BloombergNEF’s latest report projects that data centers will consume one-fifth of all electricity generated in the United States by 2035, representing a fourfold increase from current levels. This dramatic surge, primarily driven by the escalating demands of artificial intelligence compute, is expected to push data center capacity to nearly 200 gigawatts within the next decade. The implications extend beyond mere consumption, signaling profound challenges for an already strained national electrical grid and highlighting the accelerating pace of AI infrastructure development.

Key Developments

  • Data centers are forecast to use 20% of U.S. electricity by 2035, a fourfold increase from today.
  • AI compute demand is the primary driver, pushing data center capacity to approximately 200 gigawatts over the next decade.
  • The U.S. is expected to host 64% of global AI chips by power demand by 2033, with nearly half of new data center capacity dedicated to AI training and inference.
  • Major grid operators like PJM Interconnection and ERCOT face significant strain, with data centers projected to consume 34% and 22% of their respective electricity, leading to a 76% price increase in PJM over the past year.
  • Globally, aggressive AI adoption could create 1,935 terawatt-hours of new electricity demand by 2033, nearly equivalent to India’s annual consumption.

What Happened

A new report from BloombergNEF reveals a stark future for U.S. electricity consumption, with data centers poised to become a dominant load. The consultancy’s updated forecast indicates that these digital infrastructure hubs will account for one-fifth of the nation’s electricity by 2035, a significant jump from present usage. This revised estimate is 83% higher than BloombergNEF’s prediction from just last December, underscoring the rapid acceleration of demand.

The primary catalyst for this exponential growth is the intense computational requirements of artificial intelligence. AI training and inference are expected to absorb nearly half of the projected 200 gigawatts of new data center capacity over the coming decade. The United States is positioned to be the epicenter of this expansion, anticipated to host 64% of the world’s AI chips by power demand by 2033. This concentrated development is already placing immense pressure on regional electrical grids, with some operators struggling to accommodate the influx of new connection requests.

Why It Matters

The escalating electricity demand from data centers, particularly those supporting AI, presents a critical challenge for national infrastructure and energy policy. The projections indicate that the majority of new data centers will connect to electrical grids that are already operating near their capacity limits. This imbalance between supply and demand has tangible economic consequences, as evidenced by a 76% increase in electricity prices within the PJM Interconnection over the past year.

The situation is particularly acute in regions like the PJM Interconnection, which spans from Virginia to Illinois, where data centers are expected to consume 34% of the electricity. Similarly, ERCOT, covering most of Texas, will need to devote 22% of its generating capacity to these facilities. These figures highlight a looming energy crisis that could impact not only the cost of AI development but also the stability and affordability of electricity for residential and industrial consumers alike.

Industry Impact

The profound shift in electricity demand has far-reaching implications across the technology and energy sectors. For AI developers and cloud providers, securing reliable and affordable power will become an increasingly complex and costly endeavor. This could lead to a geographic redistribution of data center investments, favoring regions with more robust or expandable grid infrastructure, or potentially driving innovation in energy-efficient hardware and cooling solutions.

The electrical utility industry faces immense pressure to modernize and expand its infrastructure at an unprecedented pace. Grid operators like PJM, which previously paused new connection applications for four years due to congestion, are now grappling with a backlog of demand. The threat by American Electric Power to withdraw from the PJM interconnection underscores the severity of the strain, indicating potential fragmentation or restructuring within the energy distribution landscape. Despite these challenges, data centers continue to seek connections, representing 38% of charges in PJM’s most recent capacity auction, signaling their unwavering commitment to expansion.

Analysis

The rapid acceleration of data center electricity demand, particularly for AI, is not merely a forecast but a reflection of current market dynamics and the foundational role AI is beginning to play across industries. The significant upward revisions by BloombergNEF, EPRI, and S&P within a short timeframe indicate that previous models underestimated the velocity and scale of AI adoption and its infrastructural requirements. This suggests a potential “hockey stick” growth curve for AI compute, where the initial phase of development quickly transitions into a period of exponential resource consumption.

The concentration of AI compute in the U.S., projected to host 64% of AI chips by power demand by 2033, positions the nation at the forefront of this technological revolution but also at the epicenter of its energy challenges. While this concentration offers strategic advantages in terms of innovation and economic growth, it simultaneously creates a single point of failure for grid stability and energy security. The strain on regional grids, exemplified by PJM and ERCOT, is not an isolated issue but a bellwether for what other regions and countries will likely face as AI adoption becomes more pervasive. This necessitates a proactive and integrated approach to energy planning, infrastructure investment, and regulatory frameworks to avoid widespread energy disruptions and escalating costs.

Future Implications

Near-term (3-6 months): Expect increased scrutiny on data center energy consumption and potential calls for policy interventions or incentives for energy-efficient AI hardware and operations. Grid operators will likely face immediate challenges in managing existing loads and processing new connection requests.

Medium-term (1-2 years): The rising cost and availability of electricity could influence data center siting decisions, potentially shifting new developments towards regions with abundant renewable energy sources or underutilized grid capacity. Investment in grid modernization and smart grid technologies will likely accelerate.

Long-term (3-5 years): The global electricity demand from data centers, projected to nearly match India’s annual consumption by 2033 under aggressive AI adoption, will necessitate a fundamental re-evaluation of global energy production and distribution strategies. This could drive significant innovation in small modular reactors, advanced battery storage, and direct-to-data-center power generation solutions.

Actionable Insights

  • Evaluate current and projected energy needs for AI workloads and factor rising electricity costs into future budget planning.
  • Explore opportunities for co-locating data centers with renewable energy generation facilities to mitigate grid strain and reduce operational costs.
  • Engage with local and regional utility providers to understand grid capacity limitations and future expansion plans in potential data center locations.
  • Invest in energy-efficient AI hardware and software optimization techniques to reduce the power footprint of compute-intensive tasks.
  • Advocate for policy frameworks that support grid modernization, renewable energy integration, and sustainable data center development.

How much will data centers increase their electricity use in the U.S. by 2035?

Data centers are projected to use four times more electricity by 2035 compared to today, reaching one-fifth of the total electricity generated in the U.S., according to BloombergNEF.

What is driving this surge in electricity demand?

The primary driver is the significant increase in AI compute, with nearly half of new data center capacity dedicated to AI training and inference. This intense computational demand requires substantial electrical power.

Which U.S. regions will be most affected by this demand?

The PJM Interconnection (Virginia to Illinois) is expected to see 34% of its electricity go to data centers, while ERCOT (most of Texas) will devote 22% of its generating capacity. These regions are already experiencing significant grid strain.

How have electricity prices been impacted?

In the PJM Interconnection, the supply-demand imbalance has pushed electricity prices up by 76% over the past year. This reflects the growing cost of managing increased demand on strained grids.

What are the global implications of this trend?

Globally, if AI adoption continues aggressively, data centers could create 1,935 terawatt-hours of new electricity demand worldwide by 2033. This amount is nearly equivalent to India’s current annual electricity consumption.

Key Takeaways

  • U.S. data center electricity consumption is set to quadruple by 2035, reaching 20% of national supply.
  • AI compute demand is the overwhelming factor behind the projected 200 gigawatts of new data center capacity.
  • Major U.S. electrical grids, particularly PJM and ERCOT, face severe strain and rising electricity prices due to concentrated data center growth.
  • Forecasts for electricity demand have been significantly underestimated and are being revised upwards by multiple industry organizations.
  • The global impact of AI-driven data centers could create new electricity demand nearly equal to India’s annual usage by 2033.