AI Policy Wiki
Dashboard

Data centers are coming for rural America — Abigail Bassett (The Verge, May 13 2026)

high confidence · updated 2026-06-06

The Verge long-form feature reporting from Jay, Maine on the proposed Sentinel Data Centers conversion of the Androscoggin paper mill and the state-legislative pushback that Maine Gov. Janet Mills vetoed. Anchors the empirical case that data centers' promised long-term jobs may not materialize — Ball State economist Michael Hicks's causal study of 254 Texas counties found net long-term data-center job creation ≈ zero. 67% of planned US data centers are headed to rural areas; 39% to counties currently without any. The 'neocloud' specialization (>100kW/rack, immersion or direct-to-chip cooling) is named here as a distinct facility class.

A long-form feature by Abigail Bassett, published in The Verge on 2026-05-13, reporting from Jay, Maine, combined with national framing on data-center siting. The piece centers on a proposal to convert the Androscoggin paper mill into a data center and on a Maine moratorium bill that passed both legislative chambers before being vetoed by Gov. Janet Mills.

Summary of argument

The article reports that the standard pitch made by data-center developers — that tax breaks granted in exchange for hosting a facility will be repaid in local jobs — is, by the causal evidence it cites, empirically thin. It argues that data centers produce close to zero net long-term local employment while imposing durable land, water, and energy costs on rural communities that often lack the governance capacity to evaluate the tradeoff. Bassett frames the proposed conversion of the Androscoggin Mill in Jay, Maine, as a concrete instance of this dynamic.

The Jay, Maine case

The Androscoggin Mill in Jay was originally proposed for sale to Godfrey Forest Products, an oriented-strand-board maker, a deal projected to create roughly 150 jobs; that sale was killed when federal tariffs disrupted its financing, prompting a pivot toward a data center. Tony McDonald, of mill owner JGT2 Redevelopment, characterized many of the bidders that followed as "data center cowboys" — fronts claiming hyperscaler clients but lacking the financial backing they advertised. McDonald eventually partnered with Sentinel Data Centers, a New York-based developer specializing in healthcare, financial, and hyperscale facilities. Sentinel Data Centers did not respond to requests for comment.

McDonald did not inform the Jay Select Board of the data-center plan until late February 2026, days before a state moratorium bill was scheduled for a floor vote. The Select Board voted 4-0 in favor of the project in March. The state moratorium passed on April 14 and was vetoed on April 24, 2026.

Bassett reports three factors that help explain why Maine became an early target: a 54% renewable energy mix (8th-highest in the US), relatively cool year-round temperatures, and lax land-use statutes.

Maine moratorium bill and the veto

The Maine moratorium bill (LD 713) would have paused permits and construction for proposed facilities larger than 20 MW for 18 months, opening a study window. It passed the Maine House and Senate and was vetoed by Gov. Mills on April 24, 2026, with Mills citing 125–150 jobs at the proposed Jay facility; that jobs figure is the developer's claim and underlies the veto rationale. The bill's sponsor, Rep. Melanie Sachs (D-Freeport), who chairs the House Energy, Utilities and Technology Committee, frames it as a planning bill rather than a ban. Sachs is quoted: "Even if it's 30 jobs, that means a lot to Jay, then, okay, but you've swept away protections for 1.4 million Mainers for 30 jobs." Had it taken effect, it would have been the first statewide US data-center moratorium.

A separate Maine law passed alongside the moratorium excludes data centers from some state-level tax breaks but leaves the door open for local municipalities to offer their own incentives. Bassett's structural point is that, with the state-level shield removed, small towns hoping for tax windfalls become the targets of incentive bargaining.

Jobs evidence

Michael Hicks, of the Ball State Center for Business and Economic Research, conducted a causal analysis of 254 Texas counties and found that data-center openings produced effectively zero net long-term employment. Hicks is quoted: "Whatever long-term jobs existed at data centers were being offset by losses elsewhere in the same sector." The article reports this result as consistent with the structural fact that operational data centers employ "only a few dozen staff members," while construction creates temporary, often non-local employment. Bassett notes that, if the finding holds under replication, it would change the political economy of data-center bargaining, with communities trading durable externalities for transient construction jobs and offset employment losses. The study is single-state (Texas) and was published on Substack rather than peer-reviewed.

Siting patterns and incentives

The article cites the Pew Research Center (April 2026) finding that 67% of planned US data centers are headed to rural areas and 39% to counties currently without any. Per NCSL data, more than 35 states offer incentives or tax breaks to attract data centers.

"Neocloud" as a distinct facility class

Bassett names "neocloud" as a distinct data-center subcategory, quoting: "If a neocloud data center is going into the old mill, it will require more than 100kW of energy per rack…will need either direct-to-chip or immersion cooling, both of which require ample space and water resources." A neocloud facility is a specialized GPU-compute facility for AI/ML workloads, typically rented to AI labs and enterprise customers, distinct from a hyperscale cloud data center (built by AWS, Azure, or GCP for general compute) in chip density, cooling, and customer profile. CoreWeave, Crusoe, Applied Digital, Lambda, and Together are named neocloud companies.

Relationships

Confidence and caveats

  • The piece is a Verge feature; Sentinel Data Centers did not respond to requests for comment.
  • Hicks's study is single-state (Texas) and Substack-published rather than peer-reviewed; it should not be over-weighted without replication.
  • McDonald's "data center cowboys" framing is one developer's characterization, useful as ground truth about market signal noise but not a structured industry observation.
  • The 125–150-jobs figure for Jay is the developer's claim and underlies Mills's veto rationale.

The following claims carry the noted confidence levels: 67% of planned US data centers headed to rural areas and 39% to counties currently without any (Pew, April 2026) — high; the Maine moratorium would have been the first statewide US data-center moratorium — high; Hicks's Texas study finding net long-term data-center job creation ≈ zero across 254 counties — medium, a novel finding pending replication; more than 35 states offer data-center incentives (NCSL) — high; Maine's 54% renewable energy mix, 8th-highest in the US — high.