Pennsylvania Puts Conditions on Data-Center Growth
Yesterday’s clearest shift was not a new campus announcement but a new condition for building one. Pennsylvania moved to make local approval, power responsibility, water safeguards, and public disclosure part of the path to state permission for proposed data centers.
That does not amount to a national turn against data-center construction. But alongside Texas’s continuing review of large new grid connections, it shows that the hardest questions are increasingly being asked before a project can convert announced capacity into operating load.
Pennsylvania Gov. Josh Shapiro signed Executive Order 2026-05, according to WGAL, applying the state’s GRID requirements to proposed data centers statewide. The order removes projects from fast-track permitting, requires local approvals before state permits, assigns developers responsibility for energy-infrastructure costs, and adds requirements around water conservation, transparency, workforce development, and community engagement. The practical reach will depend on agency implementation and local decisions, but the direction is clear: the state is moving beyond investment promotion toward enforceable project conditions.
Texas remained the other major constraint. The pause on new data-center connections while ERCOT reviews proposed load has not been resolved, and its scale is striking: roughly 474GW of requested capacity sits against a recent Texas peak demand of about 91GW. Those queue figures are not a construction forecast—many proposals may never become viable—but they explain why grid access is becoming a screening process. Senate Bill 6’s provisions on cost sharing and emergency curtailment further put reliability and financial responsibility at the center of large-load development.
The financing picture for OpenAI’s proposed 10GW Ohio campus became less settled than its headline scale suggests. Nvidia is reportedly considering an investment of up to $3 billion in SB Energy, while credit support, Nvidia’s guarantee, and OpenAI’s lease terms are being revised. Recent reporting had pointed to a more developed financing framework; yesterday’s accounts made clear that the allocation of risk among tenant, developer, equipment provider, and lenders remains under negotiation. No investment, lease, financing close, or power-delivery schedule has yet been confirmed.
A Wisconsin court ruling added a different form of pressure: disclosure. Midwest Environmental Advocates reported that a Dane County judge ordered the state Public Service Commission to release previously redacted load projections for Meta’s Beaver Dam data center, rejecting a trade-secret rationale. The contents of those projections and any appeal remain unknown, but access to the estimates could sharpen scrutiny of utility planning, generation needs, and customer cost exposure.
Key Points
- The policy center of gravity is shifting from whether communities welcome data centers to what developers must prove before they proceed. Pennsylvania’s order expands this logic to statewide permitting; Texas applies it at the interconnection stage. Neither establishes a uniform U.S. regime, but together they reinforce a jurisdiction-by-jurisdiction move toward developer-funded infrastructure, local accountability, and more explicit reliability safeguards.
- For very large AI campuses, capital structure is now as consequential as land control or an announced power target. The Ohio discussions illustrate why: a 10GW plan needs credible leases, credit support, generation arrangements, and grid delivery—not simply a named tenant or a prospective investor. Announced capacity should increasingly be read as an ambition until those commitments are executable.
- Transparency is becoming an operational issue rather than a peripheral political demand. The Meta disclosure ruling could provide concrete load information to regulators and local stakeholders, while TBC Partners’ request for an Environmental Impact Assessment screening decision for a proposed 300MW Cambois campus in Northumberland shows environmental review beginning early in the siting process. The UK proposal remains preliminary, not an approval or construction commitment.
Implications
Developers in states adopting stricter conditions may need to sequence local approvals, power plans, water strategies, and community commitments earlier in project development. That can lengthen preconstruction work, but it may also make the eventual allocation of grid and environmental costs more visible before utilities and ratepayers are exposed.
The widening gap between proposed gigawatts and executable projects is likely to become more important for investors, utilities, and host communities. Texas’s queue review and Ohio’s financing revisions point to the same practical reality: interconnection rights and committed capital are the scarce assets, not announced demand.
Watchpoints
Watch
How Pennsylvania agencies define and enforce the order’s local-approval, power-cost, water, and disclosure provisions—and whether developers challenge its legal reach.
Watch
The timetable and project-level results of Texas’s ERCOT interconnection audit, including which proposed loads clear review and on what conditions.
Watch
Whether Nvidia, SB Energy, and OpenAI execute investment, credit-support, lease, generation, and grid arrangements for the Ohio campus.
Watch
What Meta’s Beaver Dam load projections reveal once released, and whether the disclosure prompts further utility or regulatory proceedings.
Watch
Northumberland County Council’s Environmental Impact Assessment screening decision for Cambois, as well as the August 26 public forum in Bloomington-Normal while both Illinois communities evaluate rules under hyperscale-data-center moratoriums.
Fallout
Data-center growth is increasingly being conditioned by demonstrable power access, financial security, local authorization, and public accountability rather than by capacity announcements alone.
Final Thought
The data-center race has not stopped being a race for scale. It is increasingly a race to show who will carry the cost, risk, and local consequences of that scale.
