On October 22, the but-for fight moved from the Senate’s report into a live federal docket. American Transmission Company, the Wisconsin transmission owner, went public with pending Federal Energy Regulatory Commission filings that would assign the transmission costs of large-load data center projects to the developers themselves — and put numbers on what the alternative costs everyone else. If FERC approves the proposal, ATC says it will bill $138 million in 2027 for transmission costs associated with large-load data center projects, and its network rate will fall approximately 6 percent from the 2026 rate. Without the mechanisms, ATC says, the 2027 network rate would be 16 percent higher than the proposed level.
That asymmetry is the story: a 6 percent decrease for existing customers under cost-causer assignment, versus a 16 percent increase over that level if the costs are socialized. Those two figures show how much transmission spending is now being driven by a handful of very large loads, and who has been in line to pay for it. For readers of this desk, this is the transmission-side counterpart to the retail battles covered here — the Pennsylvania consumer advocate’s recommendation that but-for customers bear 100 percent of their upgrade costs, the Wisconsin disputes over whether data center loads will move retail rates. Now the same principle is being tested at FERC, where the ruling will echo in every regional transmission organization’s cost-allocation fight.
The two charges
ATC’s proposal has two complementary mechanisms, designed for different phases of a large-load project’s life.
The first operates during construction. Under the Construction Work in Progress charge — CWIP — the financing costs and associated taxes for transmission facilities built to serve qualifying large-load projects would ultimately be assigned to the large customer. This is the phase most cost-allocation regimes handle worst. A transmission project serving a data center campus takes years to build, and the capital costs accumulate long before a single megawatt of load shows up. Under traditional ratemaking, those costs sit in rate base and earn a return funded by existing customers while construction proceeds. CWIP flips that: the carrying costs of the build follow the customer that caused the build.
The second mechanism, the Minimum Transmission Charge — MTC — operates after facilities enter service. It would ensure that new large-load customers pay 100 percent of the costs incurred to build the facilities needed to serve them, applying when regular payments for network service do not reflect the full capacity the customer requested. The problem the MTC addresses is subtler than it sounds. Network transmission charges are paid through the load-serving entity’s load ratio share — a large load embedded inside a utility’s network service footprint increases the denominator, and everyone sharing the system shares the cost of the facilities. If the data center ramps slowly, or never reaches its contracted peak, the facilities were sized for demand that never arrived, and the shortfall is borne by other customers. The MTC is a floor under the cost-causer’s responsibility.
Both mechanisms apply to qualifying projects involving a new or expanded load of at least 100 megawatts or requiring at least $100 million in transmission facilities. And both are explicitly designed for the scenarios consumer advocates have been warning about: projects that are canceled, substantially modified, or simply never use as much electricity as originally projected.
Why the construction phase is the hinge
It is worth pausing on the CWIP charge, because it answers the question the Senate report left open. The report’s investigators found that all seven data center companies surveyed acknowledged they would pay some, but not all, of the upgrade costs attributable to their facilities, and that the companies opposed but-for standards for shared infrastructure. That posture is cheap to hold when the costs in question are still hypothetical. Construction-phase financing is where the money becomes real before the load exists — and where the stranded-cost risk is sharpest.
Consider the arithmetic ATC is implying. A transmission build serving a data center campus costs on the order of hundreds of millions of dollars and takes years. The financing costs on that capital, plus the associated taxes, are real money spent in the ground before the first server rack draws current. If the project is canceled mid-build, or the facility ramps to half its contracted capacity, someone has paid for wires sized for a load that does not exist. Traditional ratemaking’s answer is everyone. CWIP’s answer is the customer whose name is on the interconnection request.
That is the precise moral of the Oracle situation in Port Washington. The campus Vantage Data Centers is building for Oracle — 902 MW of computing load, 1.3 GW of total electricity demand — cannot operate until ATC builds the high-voltage interconnection, and ATC refiled its Public Service Commission application in September after regulators revoked the completeness finding. Every month of delay is a month in which capital sits in the ground earning returns funded by existing ratepayers — exactly the exposure CWIP is designed to assign elsewhere.
The 6 percent and the 16 percent
Now the figures, which deserve a careful reading because ATC is deploying them as an argument, not a neutral disclosure. The company projects its 2027 network rate roughly 6 percent below the 2026 rate with the proposal approved — ATC’s release says “approximately 6 percent”; local trade coverage on October 22 characterized it as 6.2 percent — and says the rate would be 16 percent higher than the proposed level without the large-load cost assignment. The $138 million is the amount ATC expects to bill in 2027 related to large-load data center projects, contingent on FERC approval.
Two cautions belong here. The first is that these are the company’s projections of its own filing, not an adjudicated outcome. FERC has not ruled; the figures assume approval of mechanisms that exist only as pending proposals. The second is that the 6 percent decrease is not purely a data center story. ATC also disclosed that over $100 million of its 2027 network rate reduction comes from costs being shared across the Midcontinent Independent System Operator region for Multi-Value Projects the company builds and owns, including the Grid Forward Project. Regional transmission that benefits the whole MISO footprint is still being socialized across the whole footprint — as it should be. The 6 percent is two stories braided: but-for assignment of data-center-driven costs, and regional sharing of regionally beneficial costs. Conflating them would overstate what cost-causer-pays is delivering.
What ATC is not assigning
That second story is worth dwelling on, because it is the honest part of the proposal — and the part that answers the strongest argument against but-for allocation. The data center industry’s objection, as presented to Senate investigators, is that upgrades triggered by a data center often provide a system benefit that should be shared among all ratepayers. ATC’s filing concedes the principle where it is true: the MISO Multi-Value Project costs are shared across the region because the projects serve the region. What the company is assigning to the cost-causer is the incremental, customer-driven portion — the facilities that would not have been built but for the large load.
This is the distinction this desk has been drawing since the but-for fight began. Dedicated, customer-specific infrastructure and but-for network upgrades belong to the customer that triggered them; facilities with genuine system-wide benefits belong in shared rates. ATC’s proposal is notable precisely because it draws that line in a FERC filing rather than a press release — with a 100 MW or $100 million threshold, a construction-phase mechanism, and an in-service floor. It is a complete architecture, not a slogan.
The Wisconsin retail counterpart
The proposal also lands in the middle of Wisconsin’s own retail reckoning. ATC has requested or received approval for between $2.63 billion and $2.98 billion in new transmission lines and substations tied to data centers, including pending projects serving Microsoft and Vantage facilities in southeastern Wisconsin and approved projects serving Microsoft’s Mount Pleasant campus and Meta’s Beaver Dam data center. Intervenors have argued that ATC’s typical cost-sharing structure would leave other ratepayers covering much of the initial construction — the Citizens Utility Board’s executive director, Tom Content, has said any tariff must pass a no-harm standard, and that “saddling customers with transmission costs does harm them.” We Energies, which will power the Microsoft and Vantage facilities, says its data center customers will pay $564 million over the next two years to fully cover their related transmission costs.
Read those together and the shape of the Wisconsin settlement is visible: the retail side assigns cost through the utility’s agreements with the developers, and the transmission side — ATC’s FERC filings — assigns cost through the tariff. Both are working toward the same no-harm standard. Neither is complete until FERC acts.
The national template
The reason this filing matters beyond Wisconsin is the calendar FERC set in June. On June 18, 2026, the Commission opened six individualized Section 206 proceedings — one each for PJM, SPP, NYISO, MISO, ISO New England, and CAISO — each preliminarily finding the operator’s tariff unjust, unreasonable, or unduly discriminatory for lacking adequate provisions for large loads. The orders direct each operator and its transmission owners to address five subjects, including protections against cost shifting with a pro forma cost recovery agreement. FERC deliberately chose individualized proceedings over a national rulemaking, citing speed and regional tailoring.
ATC is a MISO transmission owner, and its proposal is an early, fully specified answer to exactly the question FERC posed in the MISO docket: what do protections against cost shifting look like in a tariff? The other regions are still drafting — CAISO has moved to a proposed 50 MW threshold ahead of its November filing; NYISO’s stakeholder process runs toward a November 16 show-cause date — which makes ATC’s CWIP-plus-MTC architecture the first concrete model on the table. If FERC accepts it, with or without conditions, every other transmission owner in every other 206 proceeding will be measured against it.
It is worth being precise about what “expected” means here. FERC has not signaled how it will rule on ATC’s mechanisms, and the Commission’s recent record on large-load cost allocation is mixed — it rejected PJM’s Interim Resource Adequacy Service on October 9 while accepting the Large Load Registry, and it suspended PJM’s backstop procurement over cost-allocation defects. The 206 proceedings themselves reflect a Commission that has decided tariffs must change but has not decided how. ATC’s proposal is a candidate answer, not a preview of the ruling.
The steel argument, and its limits
State the counterargument fairly, because it will be made in the docket. A transmission line built to serve a data center cluster will typically remain in service for decades, carry power for other customers, and improve reliability for the surrounding system. Assigning 100 percent of its cost to the triggering customer can overstate that customer’s responsibility — and if the load does arrive as contracted, a minimum charge calibrated to requested capacity is a floor the customer was going to pay anyway. The MTC only bites when the customer’s actual use falls short of what it asked the system to build for. That is the design’s virtue: it prices the risk of the forecast, not the electricity.
The limit of that virtue is threshold design. A 100 MW or $100 million floor catches the hyperscale campuses and misses the mid-size facilities whose cumulative impact is real. Nor does a FERC-approved tariff answer the retail side’s harder questions: the minimum billing demands, the exit fees, the contract lengths that keep a departed customer from stranding costs. Those are state commission matters, and they are being litigated in parallel. ATC’s filing is the transmission half of a settlement whose other half is being written in dockets from Harrisburg to Madison.
What to watch
Four things will determine whether this filing is remembered as the template or a footnote. The first is FERC’s ruling on the mechanisms themselves — expected in the ordinary course of a Section 205 proceeding, but never certain, and conditioned or rejected as easily as accepted. The second is the November filing calendar in the 206 proceedings, where the other RTOs’ answers to the cost-shifting question will show whether ATC’s architecture gets copied or undercut. The third is the Wisconsin PSC dockets, where the refiled Port Washington interconnection application will test whether the state and federal tracks converge. The fourth is the copycat watch: the first transmission owner outside MISO to file a CWIP-style construction-phase charge will tell you whether cost-causer-pays has become the industry’s default opening bid — or whether ATC is negotiating alone.
The figures ATC chose to lead with — $138 million, down 6 percent with the proposal, up 16 percent without it — are a company making its case in public before the Commission rules. Treat them as an argument. But the argument’s structure is what matters: construction-phase costs assigned during construction, in-service costs floored at requested capacity, system-benefit costs still shared. Whoever pays for the grid the AI buildout requires, the bill is now itemized. FERC gets to decide whether the itemization stands.
Sources: American Transmission Company, “ATC projects potential 6% decrease in 2027 transmission network rates” (press release); ATC Large-Load Cost Allocation Business Practice (MTCA terms, atcllc.com); BizTimes, “ATC projects 6.2% rate decrease with proposed data center charges” (Oct. 22, 2026); Daily Energy Insider and T&D World coverage of the ATC proposal (Oct. 2026); WisBusiness MON AM News on Wisconsin data center transmission costs; FERC Section 206 large-load proceedings (June 18, 2026 orders, dockets EL26-67 through EL26-72); prior Fracmoo coverage: “The But-For Fight Goes to Congress” (Oct. 17, 2026), “The $4.5 Billion Quarter” (Oct. 19, 2026).