US Data Centers Face 49 GW Power Shortfall by 2028

U.S. data centers will need 74 GW of power by 2028, but only 25 GW of capacity is accessible, leaving a 49 GW shortfall that threatens to stall AI infrastructure buildout and raise electricity costs for every ratepayer in the country. Morgan Stanley Research published the analysis on February 27, 2026.

The Scale of the Gap

Data centers already account for nearly one-fifth of global electricity demand growth. Morgan Stanley projects that data center power consumption will increase by approximately 126 GW annually through 2028, a figure that approaches Canada’s total annual electricity consumption. The 49 GW shortfall represents power that data center operators need but the grid is not positioned to deliver within the required timeline.

Equipment costs have risen 30% since 2019. Transformers, switchgear, and generation assets that once had 12-month lead times now stretch to 36 months or longer. The bottleneck is not just generation capacity. Transmission infrastructure, substation upgrades, and interconnection queues all create compounding delays.

Why This Matters

The power gap affects more than data center operators. Utilities are spending billions on grid upgrades, and those costs flow directly to residential and commercial ratepayers. In central Ohio, where approximately 130 data centers operate, residential electricity rates jumped from 11-12 cents per kilowatt-hour in 2020 to 19 cents in 2025, a 60% increase. Nationally, retail electricity prices have risen 42% since 2019, outpacing the 29% increase in the Consumer Price Index.

Rising power costs are becoming a political issue heading into the 2026 U.S. midterm elections. Consumer backlash against data center-driven rate increases is intensifying in Virginia, Ohio, and Georgia, states with the highest data center concentrations.

Investment and Infrastructure Response

Global power industry investment reached $1.5 trillion in 2025. Hyperscaler companies, the largest data center operators, are projected to spend over $1 trillion in 2025-2026 combined on AI infrastructure. Morgan Stanley estimates $350 billion in potential value creation across the power supply chain as utilities, developers, and equipment manufacturers race to close the gap.

Natural gas is projected to meet roughly one-fifth of new power needs outside China, filling the gap while nuclear, geothermal, and long-duration storage projects work through multi-year development timelines. Power market spreads are expected to increase approximately 15%, reflecting tighter supply-demand conditions through 2028.

The 2026-2028 window is critical. Financing decisions made this year determine whether the infrastructure exists to support AI workloads at scale, or whether power constraints force operators to delay, relocate, or curtail expansion plans.

Critical Perspective

The 49 GW announced here reflects the press-release headline; the conditions, timelines, and cost structures that determine actual delivery are not addressed. Projects of this type in comparable markets have historically run 15-25% over initial timelines due to interconnection review, permitting, or equipment supply constraints. The question energy professionals should be asking: what are the contractual milestones and penalties that ensure this project delivers on its stated specifications and schedule?

Related Coverage

• Data Center Grid Connections Now Take Up to 7 Years in Virginia: data center grid connection delays exacerbate power shortfall
• Big Tech Signs 10 GW of Nuclear Deals to Power AI Data Centers: nuclear power deals to meet data center energy demand
• The Silicon Shock: How AI Hunger: AI’s hunger for power drives data center demand
• PJM 2026/2027 Capacity Prices Reach $329/MW-Day as Data Centers: data centers drive up PJM capacity prices

Sources

• Morgan Stanley Research: Powering AI Energy Market Outlook (2026-02-27)
• Morgan Stanley: Data Center Electricity Consumption Podcast (2026-02-27)
• Yahoo Finance: Morgan Stanley Sees 20% Shortage (2026-02-27)
• Fortune: Morgan Stanley Warns AI Breakthrough Coming in 2026 (2026-03-13)