BloombergNEF forecasts that data center electricity demand will increase 2.7 times by 2035, reaching 106 gigawatts from the current 40 gigawatts, driven by AI expansion and a surge in announced projects, primarily in rural U.S. areas within the PJM Interconnection and Texas ERCOT regions.
The report details how this growth reflects the sector’s rapid evolution, with planned construction adding substantial capacity over the next decade. Facilities are shifting toward rural locations because urban sites are becoming limited as data centers expand in scale. This relocation allows for larger builds away from densely populated zones, accommodating the infrastructure needs of modern computing demands.
Current data center power usage shows that only 10 percent of facilities consume more than 50 megawatts. In contrast, the average new data center coming online over the next decade will require more than 100 megawatts. This shift underscores the increasing size of installations, with Qnearly a quarter of these new facilities exceeding 500 megawatts in power draw. A small number will surpass 1 gigawatt, representing the largest hyperscale operations planned for high-performance computing tasks.
Alongside size increases, operational efficiency will improve, as utilization rates for data centers rise from 59 percent today to 69 percent by 2035. This uptick stems from greater demand for compute resources, particularly in artificial intelligence applications. AI training and inference processes will account for nearly 40 percent of total data center compute by that year, optimizing existing infrastructure and justifying the power expansions.
Global investment in data centers has reached 580 billion dollars this year, exceeding worldwide expenditures on developing new oil supplies. This funding supports the construction and equipping of these facilities, enabling companies to meet rising computational needs. The investment surge aligns with efforts by AI firms to deploy more powerful systems, fueling the overall capacity buildup.
The BloombergNEF report marks a significant update from its April publication, with projections revised upward based on recent developments. A key factor is the increase in newly announced projects since that time. As the report states, “With an average seven-year timeline for projects to come online, developments in earlier stages affect the tail end of our forecast the most.” This means early-phase announcements influence long-term estimates, extending their impact into the 2030s.
Early-stage projects have more than doubled in number between early 2024 and early 2025. These differ from committed or under-construction initiatives, representing potential future capacity that bolsters the forecast. The seven-year average timeline covers planning, permitting, construction, and grid integration, during which projects progress from concept to operation.
New capacity planning concentrates in specific U.S. regions. Virginia, Pennsylvania, Ohio, Illinois, and New Jersey form the core of the PJM Interconnection area, which also includes Delaware, West Virginia, and parts of Kentucky and North Carolina. The PJM Interconnection operates the electrical grid across these states, managing transmission and ensuring reliability. Texas’s ERCOT grid, the independent system operator for most of the state, will host a substantial portion of additions as well, drawn by available land and energy resources.
The PJM Interconnection faces examination from Monitoring Analytics, its independent monitor. This group submitted a complaint to the Federal Energy Regulatory Commission, asserting that PJM must limit new data center connections to instances where grid capacity suffices. Monitoring Analytics emphasizes PJM’s reliability obligations, writing, “As part of its obligation to maintain reliability, PJM has the authority to require large new data center loads to wait to be added to the system until the loads can be served reliably.” The complaint further recommends establishing a load queue to sequence additions systematically.
Data centers contribute to elevated electricity prices in the PJM region currently. Monitoring Analytics attributes this to the influx of high-demand loads without corresponding grid upgrades. The organization criticizes PJM’s approach, stating, “PJM’s failure to clarify and enforce its existing rules and to protect reliable and affordable service in PJM is unjust and unreasonable.” This filing highlights tensions between rapid data center growth and grid management practices.
