Trump Forces Tech Giants to Bankroll Their Own Power as Grid Strains Under AI Demand
White House event on March 4 will formalize pledges from Amazon, Google, Meta, and others to build or buy their own electricity supply for data centers as consumer bills surge nationwide.
President Trump will host executives from Amazon, Google, Meta, Microsoft, xAI, Oracle, and OpenAI at the White House on March 4 to sign what the administration calls the ‘Rate Payer Protection Pledge’—a commitment requiring tech companies to supply their own power for new AI data centers rather than drawing from the strained public grid.
The initiative requires tech companies to ‘build, bring, or buy their own power supply for new AI data centers, ensuring that Americans’ electricity bills will not increase as demand grows,’ according to White House spokeswoman Taylor Rogers. Trump announced the agreement during his State of the Union address Tuesday, calling it the Rate Payer Protection Pledge, positioning it as both economic protection and strategic necessity in the AI race against China.
The move marks a turning point in how the federal government manages the collision between Silicon Valley’s computational ambitions and America’s aging power Infrastructure. Energy demand growth from data centers has become politically precarious ahead of the midterms as it pushes up power bills over a wide swath of the country, transforming what began as a local zoning issue into a national affordability crisis.
The Scale of the Energy Crisis
Amazon, Microsoft, Meta, and Google are committed to spend $650 billion this year combined to develop their AI capabilities, including new data center construction. That investment wave is crashing into a grid that wasn’t built for it. U.S. data center annual energy use in 2023 was approximately 176 terawatt-hours (TWh), approximately 4.4% of U.S. annual electricity consumption that year, but some projections show that data center energy consumption could double or triple by 2028, accounting for up to 12% of U.S. electricity use.
The human cost is already visible. New Jersey customers paid an average of 19 percent more for electricity in 2025 compared with 2024. In Virginia, utilities that implemented 30 percent rate increases between 2020 and 2023 have received approval for additional rate increases of up to 21 percent by 2027. In the PJM market covering 13 states, the cost to secure power supplies has exploded in recent years, with $23 billion attributable to data centers—costs that get passed down to consumers.
Voluntary Pledge or Implicit Coercion?
The White House frames the March 4 signing as a negotiated agreement. Trump said during his State of the Union address that ‘we’re telling the major tech companies that they have the obligation to provide for their own power needs’—language that suggests directive rather than discussion.
The pledge under discussion is expected to resemble commitments already offered earlier this year by Microsoft to invest in new electricity generation and efficiency measures. In January, Microsoft rolled out a ‘community-first AI infrastructure plan’ that included a commitment to ‘pay our way’ and ask utilities to set rates high enough to cover electricity costs for their data centers. Anthropic made a similar pledge earlier this month, promising to cover any electricity price increase stemming from their facilities.
Microsoft and Anthropic announced voluntary pledges to cover consumer rate increases before Trump’s announcement. The White House event formalizes what some companies already committed to—but now with presidential branding and political pressure for laggards to comply.
But voluntary cooperation may be getting a push from regulatory pressure. The Trump administration announced a bipartisan agreement directing power grid operator PJM Interconnection to force data centers to foot their own energy costs. Some Democrats pressed the Trump administration to do more in ensuring tech companies are held to their pledges, with one senator calling a ‘handshake agreement with Big Tech over data center costs’ insufficient.
Infrastructure Reality Check
The solution in the pledge lies in tech companies constructing new power plants to support the energy needs from their individual technologies, producing their own electricity. Trump noted during his speech that ‘we have an old grid’ that ‘could never handle the kind of numbers, the amount of electricity that’s needed’.
Tech companies are already moving in this direction out of necessity. 62% of data centers are considering on-site power generation, with estimates that data centers will use 35GW of behind-the-meter power by 2030. Meta has agreed to build a facility adjacent to a planned 400-megawatt natural gas plant in Ohio. OpenAI ordered 29 gas turbines capable of producing 34MW each for its Stargate data center in Texas, totaling 986MW of power.
But building private power infrastructure at this scale raises environmental and equity questions. Tech companies are building data centers with their own private power plants, a risky bet that will increase carbon emissions and other pollution, according to reporting on off-grid projects in Texas and West Virginia.
The AI Growth Calculus
AI servers use up to 10 times the power of a standard server, and companies are deploying them at an unprecedented scale. A typical ChatGPT query consumes approximately 2.9 watt-hours, while a traditional Google search uses about 0.3 watt-hours—roughly 10 times as much electricity.
The computational demands only intensify at scale. Training OpenAI’s GPT-4 reportedly consumed approximately 50 gigawatt-hours of electricity—enough to power San Francisco for three days—at an estimated cost exceeding $100 million. Inference, not training, now represents 80–90% of computing power for AI and will continue to dominate energy demands.
In 2026, more than 300 state data center legislation bills have been filed across 30+ states in just six weeks, marking a shift from incentive-focused policies to regulatory oversight as energy demands become clearer. States that once competed aggressively to land data centers with tax breaks are now drafting consumer protection measures and demanding transparency on energy consumption.
What to Watch
The March 4 signing ceremony will test whether tech companies accept federal coordination of their infrastructure plans or push back on specifics. Key unresolved questions: Will the pledge apply to existing data centers or only new construction? What enforcement mechanisms exist if companies fail to deliver promised power generation? And will other hyperscalers beyond the announced attendees—including Anthropic, mentioned in some reports but not the official guest list—be compelled to join?
Many states are reconsidering their ‘open-door’ policy towards data centers as the scale of energy demand becomes clearer, with measures ranging from consumer protections to environmental requirements. The federal pledge may preempt state-level restrictions, but it also creates a template for holding tech accountable for infrastructure costs.
Longer term, watch whether this shifts AI development economics. If tech companies must build dedicated power plants rather than tapping subsidized grid power, the capital requirements for training frontier models increase dramatically—potentially consolidating the industry around the few players with sufficient resources to build both computational and energy infrastructure at scale. Failure to address bottlenecks may compel U.S. companies to relocate AI infrastructure abroad, potentially compromising U.S. competitive advantage in compute and AI, creating national security implications that will outlast this administration.