interconnection of Data Center Large Loads
CalCCA generally supports PG&E’s efforts to attract new load by streamlining and expediting interconnection of new customers to PG&E delivery system. Greater clarity and coordination regarding new loads among all interests – PG&E, CCAs, and new customers – will serve this goal. The coordination should extend to information-sharing between PG&E and the affected CCAs during the interconnection process to enable timely procurement of generation supply to the new load.
Read more here.
Can Data Centers Flex Their Power Demand?
Do data centers really need to strain the grid? In the vast majority of hours, there is plenty of spare capacity to serve these new loads. But in a small number of hours each year, the extra demand could lead to customer-crushing prices and even shortages. There are endless reports about the electricity demand from data centers, but hardly any of them mention that the strain on the grid is limited to a very small number of hours. Couldn’t data centers ramp down during those hours demand pushes the limits of supply?
Read more here.
Data Demand Surge
As new data centers come on line, many will fall within the service areas of CCAs. Silicon Valley Clean Energy (SVCE) and San José Clean Energy (SJCE) are both grappling with increasing energy loads, as a significant portion of this AI expansion is happening in their territories. Pacific Gas & Electric anticipates 3.5 GW of growth in data center energy demand by 2029, with over half targeted for San José.
Data centers present both a challenge and an opportunity for CCAs. “Data centers offer potential benefits for cities and, from a fiscal standpoint, bring in significant revenue while requiring few services,” explains SJCE Director Lori Mitchell. “However, the siting and development of new data centers require careful consideration of land use, water consumption, and energy resources.”
CCAs are working to meet these challenges by securing clean energy sources to power their expanding loads, ensuring that their growing customer bases remain sustainably and reliably served. “We are working closely with our member cities to understand how AI and data center demand impact our load,” said Monica Padilla, CEO of SVCE. “We are continuing to procure clean power to meet the forecasted demand.”
However, CCAs need transparency and early information about these new energy demands to properly plan and mitigate any potential risks to their existing customers. Sabrinna Soldavini, director of policy at MCE, highlights this need for advanced notice, saying, “CCAs should be provided information on potential new large loads, particularly data center load, as early as practical to minimize impacts to other costs, compliance risk, and impact to other customers.”
From Strain to Gain
Interestingly, AI itself may offer solutions to help CCAs better manage the impact of growing demand. As CCAs continue to scale their renewable energy portfolios, AI and data centers can improve efficiency and grid optimization. Machine learning algorithms can identify grid inefficiencies, AI-driven load forecasting can help predict energy demand more accurately, and AI analysis of usage data can produce better CCA demand response programs.
Data centers themselves can alleviate demand strains, too. The waste heat generated from the centers can be used for residential heating, and potential demand flexibility would turn the data centers into resources for demand response. “While we’re excited by the growing interest in data center development, we’re also encouraged to see some data centers catalyze other developments,” said Mitchell. “For example, district energy systems can harness their energy more sustainably and efficiently, contributing to our broader environmental goals and resource management strategies.”
CCAs face challenges with the expanding energy demand of data centers, but they also have an opportunity to lead the way in integrating AI and clean energy solutions. The rise of AI highlights the importance of continued innovation, collaboration, and transparency to ensure that the needs of both new and existing customers are met without compromising clean energy objectives or grid reliability. With the right strategies and adequate information, CCAs can not only meet growing demand but also uncover solutions that usher in California’s clean energy goals.
