Most data center developers have a utility strategy. Very few have a commission strategy. Those are not the same thing. The gap between them is where billion-dollar projects go to stall.
There is a particular kind of confidence that comes from moving fast, deploying capital at scale, and watching governors line up to hand you a shovel at a ribbon cutting. The hyperscalers have had a lot of that confidence lately.
State public utility commissions are not impressed.
The surge in AI infrastructure investment has produced one of the most consequential and most underestimated regulatory dynamics in the energy sector right now. Billions are being announced. Load forecasts are going vertical. And the quiet, procedurally meticulous bodies that actually control whether any of this gets built are asking questions that no press release was designed to answer.
Here’s what the dockets are actually showing.
The question commissions are asking is not the one developers expect.
When major data center load hits a state commission proceeding, tech companies tend to assume the reaction will mirror what they get from governors and economic development offices. Jobs. Investment. Innovation. The full ribbon-cutting package. That framing works in political contexts.
It does not work in administrative law.
State public utility commissions are ratepayer protection bodies with statutory obligations and evidentiary records. When significant new load enters their proceedings, they are not asking: how do we attract this investment?
They are asking: who is paying for this, and what happens to existing customers if it goes sideways?
That shift is visible right now in commission dockets across the country. Load forecasting credibility is under active scrutiny. Commissions are openly skeptical of the hockey stick projections utilities are submitting on behalf of hyperscaler commitments. Cost causation principles are being applied aggressively. Stranded asset risk, what happens if a hyperscaler pulls out after infrastructure has been built, is now a standard line of questioning.
AI didn’t just add demand to the grid. It introduced uncertainty at a scale commissions are structurally designed to resist.
KEY TERM
Cost causation: The regulatory principle that the party whose actions necessitate infrastructure investment should bear the costs of that investment. In the context of AI data centers, it means: you built the demand, you pay for the grid upgrades.
Virginia is the pressure cooker. And the data centers didn’t get what they wanted.
The Virginia State Corporation Commission proceedings tied to Dominion Energy’s load growth are the clearest illustration of this collision in real time. The numbers are not subtle: AI data center demand in Virginia contributed to an 833% increase in PJM capacity auction prices for the 2025-2026 market. Virginia’s energy demand is projected to rise 183% by 2040. The load forecasts are, by any measure, extraordinary.
The Commission’s response was methodical, pointed, and notably unmoved by the scale of investment being promised.
In November 2025 the SCC created a new GS-5 rate class for large customers…primarily data centers using over 25 megawatts. These customers are now required to pay a minimum of 85% of contracted distribution and transmission demand and 60% of generation demand. They must sign 14-year contracts covering the costs of grid investments, whether or not their data centers actually get built at the scale projected.
Here’s the part worth underlining: a coalition of data center companies proposed their own alternative plan to the Commission: a less stringent version, naturally, with more favorable terms for the industry. The SCC reviewed it and declined.
The commission wrote its own rules. As it is entitled to do.
That’s not a hostile regulatory environment. That’s a commission doing exactly what it was designed to do, and doing it regardless of how much capital is on the table.
Tech companies think this is a speed game. It isn’t.
The most dangerous assumption technology companies make when they enter regulated energy markets is that capital commitment produces regulatory approval. The logic is intuitive and, in most contexts, correct: we’re spending billions, creating jobs, building critical national infrastructure. Surely the approval follows.
It doesn’t. Not automatically. Not quickly. And not without a defensible evidentiary record built over months of proceedings that move at the pace administrative law has always moved. which is to say, not the pace of a product launch.
Public utility commissions operate under administrative law. They have intervenors: consumer advocates, industrial customers, environmental organizations, and competing interests. All of them have every right to participate in proceedings, cross-examine witnesses, and file testimony that reshapes outcomes. Once something is in a docket, the utility has limited control over where it goes.
The technology company has even less.
The companies that learn this the hard way learn it on the record, under oath, with timelines they don’t control. And in Virginia, they learned it while watching the commission reject the alternative rules they tried to propose for themselves.
What a commission strategy actually looks like.
There is a specific kind of practitioner who knows how to operate in this environment. It is not a lobbyist. It is not general counsel. And it is definitely not whoever handles government affairs at the state level by sending LinkedIn messages to legislative staffers and calling it stakeholder engagement.
Lobbyists work around the edges of a proceeding. Lawyers react to the record once it exists. What regulated markets require is someone who shapes the record before it becomes a problem.
That means understanding when to engage commission staff versus commissioners, and why those are very different conversations. It means knowing how to sequence contacts before filings hit. It means anticipating which intervenors will appear, what arguments they’ll make, and structuring the project narrative to address those arguments before they’re raised in testimony.
Most technology companies show up to commission proceedings with jobs, investment, and innovation. Commissions want to know who pays, what happens if you leave, and why this won’t increase costs for the 2.8 million residential customers who have been ratepayers since long before your data center was a line item in a capital plan.
Bridging that gap early, before the proceeding, before the filing, before the record is established, is the entire game. The practitioners who understand that hold decisive advantage over everyone else in the process.
The actual downside.
Getting commission strategy wrong is not a speed bump. It is multi-year proceedings. Unexpected cost allocation decisions that restructure project economics after the capital has been committed. Mandatory infrastructure contributions that weren’t in the original underwriting. Rate structures that destroy margin assumptions. In the worst cases, denial, or indefinite deferral while the commission works through a methodology question it decided to defer to the next rate case.
In Virginia, the SCC deferred the full cost allocation methodology question to its next rate case two years away. Consumer advocates called it “a good first step” and immediately signaled they want more. The 14-year contract requirement exists specifically because commissions learned they can’t trust load projections that don’t come with financial commitment attached.
For a hyperscaler or developer, the endpoint of getting this wrong looks like this: you secured land, announced billions in investment, held a press conference, and still cannot get reliable, timely power at a predictable cost.
At that point, you are not building a data center. You are sitting on stranded strategy.
Commissions are not anti-AI. They are not anti-growth. They are anti-being left holding the bag when a speculative load commitment doesn’t materialize and existing ratepayers are stuck with the infrastructure bill.
Eight hundred and thirty-three percent. That’s how much PJM capacity prices jumped in one year, driven substantially by data center demand. The commission that approved the GS-5 rate class knew that number. They wrote 14-year contracts and 85% demand minimums into their order specifically because they knew it.
Anyone who walks into a commission proceeding without understanding that is going to learn it on the record, with timelines they don’t control, and outcomes they didn’t model.
The practitioners who understand it before the filing are the ones whose projects actually get built.
You are not building a data center. You are sitting on stranded strategy.
Michael-Christopher Warren is a government affairs and external affairs professional and the founder of RegulatorIndex.com a practitioner-built intelligence platform mapping every U.S. Public Utility Commission for the professionals who can’t afford secondhand analysis.
