For two years, the fight over who pays for the grid’s data center buildout has been waged docket by docket: in capacity auctions, interconnection queues, state rate cases, and FERC orders about which loads get curtailed first. This week, Congress proposed to settle the question by statute. A bipartisan Senate bill would require data centers and other large computing loads to pay for both their share of the existing transmission system and the incremental cost of expanding it to serve them — two charges that federal policy has, since 1994, treated as alternatives.
The bill is the Bipartisan American Affordability and Jobs Act, known as BAAJA, introduced in the Senate on September 30 and referred to the Energy and Natural Resources Committee. It is a sweeping permitting-reform package — more than 400 pages — negotiated by Senate Environment and Public Works and Energy and Natural Resources committee leaders, and it is not law. It still needs Senate passage, House agreement, and a presidential signature, and a procedural vote is expected November 9. Its data center provisions, however, are the most direct legislative answer yet to the cost-allocation fight this desk has been covering since the beginning: who bears the cost of the wires that a new class of enormous customer requires.
The 1994 rule the bill would rewrite
To understand what the bill changes, start with the rule it replaces. Under a Federal Energy Regulatory Commission pricing policy dating to 1994, a transmission customer generally pays either the embedded cost of the existing system — its share of the wires already built — or the incremental cost of the expansion needed to serve it, whichever is higher. Charging both was, in effect, double-charging for the same capacity, and the policy forbade it.
That either-or framework was built for a world of slow, predictable load growth, in which a new customer’s incremental expansion was modest and the existing system was shared broadly. It fits poorly with a 500-megawatt campus that requires a new substation, reconductored lines, and upgrades deep into the regional network. The bill’s transmission-pricing provision, Section 2107(a), would direct FERC to revise that policy so that computational loads — data centers, cryptocurrency miners, and other high-density computing facilities — larger than 20 megawatts pay both: embedded costs for the system they use, plus incremental costs for the expansion they cause. Incremental payments would be credited against the utility’s transmission revenue requirement, so the same dollars are not also collected from other customers.
The bill pairs the pricing change with several companion protections. FERC would have to offer a lower rate to computational loads that choose non-firm transmission service — the curtailable option this desk has covered as flexible interconnection — giving loads that can accept interruptions during grid stress an incentive to do so. States could place covered loads in a separate service class, with rates, terms, or service priorities that differ from, and may be less favorable than, those for other large commercial and industrial customers. And utilities would hold covered loads liable for capital costs even if they exit their agreements early, so ratepayers are not left holding the cost of stranded infrastructure. The provisions would generally apply to loads that interconnect after enactment; previously approved arrangements would remain in effect.
The $55 billion estimate, and what it assumes
Grid Strategies, a power-sector consulting firm, published a white paper on November 2 estimating what the federal pricing change alone would collect. Its headline figures: about $5.53 billion a year in transmission charges from data centers, or roughly $55 billion over ten years, compared with roughly $2 billion a year that data centers currently pay. Of the $55 billion, about $35 billion would be embedded costs and about $20 billion incremental costs. The white paper was developed with support from Arnold Ventures.
The geography of the estimate is concentrated. PJM and MISO, the two grid operators whose regions hold the largest data center pipelines, would account for about 82 percent of the projected annual charges — about $2.31 billion in PJM and about $2.25 billion in MISO. Grid Strategies projects 13.4 gigawatts of new data center capacity in PJM and 16.7 gigawatts in MISO. PJM’s higher charges, despite its smaller projected pipeline, reflect the region’s higher transmission rates, according to Grid Strategies president Rob Gramlich.
The estimate rests on assumptions that matter as much as the headline. It assumes 40.3 gigawatts of new data center capacity outside Texas would be subject to the policy, drawn from Enverus’ third-quarter 2026 high-confidence project tracker. It assumes new data centers take firm transmission service; developers that build entirely off-grid would generate none of these charges. It excludes Texas’ ERCOT market, where FERC does not traditionally regulate transmission service. And its incremental-cost estimates are built largely on MISO data: 163 approved transmission projects representing about $11.4 billion in investment to connect 43.5 gigawatts of large loads, from which the analysts calculated an average incremental cost per megawatt and applied it across the regions studied. Grid Strategies estimates the projected revenue could support nearly 10,000 miles of new 765-kilovolt transmission lines over ten years — a construction program that would depend, as the firm notes, on when and where data centers actually connect.
Every one of those assumptions is a live question. A forecast of 40.3 gigawatts of high-confidence capacity is not a commitment, and this desk has documented how much of the forecasted data center wave evaporates: PJM’s own market monitor calculated that more than 4,000 megawatts of forecast data center load in that region was canceled or delayed as of September. If the load does not arrive, the charges do not either — and transmission built in anticipation of it becomes the stranded-cost problem the bill is trying to prevent. Actual collections would also depend on FERC’s final rules, on applicable tariffs, and on the infrastructure requirements of individual projects, none of which exist yet.
Why the industry is, so far, not fighting it
The most striking feature of the bill’s reception is how little resistance it has drawn from the companies that would pay. At least two industry groups representing companies including Amazon, Google, Meta, and Microsoft have endorsed the legislation: the Corporate Energy Buyers Association and the Electricity Customer Alliance. CEBA called the bill a “generational opportunity” for faster permitting and improved transmission planning. Jeff Dennis, executive director of the Electricity Customer Alliance, said data center developers have already said they expect to pay their own costs, and then some. “They are willing to pay more than their own costs to make sure other customers are treated fairly,” he said.
The Data Center Coalition, by contrast, has not endorsed the bill and said it is still reviewing the legislation with its members. Amazon, Google, and Meta did not respond to requests for comment; Microsoft declined to comment.
There is a reason the hyperscalers might accept a larger transmission bill. The same package that charges them more also promises to make it easier to build: expanded FERC backstop siting authority for interstate lines, requirements for regional and interregional planning with benefits-based cost allocation, and faster permitting for the generation and transmission infrastructure their campuses need. For a developer whose binding constraint is the five-to-ten-year timeline for a new line rather than the transmission rate, a higher charge paired with a faster build is a trade worth taking.
The off-grid escape hatch
The estimate’s firm-service assumption points to the bill’s central tension: the harder grid-connected service becomes, the more attractive it is to leave the grid altogether. Grid Strategies’ Gramlich said he does not expect most data centers to abandon grid connections, because reliable power from an entirely off-grid system generally costs more and presents greater operational challenges, and he expects behind-the-meter generation to supplement utility service rather than replace it. But some projects are already going that route, mainly powered by gas turbines and engines with backup diesel generators, and the Electricity Customer Alliance’s Dennis warned that a final bill must ensure data centers are not forced to shoulder so much cost that they decide to build off-grid. “That could pose negative impacts to customers and removes a lot of investment in the grid we all rely on,” he said.
The off-grid question matters because it determines whether the $55 billion is a transfer or a mirage. A developer that builds its own generation pays no transmission charges at all under the bill’s framework — but it also contributes nothing to the shared grid, and its departure removes exactly the deep-pocketed customer whose payments were supposed to subsidize everyone else’s wires. The bill’s answer is to make the grid-connected option more valuable, not merely more expensive: faster lines, planned capacity, and a lower non-firm rate for flexible loads. Whether that combination holds developers on the system is a question the FERC rulemaking will have to answer, and Nidhi Thakar, CEBA’s senior vice president for policy, acknowledged the stakes: “A lot of this will have to get worked out in a FERC rulemaking. That is where the deep details on how to allocate costs for new transmission lines will occur.”
The second, larger number
Section 2107(b) of the bill would go further. It would establish a Ratepayer Protection Floor requiring states to address the costs large computing facilities impose on generation, storage, transmission, and distribution infrastructure. Grid Strategies estimates that provision could result in another $136 billion in incremental generation costs recovered from data centers over ten years if states fully implement the policy. That figure is potential cost exposure under the researchers’ assumptions, not a forecast of actual collections or a verified increase in total electricity-system costs — and it depends entirely on state implementation, which the bill invites but does not itself complete.
The distinction matters. The $55 billion transmission figure flows from a federal pricing change FERC would be directed to make; the $136 billion generation figure depends entirely on state implementation, which the bill invites but does not itself complete.
What the bill does not solve
Congress can change who pays for the transmission line. It cannot legislate the data center’s demand into existence, as one analyst put it — and that is the gap the bill leaves open. The pricing change addresses cost allocation, not load credibility. It does nothing, by itself, to verify that a queued project is real before a utility commits billions on its behalf, or to ensure that a developer’s financial security covers the utility’s full exposure if the project slips or dies. Those protections — binding capacity commitments, deposits and collateral tied to actual exposure, minimum-billing terms, early-termination liability — exist in the bill in partial form, through the early-exit liability provision, but the deeper discipline of tying major spending decisions to verified milestones remains a utility and regulator task.
That is the same lesson this desk drew from the phantom-load reckoning: a planning system that treats every queued reservation as a firm commitment will systematically overbuild, and overbuilding is not free. The bill’s pricing reform makes the overbuild cheaper for everyone except the data center — which is progress for ratepayers — but it does not make the forecast more accurate.
There is also the over-collection question Gramlich himself flagged: the bill errs on the side of over-collection, meaning data centers could pay more than their impact on the system and effectively subsidize other customers. For ratepayer advocates, that is a feature; for developers, a cost to weigh against the permitting gains; for FERC, a design problem for the rulemaking.
What to watch
Four things will determine whether the proposal becomes more than a negotiating position. The first is the Senate vote expected November 9, and whether the bill commands the 60 votes it needs to advance; negotiators are reportedly working on a managers’ amendment, though it is unclear whether it would revise the ratepayer protection section. The second is the House, which passed its own permitting bill, the SPEED Act, in December 2025 — the two chambers would still need to reconcile their versions. The third is the FERC rulemaking the bill would trigger, where the industry’s current equanimity will be tested against actual rates. The fourth is the states: whether they use the Ratepayer Protection Floor aggressively, and whether their separate service classes for computing loads survive legal and political challenge.
The deeper question is whether Congress has found the durable answer to the who-pays fight, or merely moved it to a new forum. For two years, the fight has been litigated in FERC dockets and state rate cases, where the question was whether a particular tariff or procurement was just and reasonable. The bill would answer it in statute: computational loads above 20 megawatts pay for the system they use and the expansion they cause. If it becomes law, the argument shifts from whether data centers should pay to how much, and by what formula — which is exactly the argument FERC rulemakings are built to have. If it does not, the dockets continue, and the $55 billion remains what it is today: an estimate of what a different cost-allocation rule would be worth, attached to a bill that has not yet passed.