AI Datacenter Energy Demands Threaten to Overwhelm US Grid, Deloitte Warns

AI Datacenters Threaten to Overload US Power Infrastructure

The energy demand from AI datacenters in the US is set to increase dramatically over the next decade, potentially straining the nation’s power grid to its limits. A recent Deloitte Insights report highlights that datacenters focused on AI could require more than 30 times their current power by 2035, with some new facilities expected to reach 5 gigawatts (GW) in capacity—comparable to the electricity used by millions of homes.

Current and Future Energy Demand

Last year, US datacenters consumed about 33 GW of energy in total, with AI-specific facilities accounting for roughly 4 GW—around one-eighth of the total. By 2035, total datacenter power needs are projected to rise to 176 GW, with AI datacenters alone drawing 123 GW, or 70% of that figure.

Not only will there be more datacenters, but they will also be larger. Currently, the biggest US hyperscale datacenters use less than 500 megawatts (MW), but some under construction may exceed 2 GW. Early planning stages for massive campuses spanning 50,000 acres could push consumption to 5 GW each.

Challenges Facing Power Infrastructure

Meeting this surge in demand is complicated. Most recent growth in datacenter power consumption has been met with increased gas-fired electricity generation, despite many operators' commitments to clean energy goals. The grid infrastructure is struggling to keep pace due to regulatory delays, supply chain disruptions, and long lead times for new power plants.

For example, some grid connection requests are facing waits up to seven years. Building new gas power plants without existing contracts is unlikely to fill the gap before the next decade. Meanwhile, datacenters can be constructed in just a few years, creating a mismatch in timing.

Supply chain issues are also impacting the availability and cost of critical materials like steel, aluminum, copper, and cement — essential for both energy and datacenter infrastructure. Tariffs on imported components add further pressure.

Strategies to Close the Energy Gap

Deloitte’s survey of datacenter operators and power companies revealed three key approaches to address this looming energy gap:

• Technological innovation: Developing more energy-efficient infrastructure, such as optical data transmission within server rooms and solid-state transformers for the grid.
• Regulatory reform: Streamlining approval processes, prioritizing ready projects, and eliminating speculative "zombie" projects could speed up grid connections. The UK’s Ofgem regulator has introduced similar queue management reforms.
• Increased funding: Massive investments across industries are necessary to expand and modernize the power infrastructure to support AI growth.

Wider Implications

If these challenges are not addressed, the US risks falling behind in AI development due to power constraints. This could affect economic competitiveness and national security, as AI infrastructure becomes increasingly critical.

Power companies also face a missed opportunity to upgrade and expand the grid. The need for better infrastructure to support AI datacenters is becoming a key topic in investment discussions, opening doors for new funding.

For those interested in AI and infrastructure trends, staying informed on how energy demands impact technology deployment is crucial. You can explore relevant AI courses and resources at Complete AI Training.