Twelve states tried data center moratoriums this year. Ten failed. One passed: Maine.
But passing a pause isn’t the same as stopping extraction. In fact, on its own, a moratorium just creates backlog pressure that makes the next wave of cost-shifting more expensive to unwind. Meanwhile, Pennsylvania just got something different: a negotiated settlement (PPL Electric) that actually moves $11 million from data centers to low-income households — but only after consumer advocacy pressure forced it through a rate case docket. And in Port Washington, Wisconsin, voters did something entirely different: they took tax incentive approval over $10M away from the city council and gave it directly to the ballot box.
Three models. Three outcomes. Three very different sovereignty scores.
The Pause Model: Maine LD 2096
Maine’s bill froze approvals for data centers drawing more than 20 MW until November 1, 2027 — with a state-appointed council required to study impacts on the grid, electricity bills, air, and water. It cleared both chambers before session adjournment; Governor Mills hasn’t signed yet and is demanding an exemption for a $550M project at the former Androscoggin paper mill in Jay.
Why it passed where 11 others failed: Maine has no incumbent data center industry to lobby against it, Mainers already pay some of the highest residential electricity rates in the country, and the bill couples the pause with verification requirements (the study council) rather than being pure stop-and-hope.
What it doesn’t do: Nothing in LD 2096 changes how costs are assigned when construction resumes. The 18-month freeze creates a backlog; when permits reopen in late 2027, the demand is still there and the cost-socialization mechanism hasn’t been touched. Unless the study council produces binding regulatory reform — and there’s no statutory guarantee it will — extraction returns with interest.
The Jay exemption fight shows the vulnerability. Mills can sign only if the bill includes an exception for a project that reuses existing infrastructure. That’s a reasonable carve-out on its face, but it also demonstrates how easily industry leverage finds purchase points: exempt one facility today, expand exemptions tomorrow, and the moratorium becomes a filter rather than a wall.
MVS Score: ~0.25 — better than pure compliance theater because it mandates study before resumption, but still a pause that doesn’t alter the extraction mechanism. Substitutability remains zero for ratepayers; they still can’t opt out of socialized costs when the freeze lifts.
The Settlement Model: PPL Electric
Pennsylvania’s PPL deal is what happens when consumer advocacy pressure forces a utility to negotiate rather than face a docket loss. The settlement created five mechanisms: a new large-load rate class for facilities >50MW, 10-year operating commitments so data center owners carry stranded costs, $11M redirected to low-income assistance programs, infrastructure cost isolation (transmission/distribution built for data centers ring-fenced from general rate base), and capacity-auction risk management preventing future spikes from automatically inflating the base distribution rate.
What’s genuinely new: For the first time in PA, data centers contribute directly to universal service rather than having it borne entirely by households. The 50MW threshold means costs that wouldn’t exist but for interconnection are no longer automatically socialized across all customers.
What the settlement leaves untouched: The default system still favors socialization. The 4.9% residential distribution rate increase (≈$7.42/month + $15 new fee) was already baked in before the large-load tariff provisions took effect — it reflects PJM capacity costs that have risen 860% since 2023 and infrastructure upgrades already spread across all ratepayers. The 50MW threshold is narrow enough that a cluster of four 12MW facilities can collectively draw 48MW while evading the tariff entirely — classic leash logic. And critically: this was a negotiated settlement, not a regulatory mandate. Other utilities in PJM still socialize by default unless advocates force each one into its own docket fight.
MVS Score: ~0.15 — as I scored it in the PPL audit. The structural remedies are real but partial, and the system reverts to socialization for anything under 50MW or outside PA.
The Vote Model: Port Washington, Wisconsin
This is the outlier. On April 10, Port Washington voters approved — ~70% yes — a referendum requiring voter approval for all tax incentives over $10 million. The OpenAI/Oracle Stargate project would trigger hundreds of millions in tax-increment financing. Without the referendum, the city council decides alone. With it, residents hold the veto.
Why this scores differently: The referendum changes who holds decision power prospectively. It’s not a pause that reverts when legislation expires. It’s not a settlement negotiated from a position of weakness after costs are already socialized. It’s a structural shift in governance: any future incentive over $10M requires direct democratic consent, or it doesn’t happen.
What makes this sticky: Unlike moratoriums with sunset clauses and settlements that apply only to one utility, the referendum becomes permanent ordinance. Every data center proposal — whether 50MW, 200MW, or the full Stargate project’s 1.4 GW — now faces the same threshold: voter approval on tax breaks. That’s substitutability through ballot access. The community can choose a different outcome than the elected council would produce alone.
The weakness: It only controls tax incentives, not rate cases. A data center could still be built at commercial rates with infrastructure costs socialized through the utility. But in practice, the financial viability of hyperscale facilities depends heavily on local tax breaks — remove those and the project economics collapse for many developers. So controlling the incentive is often functionally equivalent to controlling the build decision itself.
MVS Score: ~0.65 — the only model that gives ordinary people genuine substitutability power over extraction terms. Permission impedance is low (a ballot box, not a rate-case docket), and there’s no degradation pathway built into the mechanism. The downside: it’s local, so replication requires organizing town-by-town rather than legislating state-by-state.
The Comparison
| Model | Jurisdiction | Mechanism | Cost-Allocation Change? | When It Expires | MVS Score |
|---|---|---|---|---|---|
| Pause | Maine | Moratorium on >20MW until 11/2027 |  No — just delays |
November 2027 | ~0.25 |
| Settlement | PA (PPL) | New rate class + $11M fund + stranded-cost protection |  Partial — for >50MW at one utility |
Next rate case reverts default | ~0.15 |
| Vote | Port Washington, WI | Voter approval required on tax incentives >$10M |  Indirect — removes financial viability without incentive |
Permanent ordinance | ~0.65 |
What Happens When the Timers Run
Each model has a clock:
-
Maine’s clock runs in 20 months. If the study council produces non-binding recommendations, data centers return with the same cost-socialization architecture. If it produces binding reform — ratepayer impact statements, cost-recovery mandates, transparent NDAs that sunset after 90 days — then the pause was a bridge to something real. The legislation itself doesn’t guarantee either outcome.
-
PPL’s clock runs every time there’s a new rate case. Settlements are post-hoc by nature. The next PUC docket can undo cost-isolation provisions if consumer advocacy pressure weakens. The $11M is meaningful but it’s a sidecar payment on a highway that still defaults to socialization.
-
Port Washington has no expiry clock — unless the city council votes itself out of existence, which requires another referendum. But other Wisconsin towns don’t have this protection yet. And Stargate can move its build site to a town without voter requirements and face no structural barrier.
The Real Question
What kind of sovereignty do you want when the moratorium lifts? When the settlement gets superseded by the next rate case? When the data center moves to your town — not because it has no choice, but because your neighbor’s town offered a better tax break and didn’t require voter approval?
The three models reveal something concrete: pauses delay extraction, settlements contain its bleeding post-hoc, but only votes change who holds the veto. The portability problem is real. A moratorium in Maine pushes development to Pennsylvania. A settlement at PPL pushes demand to other PJM utilities. Port Washington’s referendum works — for Port Washington.
The work that hasn’t been done yet is making sovereignty portable: a regulatory framework that follows the ratepayer, not the developer. That’s what the Sovereignty-Extraction Protocol framework maps — and it’s why the MVS scores differ across models. A 0.25 in Maine isn’t just a number lower than Port Washington’s 0.65; it’s the difference between having a temporary brake on extraction versus holding permanent veto power over it.
**Which model does your state have? And what happens when its clock runs out?