Trump's AI Data Center Power Pledge Faces a Reality Check as Electricity Bills Rise

President Donald Trump expanded his AI data center power pledge on July 23, despite a 75.5% annual increase in wholesale costs across the PJM grid. The expansion adds governors, utilities, electric cooperatives, and data center developers to a commitment previously signed by major technology companies. As highlighted by tom hardware, the central promise is simple: households should not finance the infrastructure required by private AI projects.

Trump claims the broader coalition will eventually lower household electricity bills by producing more power than data centers consume. The White House says participants will build, bring, or buy new generation while covering their connection and transmission expenses.

However, the pledge remains voluntary. That distinction matters because electricity costs are already rising across PJM Interconnection, the largest regional grid operator in the United States. Maryland officials also claim their customers face two billion dollars in capital costs for transmission projects driven mainly by data center demand elsewhere.

The expansion therefore creates a clear test. A much larger coalition has accepted the principle that data centers should pay their way. Yet existing grid rules can still distribute infrastructure costs among households that never requested the new demand.

Trump Put More of the Power Industry Inside the Pledge

The July expansion turns a technology company promise into a broader agreement covering the institutions that regulate, build, and deliver electricity.

Trump introduced the first version of the Ratepayer Protection Pledge on March 4, 2026. Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI were among its original corporate participants.

Those companies committed to procuring the power needed for new data centers. They also agreed to pay for related generation, delivery systems, and network upgrades. Separate rate structures would make companies pay for reserved electricity and infrastructure even when they did not use the full allocation.

The July expansion brought the other major participants in the electricity system into the arrangement. According to the expanded pledge, it now includes 23 governors, more than 150 utilities and cooperatives, and 29 data center developers.

The White House lists Equinix, Digital Realty, Aligned Data Centers, and other developers alongside utilities such as American Electric Power. State governors represent the regulatory layer because utility rates and infrastructure approvals often depend on state authorities.

The administration says the coalition contains more than 200 organizations. It reportedly covers 80% of the electricity delivered to American homes and businesses. The White House also claims 263 million Americans are covered when a data center is built nearby.

Coverage does not mean that every resident has received a contractual guarantee. Instead, it describes the population served by participating utilities and governments. The pledge establishes common commitments, while actual protection depends on tariffs, contracts, regulatory orders, and cost-allocation rules.

The five commitments require participants to add power, finance delivery infrastructure, and accept separate payment arrangements. They also cover local workforce investment and grid resilience. Data centers are encouraged to make backup generation available during shortages when practical.

Trump said the resulting construction would leave utilities with excess electricity for the wider grid. “Electricity bills for American families will actually come down,” he said during the announcement, according to pledge coverage.

That prediction goes beyond preventing cost transfers. It assumes new data center generation will add enough usable supply to reduce costs for other customers.

The White House cited several arrangements as evidence. It said NiSource agreements involving Amazon and Alphabet should return at least 1.4 billion dollars to Indiana customers over 15 years. It also credited an Entergy agreement with Amazon for roughly two billion dollars in customer benefits in Mississippi.

Georgia Power has committed to freeze base rates through 2029 under an agreement cited by the administration. Alliant Energy arrangements in Wisconsin require developers to cover their energy and infrastructure needs, according to the White House.

These examples show how the pledge is intended to work. A utility negotiates a dedicated tariff or contract for a large customer. The data center then accepts minimum payments and construction obligations instead of placing uncertain demand risks on the general customer base.

However, each example operates under different state rules. A collection of individual agreements is not the same as a binding national standard. That gap separates the pledge’s broad political message from its implementation across dozens of utility jurisdictions.

Tom Hardware Numbers Put the Affordability Claim Under Pressure

The pledge promises lower bills while the country’s largest regional power market is reporting sharply higher wholesale costs.

PJM Interconnection coordinates electricity markets and transmission across all or parts of 13 states and Washington, D.C. Its territory includes Northern Virginia, which contains the country’s largest concentration of data centers.

PJM does not own the power plants or household distribution networks within its region. It manages wholesale markets, plans regional transmission, and ensures that enough generation remains available to meet future demand.

An independent market monitor reported that PJM’s total wholesale power cost averaged 136.53 dollars per megawatt-hour during the first quarter of 2026. That was up from 77.78 dollars during the same period in 2025.

The increase was 75.5%, a figure emphasized in tom hardware reporting about the region’s data center demand. Higher energy costs contributed heavily, while rising capacity expenses and transmission congestion added further pressure.

Capacity payments compensate generators for remaining available during future periods of high demand. They do not represent a household’s complete electricity bill. However, utilities can eventually pass those costs through to customers under applicable regulatory rules.

PJM’s recent capacity results show why officials are concerned. Forecast demand has grown faster than available supply, with data centers forming a major part of the increase. Retiring plants, electrification, and delays in adding generation also affect the balance.

The grid operator’s auction for the 2028 to 2029 delivery year cleared at its price cap. It also produced a capacity shortfall of about 6.8 gigawatts against PJM’s target reserve margin. A reserve margin represents additional capacity intended to cover outages and unexpected demand.

That result weakens any claim that surplus electricity is imminent across the entire region. Developers have announced new generation, but proposed power plants can take years to receive permits, secure equipment, connect to the grid, and begin operating.

Demand can arrive faster. A large data center campus can request hundreds of megawatts, while several planned campuses can reshape a utility’s long-term forecast. Artificial intelligence facilities also operate at high utilization because expensive accelerators generate value only when they remain active.

The White House has blamed PJM’s governance and planning process for failing to respond quickly enough. A spokesperson said the operator had not implemented principles supported by the administration and the region’s 13 governors.

PJM faces a genuine timing problem. It must plan for demand that developers say is coming, even when final construction remains uncertain. Underbuilding risks shortages, while overbuilding can leave customers financing infrastructure for projects that never reach their proposed size.

The pledge tries to move that forecasting risk toward developers. Its take-or-pay principle requires companies to cover contracted infrastructure even when their facilities use less electricity than expected.

That mechanism works best when it appears in an enforceable tariff or service agreement. A voluntary signature alone does not automatically rewrite an existing utility contract or PJM’s regional cost-allocation method.

The White House describes the expanded coalition as an answer to enforcement concerns. Governors influence state policy, utilities write tariffs, and developers sign service arrangements. Bringing them together should make implementation easier.

Still, cooperation does not erase conflicting financial incentives. Utilities can earn regulated returns on approved infrastructure. Developers want rapid connections and competitive rates. State officials want investment without provoking voters who already face higher bills.

These incentives explain why tom hardware data about PJM matters more than the coalition’s size. The test is not how many organizations support the principle. The test is whether future regulatory filings place identifiable data center costs on the customers creating them.

Maryland Shows How Grid Costs Can Escape State Protections

Maryland’s dispute demonstrates that a local data center tariff cannot stop every regional transmission charge from reaching ordinary customers.

In May 2026, the Maryland Office of People’s Counsel filed a complaint with the Federal Energy Regulatory Commission, or FERC. The agency represents residential utility customers in state and federal proceedings.

The complaint challenges PJM’s method for allocating regional transmission expenses. Maryland says PJM advanced 22 billion dollars in transmission projects during three competitive planning windows, largely to serve expected data center growth.

Under the existing allocation method, Maryland customers were assigned two billion dollars in capital expenditures. The state consumer advocate estimates those investments will add 1.6 billion dollars to Maryland electricity bills over ten years.

Residential customers would bear an estimated 823 million dollars of that amount. Commercial customers would carry 146 million dollars, while industrial customers would face 629 million dollars.

The consumer advocate estimates an average residential impact of roughly 345 dollars over the decade. These are allegations and projections presented in a pending regulatory complaint, not a final FERC judgment.

Maryland argues its customers did not cause most of the projected demand. The Office of People’s Counsel says anticipated load growth is much higher in Virginia, Ohio, Pennsylvania, and Illinois, where major data center development is concentrated.

PJM’s current method distributes some high-voltage project costs across its territory according to regional demand. It allocates another portion through power-flow modeling, which estimates where electricity using a transmission project will travel.

This approach reflects the shared nature of a regional grid. A transmission line built in one state can improve reliability or transfer capacity elsewhere. Assigning every project only to its immediate location can overlook those wider benefits.

Maryland’s complaint says the current calculation fails when a distinct group of very large customers creates most of the new need. It asks FERC to assign data center-driven costs to the zones hosting those facilities.

As an alternative, Maryland wants PJM to charge large data center customers directly. The state’s regulatory complaint says regional rules currently undermine the pledge’s promise that households will not fund data center upgrades.

This is the most important reversal in the story. A data center can accept a special utility tariff within its host state while neighboring customers still receive part of a regional transmission bill.

Large-load tariffs generally govern the relationship between a utility and a major customer. They can impose minimum payments, collateral requirements, longer contract terms, and exit fees. Those measures protect customers when a planned facility is canceled or uses less power than forecast.

However, regional transmission expenses can enter bills through a different route. PJM plans facilities across multiple utility zones, and FERC regulates the interstate allocation of those costs.

The distinction makes the pledge harder to implement than its language suggests. Governors and utilities can address local generation and connection expenses. They cannot individually replace a federally regulated regional cost methodology.

The Trump administration can pressure PJM and support changes before FERC. Congress can also establish a broader rule. Until one of those pathways produces enforceable terms, the Maryland case remains a direct challenge to the White House’s affordability claim.

FERC has separately moved to accelerate connections for large loads while requiring those customers to cover necessary connection upgrades. That action addresses one important category of expense.

It does not automatically resolve every previously approved regional project or determine which zone caused each investment. The Maryland proceeding therefore deserves attention beyond one state’s bills.

If FERC accepts Maryland’s argument, other states with modest data center growth could challenge costs assigned through regional planning. If FERC rejects it, the pledge will need another mechanism to prevent cross-state cost transfers.

The dispute also exposes a political problem. Residents are unlikely to distinguish among generation, capacity, local distribution, and regional transmission charges. They will judge the pledge by the total amount on their monthly bill.

A Voluntary Promise Still Needs Enforceable Utility Rules

The coalition can guide negotiations, but only binding contracts and regulatory decisions determine who ultimately pays.

The White House’s expanded pledge addresses a weakness in the March announcement. Technology companies cannot control electricity prices alone because utilities, state commissions, grid operators, and federal regulators govern different parts of the system.

Adding those institutions increases the chance that the commitments reach actual tariffs. It also gives voters a public standard against which they can compare future utility proposals.

The core promise includes several useful protections. Developers should pay for new generation. They should cover delivery upgrades, accept minimum obligations, and remain responsible when expected demand fails to materialize.

Those provisions address risks that have appeared in earlier infrastructure booms. A utility can build around ambitious demand forecasts, only to recover its costs from remaining customers if a large project is delayed or abandoned.

Long contracts and minimum payments reduce that exposure. Credit requirements can prevent a thinly capitalized development entity from leaving customers with unpaid infrastructure.

Separate accounting also improves transparency. Regulators can compare a data center’s payments with the generation, transmission, and distribution expenses attributed to that facility.

The problem is that the pledge does not preempt state law, change PJM’s tariff, or create an automatic federal enforcement process. Its text says companies will “voluntarily negotiate” the relevant structures.

That wording leaves the final terms to individual proceedings. Utilities can define eligible costs differently, and state commissions can accept different risk allocations. Regional projects introduce another layer of disagreement.

California provides an example of the political tension. Matthew Freedman, an attorney with The Utility Reform Network, told the Associated Press that technology companies supporting the pledge were also opposing state legislation intended to make its principles mandatory.

The criticism does not establish that every signer opposes enforceable protections. It does show why public commitments must be compared with lobbying positions and regulatory filings.

A company can support the general principle that data centers should pay their way while disputing a specific bill’s calculation. It might challenge which facilities qualify, how causation is measured, or whether an expense benefits the wider grid.

Utilities can raise similar objections. A dedicated customer may cause a new project, but other customers can receive reliability benefits from it. Regulators must decide how to value and distribute those benefits without disguising a subsidy.

The pledge also cannot guarantee lower bills because data centers are not the only factor affecting electricity costs. Fuel prices, extreme weather, power plant retirements, transmission congestion, financing expenses, and local distribution investments all influence rates.

Even if signers cover every data center-specific expense, other costs can rise. Trump’s stronger claim that household bills will decline therefore requires more than successful cost separation.

New generation must enter service quickly enough to improve the supply balance. It must also connect where the grid needs it and remain available during periods of peak demand.

On-site generation presents another complication. A campus can become less dependent on the grid while using natural gas turbines, batteries, or other local resources. That arrangement can reduce connection pressure, but it does not necessarily provide surplus power to households.

Exporting electricity requires appropriate interconnection equipment, permits, market participation, and operating agreements. Backup generators designed for emergencies might face emissions restrictions or lack permission for regular operation.

The administration points to a 900-megawatt Crusoe campus in Abilene, Texas, powered through on-site natural gas generation and battery storage. That model shows how developers can reduce direct reliance on existing supply.

It does not prove that every region can replicate the approach. Fuel access, land, air permits, transmission conditions, and state regulations vary substantially.

A recent analysis from ICF estimated that rising electricity demand could increase monthly utility bills by 15% to 40% by 2030. Forecasts at that range contain uncertainty, but they illustrate the scale of the risk policymakers are addressing.

Research can also produce different conclusions depending on the period and method studied. One recent working paper found that data center development modestly reduced average retail rates between 2015 and 2024.

The paper’s argument reflects economies of scale. Large customers can spread fixed grid costs across more electricity sales when infrastructure is used efficiently and demand persists.

That historical result does not settle the current debate. The latest AI campuses are larger, planned growth is faster, and constrained regions like PJM face different supply conditions.

Both outcomes are plausible under different contracts. A durable large customer can broaden a utility’s revenue base and reduce average costs. A speculative project can leave households supporting oversized infrastructure.

The pledge is therefore best understood as a risk-allocation framework. Calling it proof of future savings goes beyond the evidence currently available.

The Next Three Signals Will Determine Whether Bills Fall

The pledge will become credible only when its commitments appear in enforceable tariffs, regional cost rules, and measurable household outcomes.

The first signal is FERC’s response to Maryland’s complaint and related large-load proceedings. That decision will show whether regional transmission costs can be assigned more directly to the zones or customers driving new demand.

A ruling for Maryland would strengthen the pledge’s central claim. It would reduce the chance that customers in one state finance data centers concentrated in another.

A rejection would expose a major limitation. States could protect customers from local connection costs while remaining vulnerable to charges distributed through a regional grid.

The second signal is the content of new utility tariffs. Signatures and press releases matter less than minimum payment terms, security requirements, exit fees, and definitions of recoverable infrastructure.

Regulators should examine whether a tariff covers generation, local delivery, regional transmission, and stranded assets. They should also determine how long a customer remains responsible after reducing or canceling its project.

The most persuasive filings will identify costs before construction begins. They will explain how expenses move if demand forecasts change, rather than relying on later negotiations.

Tom hardware readers should also watch whether the House advances federal legislation requiring large data centers to bear grid upgrade costs. A bipartisan proposal approved by the House Energy and Commerce Committee would create a national standard beyond the voluntary pledge.

Legislation could direct FERC to establish consistent rules across regional grid operators. However, statutory language would still need to distinguish customer-specific upgrades from projects delivering broad system benefits.

The third signal is the gap between wholesale costs and residential bills. PJM’s 75.5% wholesale increase does not translate directly into an identical household increase, but it creates pressure that will appear through future rate adjustments.

The relevant comparison is not a single month. Observers should track capacity charges, transmission riders, utility rate cases, and average residential prices across participating states.

If those measures stabilize while data center demand grows, the White House will have evidence that the pledge is working. If they continue rising, officials must separate data center costs from other drivers before claiming success.

PJM’s next planning decisions will offer another practical test. The grid needs enough generation and transmission to support projected demand, but it must avoid charging customers for facilities based on unreliable forecasts.

Developers can improve that process by providing deposits, binding schedules, and realistic power requirements. Utilities can publish clearer information about which investments serve an individual campus and which benefit the wider system.

Communities will also judge the pledge through local outcomes. They will compare construction employment and tax revenue with water demand, land use, noise, emissions, and electricity costs.

Trump has urged local leaders to persuade residents that data centers will create wealth. That political argument becomes harder when households see bill increases before the promised tax revenue or power supply arrives.

The expanded pledge gives those communities a useful question to ask: which signed agreement guarantees that this project covers its full cost?

That question is more productive than debating whether every data center is inherently beneficial or harmful. Projects differ, and their contracts determine how risks and gains are divided.

The administration has assembled participants representing most stages of data center development. It has also made a measurable promise that ordinary customers will not finance private AI infrastructure.

Now the coalition must translate that promise into enforceable numbers. The next utility filings and FERC decisions will reveal whether the expansion changes cost allocation or simply broadens political support.

For readers following tom hardware coverage, the Maryland case provides the clearest benchmark. Watch where each infrastructure charge lands, how long developers remain liable, and whether residential bills decline after new power enters service. The pledge succeeds only if households receive protection before projects create costs, not compensation after those costs have already entered rates.