On September 30, the U.S. Senate voted 57–43 to advance the Ratepayer Protection Act — three votes short of the 60 needed to break a filibuster. Two weeks earlier, the House had passed the identical bill 417–3, a margin that united the chamber’s most progressive and most conservative members behind a single proposition: data centers should pay for the grid upgrades they cause. The Senate killed it on a procedural vote anyway. But the principle at the bill’s core — that the largest new loads on the American grid should bear the full cost of serving them — didn’t die with it. Within hours, it had resurfaced inside the biggest energy bill of the year, in a tougher form than the one that failed.
The stakes here are pure cost allocation, the quiet machinery that decides whose bill absorbs each dollar of grid investment. As we traced in our wholesale-to-retail piece, every dollar of generation, transmission, and distribution spending lands somewhere: on the customer whose demand caused it, or socialized across every ratepayer on the system. The Ratepayer Protection Act was Congress’s first attempt to answer that question for the AI era. Its failure and its afterlife tell you where the politics of “who pays” is heading.
What the bill would have done
H.R. 9340, led in the Senate by Jon Husted (R-OH) and in the House by Gabe Evans (R-CO) and Kathy Castor (D-FL), would have amended the Public Utility Regulatory Policies Act of 1978 to create a new federal large-load standard. The mechanics were straightforward. For nonresidential customers with peak demand of at least 100 megawatts at a single site or campus — in practice, AI data centers — utilities would be required to recover the full incremental cost of the generation, transmission, and distribution upgrades needed to serve them. Utilities would also have to obtain financial assurances or upfront contributions before building those upgrades, and the large customer would remain on the hook for cost recovery even if it terminated service early — the stranded-asset protection, aimed at the utility that builds a substation for a campus that never fills up.
The catch was in the verb. States would only have to consider the standard: utility commissions would get up to two years to run a proceeding and decide, with final say resting with the states. The bill codified the voluntary commitments of President Trump’s Ratepayer Protection Pledge from March 2026, under which major technology companies promised to cover their own infrastructure costs rather than shifting them to households. It was designed as a nudge, not a command — federal suggestion layered over state authority, deliberately built to win 417 House votes.
Why “consider” killed it
The Senate math was simple: 57–43 on cloture, three votes short of 60, after an earlier unanimous-consent attempt had already been blocked. But the substance of the opposition is what matters for the idea’s future. Senate Democrats argued that a bill requiring states to merely consider a standard was protection in name only — the status quo with better branding. Their position, in essence: if large loads paying their own way is good policy, make it mandatory; if it isn’t, don’t pretend a suggestion fixes it.
There is a real irony in how this played out. The bill’s voluntary design was its bipartisan strategy — the feature that produced the 417–3 House margin became the weapon Senate opponents used against it. The closer the bill got to unanimous, the less it actually required, and the easier it was to dismiss as inconsequential. A measure can be too bipartisan to survive: consensus on the principle masked a live disagreement about whether the federal government should compel it, and the filibuster gave that disagreement a veto.
The weakness critique also had a factual tailwind: states weren’t waiting. Roughly two dozen states were already redesigning large-load tariffs, special contracts, and exit fees on their own — and H.R. 9340 explicitly exempted states that had already acted from repeating the process. That made the federal “consider” standard look, to skeptics, like a mandate to contemplate what leading commissions had already done. A bill whose practical effect was concentrated in the states doing the least was always going to struggle to justify 60 votes.
The idea’s new home
On the very same day the Senate vote failed, four senators unveiled the vehicle that now carries the idea: the Bipartisan American Affordability and Jobs Act of 2026, a comprehensive permitting reform package negotiated by Sens. Shelley Moore Capito (R-WV) and Sheldon Whitehouse (D-RI) — the chair and ranking member of the Environment and Public Works Committee — and Sens. Mike Lee (R-UT) and Martin Heinrich (D-NM), who lead the Energy and Natural Resources Committee. The headline provisions speed federal permitting, strengthen FERC’s backstop transmission siting authority, and require grid-enhancing technologies in interconnection. Buried in the bill’s later sections are the data center cost provisions, and they go further than H.R. 9340 did.
This is vehicle logic, and it’s worth understanding as a market participant. A standalone bill needs 60 Senate votes on its own merits; a must-pass package on permitting and affordability carries passengers that could never clear the filibuster alone. The cost-causation idea went from the marquee to the engine room — less visible, harder to kill, attached to legislation both parties want for other reasons. Ideas that lose floor votes often win this way, folded into bills too big to fail.
What’s tougher this time
The new provisions upgrade the old bill in three ways, and the third is the one to watch. First, the threshold drops from 100 MW to 20 MW. H.R. 9340 caught only hyperscale campuses; the permitting bill’s language reaches far down the load stack — enterprise data centers, large commercial and industrial loads, anything with serious computational density. The net is roughly an order of magnitude wider in the number of covered facilities.
Second, suggestion becomes requirement. Rather than asking states to consider a standard, the bill directs FERC to revise its transmission pricing policy within 270 days of enactment so that covered loads pay their costs, establishes what analysts are calling a ratepayer protection floor that bars utilities from allocating incremental costs to other customers, and requires financial assurances before new facilities are built — with the large customer’s obligation surviving even if it later abandons the load. State authority to set separate rates and service conditions for large loads is explicitly preserved, but the federal baseline is no longer optional.
Third — and most consequential — the bill reaches beyond incremental costs to embedded costs. Incremental cost is the new substation built for your campus. Embedded cost is your share of the grid that already exists: the transmission network built over decades and paid for by everyone. Charging large loads for incremental upgrades is cost causation; charging them embedded costs is beneficiary-pays taken seriously — the principle that using the shared system means paying for the shared system. That single word, “embedded,” represents the largest philosophical shift in federal transmission cost allocation in a generation, and it will be the most contested provision if the bill moves.
What it means for the wholesale-to-retail flow
Return to the anatomy of the retail bill. Transmission is roughly 10–15% of what households pay, and today those costs are overwhelmingly socialized — recovered through postage-stamp or load-ratio-share rates spread across all customers in a zone. The permitting bill’s provisions would carve large computational loads out of that socialization. Follow the money: every dollar of transmission revenue requirement assigned directly to a data center is a dollar that doesn’t have to be recovered from residential and small commercial bills.
The magnitudes, though, will be decided in implementation, not in statute. FERC’s 270-day pricing-policy revision is where abstract principle becomes tariff language — which costs count as “associated with” a covered load, how embedded cost shares are calculated, what counts as adequate financial assurance. Anyone modeling retail rate impacts should watch that docket, not the bill text: the statute sets the direction, but the dollars get decided in the compliance filings.
One underappreciated provision deserves attention: the bill directs the Energy Information Administration to begin collecting electricity demand data from covered data center entities through Form EIA-861. That sounds like paperwork. It isn’t. You cannot allocate costs to loads you cannot see, and today the federal government has no systematic measurement of data center electricity demand — the fastest-growing load category on the grid is, statistically, a rumor. Durable cost-allocation policy requires durable data; the EIA provision builds the metering before the billing.
There is also a second-order effect on utility planning worth tracing. The financial-assurance requirement — money or guarantees posted before the utility builds — de-risks building ahead of load. Today, a utility asked to serve a speculative 500 MW campus faces an asymmetric bet: build now and risk stranded costs in the next rate case, or wait and watch the customer walk to a competing utility or behind-the-meter option. Assurances flip that calculus, which means faster interconnection for creditworthy loads and less regulatory lag in getting steel in the ground. Cost allocation, done right, isn’t just about fairness — it’s a construction accelerator.
What happens next
The honest answer is: nothing, until after the election. The Senate is scheduled to recess after this week until after the midterm elections, which means no Senate vote before November — and the House would still need to pass the bill after it returns. The idea is alive, tougher, and parked. In the meantime, the states keep moving on their own: commissions were already redesigning large-load tariffs before Congress acted, and the federal stalemate pushes more of the action back to them.
Watch three things. FERC’s posture — the 270-day clock starts at enactment, but the Commission’s early signals on cost allocation will shape expectations long before. State separate-rate proceedings, explicitly preserved by the bill, where the near-term dollars are actually being decided. And the 20 MW threshold, the provision most likely to be negotiated if the bill reaches conference — every megawatt of threshold is a lobbying battle over which facilities pay.
Finally, notice the electoral logic keeping this idea alive. “Make data centers pay their own way” is one of the rare energy positions with no partisan cost: Republicans get to defend households while championing the AI buildout, Democrats get to protect ratepayers while taxing Big Tech’s infrastructure. Both parties will claim the mantle through November, which is precisely why the idea survived its bill. When a principle is this cheap to endorse and this expensive to oppose, it doesn’t need 60 votes — it needs a vehicle.
The bill died. The idea didn’t — it got stronger, moved into bigger legislation, and is now waiting on an election. That’s the normal lifecycle of cost-allocation reform: the physics of who uses the grid never changes, the economics of who should pay are endlessly contested, and the politics eventually find a vehicle. September’s lesson isn’t that Congress rejected the principle. It’s that the principle needed a bigger bill.