Congress Sent AI Data Centers the Tab, 417 to 3

Policy & Markets By EnergyInnoCore Editor · Published Oct 9, 2026 · 11 min read

Last week, something happened in Washington that almost never happens anymore. The House of Representatives agreed on something. On September 16, 2026, the House passed the Ratepayer Protection Act, by a vote of 417 to 3. Only three members voted no.

It would be easy to file this under "another bill that may or may not survive the Senate" and move on. I think that would be a mistake.

This is the first bill in this Congress to pass the House that speaks directly to how data centers and their developers pay for, and ultimately secure, the electricity they need. It does not ban anything. It does not cap growth. What it does is put a very clear federal marker on a question that has been bouncing around utility commissions for the past two years.

Who pays for the grid that AI needs?

The answer Congress is signaling is simple. The data center does.

In previous articles, I covered how AI data centers affect household electricity bills and why Behind the Meter (BTM) power is becoming a mainstream strategy. This bill connects those two conversations. If electricity from the grid becomes more expensive and more conditional for large data centers, developers will look harder at bringing their own power. And from the utility side, the same bill gives utilities a tool they have badly wanted, a way to tell real demand apart from phantom demand sitting in their interconnection queues.

What the Bill Actually Says

The Ratepayer Protection Act was introduced by Representative Gabe Evans of Colorado and Representative Kathy Castor of Florida, one Republican and one Democrat. A Senate companion is led by Senator Jon Husted of Ohio.

The bill works through an old and somewhat obscure law, the Public Utility Regulatory Policies Act of 1978, better known as PURPA. PURPA contains a section that allows Congress to create federal standards that every state utility commission must formally consider. That is the mechanism here.

The standard applies to data centers with loads of 100 MW or more, which covers essentially every hyperscale AI campus being planned today.

Under the proposed standard, those large data centers would:

• Pay the full incremental cost of any generation, transmission, or distribution upgrades their operations trigger.
• Provide financial assurances before the utility commits capital to build that infrastructure.
• Remain responsible for those costs if the project is canceled, downsized, relocated, or stops taking power early.

One detail deserves attention. The bill reaches generation costs, not just the wires. Historically, most large load agreements focused on transmission and distribution facilities. Including power supply costs is a meaningful expansion, and it matters a great deal in markets like PJM where capacity has become scarce and expensive.

The timeline is also defined. State commissions would have one year from enactment to open a proceeding and two years to reach a final decision. States that already have comparable standards in place, including Virginia, Oregon, Ohio, and Oklahoma, would not need to start new proceedings.

The Word That Matters Most: “Consider”

Here is the part that often gets lost in the headlines. The bill does not require states to adopt the standard. It requires them to consider it.

That distinction is real. Under PURPA, a state commission can hold a formal proceeding, review the evidence, and decide not to adopt the federal standard, as long as it documents its reasoning. Many states did exactly that with the PURPA standards Congress added in 2005 and 2007.

Some analysts have described the bill as a procedural floor rather than a substantive mandate. ClearView Energy Partners observed that it largely reinforces a transition that is already underway, noting that only 13 states still lack any form of large load tariff.

So why does it matter?

Because direction matters as much as mandate. When 417 members of Congress agree that data centers should pay their own way, every state commission, every utility, and every developer hears the message. Commissions that were on the fence now have federal cover. Utilities that wanted stronger tariffs now have a template. And developers now know that the era of spreading grid costs across all ratepayers is ending, whether their state adopts this exact standard or something similar.

Why Congress Moved Now

U.S. electricity prices have risen roughly 27 percent since 2019. Data center construction increased by about 57 percent between July 2025 and July 2026. A recent poll found that 61 percent of voters oppose new data center construction in their communities. And the midterm elections are only a few weeks away, with 36 governor races and 35 Senate seats in play.

PJM has become the clearest example of the pressure. According to PJM's independent market monitor, data centers were responsible for about $6.3 billion, or 38 percent, of the charges in the most recent capacity auction, and about $29.4 billion across the last four auctions. PJM is now running a backstop procurement to cover a shortfall of roughly 6.8 GW. When households see those numbers turn into higher bills, elected officials feel it quickly.

House Majority Leader Steve Scalise framed the bill as protecting ratepayers while AI infrastructure is built out. Energy and Commerce Chairman Brett Guthrie tied it directly to national competitiveness, describing U.S. leadership in AI as critical to economic prosperity and national security. That framing tells us something. Congress is not trying to slow AI down. It is trying to make the buildout politically sustainable.

Impact One: Grid Power Gets More Expensive and More Conditional

For developers, the most immediate effect is on cost and certainty.

If a state adopts the standard, a 300 MW campus that triggers a new substation, transformer banks, and transmission reinforcements would carry those costs directly instead of sharing them with the broader rate base. On top of that, the developer would need to post collateral or other financial assurances before the utility begins construction.

Put simply, grid power stops being a utility service that shows up eventually and becomes a capital commitment the developer must fund and guarantee up front.

It also creates a patchwork. Some states will adopt the standard as written. Some will adopt tougher versions. Some may decline and try to attract projects with lighter rules. Developers will need to read state utility proceedings as carefully as they read land deals.

Impact Two: The Push Toward Behind the Meter Power

This is where the bill connects directly to the BTM discussion.

Interestingly, the Ratepayer Protection Act does not require BTM generation. It contains no self supply mandate and no curtailment obligation. It is fundamentally a cost allocation bill.

But cost allocation drives behavior.

Consider the choice a developer faces. Option one is to wait three to seven years for a grid connection, pay the full incremental cost of the upgrades, post financial assurances, and take the risk that the grid timeline slips again. Option two is to build or contract on site generation, such as gas turbines, reciprocating engines, or fuel cells, and energize in a fraction of that time, possibly bridging to the grid later.

When the grid option carries more of its true cost, the BTM option looks more competitive. Not always cheaper, but more competitive, especially once Time to Power is priced in.

The bill does not stand alone either. PJM has proposed curtailing new large loads that do not bring their own supply during grid emergencies and excluding them from future capacity forecasts. Australia now requires large data centers to bring clean, firmed power and pay full network costs. The direction is consistent. Bring your own power, or pay the full price of ours.

There is a caveat worth stating plainly. BTM has its own bottlenecks. GE Vernova's gas power backlog and reservations reached about 116 GW by the second quarter of 2026. Transformers, switchgear, and air permits all sit on the critical path. ClearView has warned that if rules move toward requiring data centers to bring new capacity, construction timelines could stretch by years, because power plants simply take longer to build than data centers.

So the winners will not be developers who simply choose BTM. They will be the developers who plan BTM early, secure equipment slots, and design the permitting strategy alongside the technology choice.

Impact Three: Utilities Finally Get a Filter for Phantom Demand

From the utility perspective, this may be the most valuable part of the bill.

Interconnection queues across the country are crowded with data center requests. Many are real. Many are not. A single developer may file requests in several utility territories for the same project, planning to build only where power shows up first. Others file early simply to hold a place in line. Utilities call this speculative or phantom demand, and it creates a serious planning problem.

If a utility plans generation and transmission for every request in the queue, it risks overbuilding infrastructure that existing customers end up paying for. If it discounts the queue too heavily, it risks leaving real projects without power. Either way, the utility is guessing.

Financial assurances change that dynamic. When a request comes with collateral and a commitment to pay for upgrades even if the project walks away, filing in five territories becomes expensive. Developers will be more selective. Utilities will see a queue that looks much more like actual demand.

That cleaner signal benefits nearly everyone. Utilities can plan with more confidence. Regulators can approve investment with less risk of stranded assets. And serious developers with real projects may actually move faster, because the queue in front of them gets shorter.

Impact Four: A Reshaped Market for Developers and Hyperscalers

Taken together, these effects point to a data center market that looks different in two or three years.

Well capitalized hyperscalers and developers can absorb upfront grid costs and post financial assurances. Smaller developers and speculative land plays will find it harder. Expect more consolidation and more partnerships between developers, utilities, and independent power producers.

Site selection will weigh regulatory climate as heavily as fiber, land, and water. States that adopt clear, predictable large load tariffs may actually attract serious projects, because predictability has value even when costs are higher.

BTM strategies will become part of the base case rather than a contingency plan. Power strategy will be decided before the building design, not after.

Not everyone in the industry is comfortable with the details. The Data Center Coalition supported an earlier, broader version of the bill that applied to all large loads, but raised concerns when the scope narrowed to data centers specifically, arguing that it creates uneven regulatory treatment. That debate will likely continue as states write their own rules.

What Comes Next

The bill still has a long road ahead. Here is what I will be watching.

The Senate. Passing before the midterm recess would likely require unanimous consent, meaning a single senator could hold it up. Most observers expect action to slip until after November.

Competing proposals. Senator Martin Heinrich has introduced the GRID Savings Act, which would direct FERC to set national rules for large load interconnection. He has said the House bill does not go far enough.

FERC and PJM. FERC's rulemaking on large load interconnection, PJM's backstop procurement, and PJM's possible ride through standards following the July disconnection of about 3.8 GW of data center load will shape the technical rules that sit underneath any cost allocation framework.

The states. If the bill becomes law, fifty separate utility commissions will spend the next two years deciding how to apply it. That is where the real decisions will be made, and where developers should be paying close attention.

Final Thoughts

The Ratepayer Protection Act will not, on its own, change a single electricity bill tomorrow. It may not even become law this year.

But the 417 to 3 vote tells us where the conversation is heading. The old model, where large new loads arrived and the cost of serving them was spread across everyone, is ending. In its place is a model where data centers pay for what they trigger, commit capital up front, and increasingly bring their own power.

For utilities, that means a clearer picture of real demand. For households, it means some protection from cost shifting. And for AI data center developers, it means the power strategy is no longer something to figure out after the land deal closes.

EnergyInnoCore Editor

Independent analysis of power, grid and policy for AI data centers. Every article lists its sources. About the editor · Editorial policy

This article is for general information only and is not engineering, legal, financial or investment advice. See the disclaimer.