The Trump administration is negotiating a new voluntary agreement with several global technology giants in an attempt to limit the impact on residential electricity prices, the grid and water resources amid the rapid expansion of artificial intelligence (AI) data centers and soaring energy prices. According to two unnamed government officials, the draft "agreement" targets major players in the data center and AI fields such as OpenAI, Microsoft, Google, Amazon, and Facebook parent company Meta, aiming to force these companies to bear more costs for the infrastructure expansion they promote.
The draft shows that the proposed agreement will take the form of a voluntary commitment between President Trump and large U.S. technology companies and data center developers, rather than a binding administrative regulation. Relevant commitments focus on several aspects: ensuring that ultra-high-energy-consuming data centers do not push up residents’ electricity bills, do not occupy local water resources, do not weaken the reliability of the grid, and require companies to bear the entire cost of new power generation and grid infrastructure. It's unclear how many companies have agreed to join or received invitations, and the parties involved did not respond to requests for comment.
This initiative is seen as an important attempt by the White House to shape the AI infrastructure landscape without directly issuing regulations. A month ago, the White House made a rare appeal to power grid operators in the Mid-Atlantic region of the United States, hoping to lower electricity prices through special arrangements. As AI data centers spring up like mushrooms after a rain, their huge power demand is said to be likely to further push up electricity prices, placing a political burden on the Trump administration, which has already invested heavily in the construction of data centers. The agreement is seen as a political card to demonstrate the government's efforts to "mitigate the impact" before the mid-term elections.
White House spokesman Tyler Rogers said that as President Trump has said before, top technology companies are working with the president to "pay the bill" for the power consumption caused by building data centers, adding that "more progress will be made soon." Another White House official said that the current draft is "outdated and no longer completely accurate," but did not specify the specific changes. Because the U.S. power grid is highly decentralized, turning the draft into enforceable specific arrangements will still require the cooperation of regional grid operators, state-level regulators, and utility companies in terms of rules and contracts.
In recent years, power companies, regulators and lawmakers have repeatedly warned that the proliferation of AI-driven data centers could overwhelm power systems in some areas and push up already worrying power bills. U.S. Energy Secretary Chris Wright admitted in an interview that people's concerns about the "increased burden" of data centers are understandable, but emphasized that the government is talking to major "hyperscale" data center developers, hoping that they will not only become a force to lower electricity prices in the long term, but also help curb existing price increases in the short term.
One of the core elements of the draft is to require AI data center developers to bear the full cost of the new power generation capacity they build and sign long-term power purchase agreements to avoid passing costs on to other users when the data center is closed or transferred. In addition, companies are required to pay the full cost of transmission upgrades to connect new data centers to the grid. The draft also proposes that companies should cooperate with federal, state and local regulatory agencies to formulate electricity prices and transmission and distribution fee structures so as to "do not hurt" residents' electricity prices as much as possible, and even under ideal circumstances can reduce residents' charges. These principles apply not only to enterprise-owned data centers, but also to facilities leased or operated on behalf of enterprises, to prevent enterprises from outsourcing the impact to third parties.
Electricity prices themselves are already on an upward trajectory. Over the past year, U.S. residential electricity prices have risen even faster than inflation, utility companies have filed for record price increases, and government forecasts show electricity prices will continue to rise in the coming years. The federal government predicts that data center power demand may increase two to three times between 2025 and 2028, and the layout of large-scale data centers has been considered an important driver of higher electricity prices in some areas. A media analysis in 2025 pointed out that electricity prices have increased more significantly in areas with dense data centers; research by the Environmental and Energy Law Project at Harvard Law School believes that consumers are paying for the power grid infrastructure that serves data centers.
However, the White House and industry allies have refuted the idea that data centers are driving up electricity prices. A report released last week by the Edison Electric Association, a trade association representing investor-owned utilities, found that most areas with large numbers of data centers are not seeing higher electricity prices. The report believes that by rationally designing data center-oriented rates and agreements, allowing large technology companies to bear more of the new power generation and transmission costs can actually help ordinary users reduce expenses. Wright specifically mentioned two states where data center demand has grown the fastest, but electricity prices have not increased simultaneously, including North Dakota, where electricity demand has increased by about 35% in the past five years. He said that local nominal electricity prices have not increased, but actual electricity prices have dropped significantly.
The draft comes on the heels of a series of similar pledges by Microsoft to pay more for the power used by its data centers, cover new infrastructure costs and reduce water consumption, while no longer accepting local tax breaks (a provision not included in the White House draft). Trump had previously promoted Microsoft's move on his social platform and said he was working with other technology companies to ensure that the American people would not "pay" for their electricity. Wright revealed that Google achieved a three-year electricity price freeze in Georgia through a deal, and that more large-scale data center projects will be "bundled and released" with lower electricity prices in the future.
Other technology companies also claim to have shouldered the costs themselves. Meta said it has covered all of its energy expenses and commissioned research that pointed out that the clean energy projects it promoted increased grid supply without raising electricity prices for users. The draft also requires that data centers be more directly integrated into grid reliability planning. Signatories are required to use non-critical backup power in data centers to improve system stability in emergencies in coordination with grid operators. Companies will also voluntarily agree to reduce data center loads when necessary to ensure that residential power is prioritized during peak demand and extreme weather conditions.
Relevant policy discussions are also heating up around the concepts of "grid flexibility" and backup power. Texas passed important legislation last year requiring large power users (including data centers) to reduce their load or exit the grid during emergencies; other states and grid operators are also exploring similar mechanisms. During last month's winter storm, Wright called on grid operators to fully mobilize backup power resources for data centers.
In addition to energy issues, the draft also seeks to ease local opposition to data centers that have sparked in some areas. Big tech companies will commit to achieving a "net positive water effect" by developing or acquiring enough water for new facilities to support operations without compromising local water quantity and quality. The agreement also encourages companies to carry out AI education and science popularization projects in surrounding communities and public schools, and adopt best practices to mitigate noise, traffic and other disturbances to neighboring residents.
The agreement could also have real-world appeal for companies looking for federal support to speed up the process of connecting to the grid. Currently, new AI infrastructure projects often encounter bottlenecks due to queues and approvals when connecting to the "backbone power system" that transmits high-voltage power. The draft proposes that the federal government will commit to assisting in accelerating the connection of data centers to these critical power grids, thereby providing additional incentives to companies participating in the agreement. At a time when the Trump administration is pushing for the expansion of AI infrastructure and is facing pressure from energy prices and public opinion, this voluntary agreement, which is still under revision, is becoming a political and policy attempt to balance development and people's livelihood.