Unpacking President Trump’s Directive to Big Tech to Build Their Own Power Plants

In his recent State of the Union Address, President Trump stated that data center companies should build their own power plants to meet their large electricity demands. In describing what he called a “ratepayer protection pledge,” Mr. Trump said, “we’re telling the major tech companies they have the obligation to provide for their own power needs and can build their own power plants so no one’s prices will go up.[1]”

Across the political spectrum, elected officials, regulators, and utilities are reacting to pressure to ensure that the costs of providing power to large data centers could come at the expense of ordinary ratepayers and electricity consumers. 

The President’s announcement offered very few details about how this idea would work in practice, or how it would interact with state level efforts to address energy affordability. Below, we explore some of the dynamics causing retail electric rates to rise, how the data center buildout is contributing to rising rates, and describe some of the implications of the President’s proposal.

What is “large load growth” and how are data centers contributing to it? 

A single hyperscaler data center can use as much electricity as a medium sized city. Between 2017 and 2023, data center power demands more than doubled because of the proliferation of data centers to power AI applications.[2] 

Electricity needs from large data centers, industrialization, building energy, and other large loads are projected to continue to rise dramatically in the coming years. Recent research from Lawrance Berkly National Lab reported that data centers consumed about 4.4% of total electricity in the US in 2023, and that that rate of consumption could rise between 6.7 and 12% of total US electric load by 2028.[3] 

There is tremendous uncertainty about the scope and scale of this buildout caused by several factors, including 1) The unknown market uptake for AI service, 2) difficulty in distinguishing between definitive and speculative grid interconnection requests by data centers, 3) the unpredictable nature of future energy efficiency gains and computer chip energy demand, and 4) a lack of transparency from data center operators themselves about their energy use.[4] Given uncertainty about whether proposed data centers will come online, utilities are increasingly seeking long term power purchase agreements from tech companies. This is an effort to avoid “stranded assets” and an expensive buildout of electric transmission infrastructure for data centers that don’t ultimately materialize.

How does the data center buildout impact electric prices? 

Electricity prices in the US are a product of complicated market structures with multiple actors operating with split incentives. Complicating this situation is the fact that it often takes less time to build a new data than it does to build the power infrastructure to connect it to the grid. Extended regulatory and permitting processes required to connect power generation to the grid in the United States pose challenges for data centers requiring new transmission infrastructure. Existing regulatory proposals from the Federal Energy Regulatory Commission (FERC) and other utility regulators often seek to trade faster grid interconnection for concessions from data center operators. The “speed to power” imperative generated by tech company’s desire to bring new data centers online quickly is a focal point for regulatory action. 

In regulated states, electricity prices in the United States are approved by state level Public Utility Commissions (PUCs), who are responsible to review and approve rate case filings by electric utilities in their jurisdiction. For utilities operating in regulated markets, companies are guaranteed a return on capital expenses such as building new power plants or transmission infrastructure by incorporating those costs into the rates that they charge all customers. 

Electric utilities passing along the costs associated with building new infrastructure to electricity demand from commercial customers such as data centers is the primary mechanism by which the data center buildout can cause retail electric rates for residential customers to rise. 

The model to provide a guaranteed return to utilities was adopted as a means of ensuring that electricity would be provided to areas where it otherwise might be too expensive to provide. It has worked on the assumption that building new infrastructure resulted in broadly distributed social benefits. This assumption is being tested by new investments in transmission and generation infrastructure being built to serve the needs of large tech companies, but whose costs are being borne by all ratepayers. 

Most researchers agree that data centers cause increases in electric rates, particularly in areas with high concentrations of data center development. For example, research by the Union of Concerned Scientists found that in 2024, electricity customers in Illinois, Maryland, New Jersey, Ohio, Pennsylvania, Virginia and West Virginia faced $4.3 billion in additional costs attributed directly to transmission projects to deliver power to data centers.[5] 

However, recent research from E3 complicates this narrative, arguing that when large load customers pay more than the minimum cost required to serve them, they can put downward pressure on residential rates by generating revenue for the utility. They cite projections from California utility PG&E stating that data center growth could reduce average household electric bills in their service territory by up to 2%.[6] 

Ultimately, electric price increases from data centers are as much a product of decisions that utilities and regulators make about how to meet AI-driven load growth as they are the size of load growth itself. 

State and Local Regulatory Efforts to Address Energy Affordability are Already Underway

A lack of transparency makes understanding how costs are being distributed by utilities difficult to assess. Contracts signed between data center developers and local power utilities are typically confidential, and rate case proceedings can be highly technical and difficult to decipher. Local policymakers often have an incentive to try to attract economic development to their region, and large load users can often negotiate preferential electric rates. Meanwhile, residential customers often have no choice in their electricity providers, and the monopolistic utility model means they often cannot seek other sources of power. Ordinary people are “captive ratepayers,” while large load users and tech companies can freely shop around for the best deal.[7]

Across states and regions, policy and regulatory efforts are already underway to ensure that big tech companies pay their fair share of the infrastructure costs associated with meeting their rapacious energy needs. One of the main policy tools being explored is the creation of “large load” electricity tariffs. These tariffs are designed to enable the utility to recover costs imposed on the electric system by large load users such as data centers. As of December 2025, there were at least 65 large load tariffs under consideration or in place across 34 states.[8]

Major tech companies such as Microsoft and Anthropic have publicly pledged to pay higher electric rates to cover the costs to meet their energy demand. However, it is unclear how these pledges would be enforced in practice, or to what extent they apply to smaller data center companies that large tech companies hire to build and operate their data centers.[9]

States are also exploring “Bring Your Own Generation” (BYOG) models like what was proposed by President Trump. By reducing the need for expensive new transmission infrastructure to meet large load growth from data centers, BYOG could help to reduce upward pressure on electricity rates.

However, there could be other major climate costs imposed by the BYOG approach that depend on the generation source at power plants data center operators build.

The Risks of Fossil Buildout Fueling “Bring Your Own Generation” 

Data center companies are increasingly looking toward off-grid natural gas plants as a means of coming online quickly in the face of persistent permitting challenges and transmission constraints. Large load growth is also causing utilities to extend the life of existing coal power plants by delaying planned retirements, and repowering unused or mothballed plants.[10] 

The United States tripled the amount of gas fired generation capacity in development in 2025. More than a third of this capacity is directly tied to on-site generation at data centers off grid, while many grid connected projects have been announced to meet AI driven energy demand.[11]  That said, the rush by data center companies seeking new natural gas plants has led to supply chain shortages in the turbines needed to operate new plants. Two-thirds of projects in development have not secured a manufacturer, which could limit the extent of gas power deployment.[12]

Increasingly, communities are raising concerns about the creation of a “shadow power grid” designed exclusively to serve data center needs and dominated by natural gas. A recent Washington Post analysis found off-grid data center projects are planned across Texas, New Mexico, Pennsylvania, Wyoming, Utah, Ohio and Tennessee, and noted that these facilities would bring “huge new industrial facilities to communities across the country and increase U.S. emission of carbon dioxide and other air pollutants.[13]

Ultimately, if BYOG proposals like those announced by President Trump end up facilitating a massive fossil fuel buildout, they could reduce electric costs on one hand while increasing emissions related costs on the other.