IndexRetired Coal Plant Data Centers: Site Criteria for Brownfield Power Campuses
Retired coal sites can solve part of the power problem, but only when developers underwrite the brownfield risk directly.
Retired coal plant data centers are moving from niche idea to active development strategy. The reason is simple: large data centers need power, and retired power plant sites already sit inside the electrical geography that greenfield projects are trying to reach.
That does not make them easy sites. It makes them worth underwriting.
The U.S. Department of Energy lists redevelopment of retired coal power plant sites as one of the tools available to meet data center electricity demand, alongside grid infrastructure expansion, clean generation, demand flexibility and existing nuclear or hydropower infrastructure. DOE's Energy Infrastructure Reinvestment Program under Title 17 also supports projects that retool, repower, repurpose or replace energy infrastructure that has ceased operations.
For data center developers, the thesis is clear. A retired coal plant may offer hundreds or thousands of acres, an industrial land-use history, high-voltage interconnection assets, water infrastructure and local familiarity with heavy infrastructure. The risk is that every one of those advantages needs proof.
Why retired coal sites are attractive
A conventional data center site search often starts with land and works toward power. Retired coal sites invert the workflow. They start with power infrastructure and force the developer to test whether the rest of the site can carry a modern compute campus.
The attraction usually comes from six attributes:
existing transmission adjacency
switchyard or substation infrastructure
large industrial parcels
rail, road and utility corridors
permissive or precedent-based industrial zoning
communities with a history of energy-sector employment
That package can be powerful. Interconnection queues in major markets are long. JLL's 2026 outlook says average wait times for grid connections in primary data center markets exceed four years. A site with a credible path through existing electrical infrastructure can move ahead of a better-looking greenfield parcel.
But developers should be precise. A retired coal plant does not automatically come with usable capacity. Existing wires, breaker positions and interconnection rights may not match the new load shape. The grid may have changed since the plant stopped operating. A generation export interconnection is not the same as a load import service path.
That distinction decides whether the site is an asset or a trap.
Homer City shows the scale of the opportunity
Pennsylvania's Homer City site is the clearest public example. The Pennsylvania Department of Environmental Protection describes Homer City Generating Plant as a decommissioned 2 GW coal-fired power plant that ceased operations in 2023. The state says the site is planned to become a more than 3,200-acre natural gas-powered data center campus designed for AI and high-performance computing demand.
That is the model many developers are studying: convert legacy power infrastructure into a compute campus, then use new generation, transmission assets and industrial land to serve large loads.
The size is the point. AI-era data centers are no longer single-building industrial conversions. They are power campuses. A retired coal site can offer the acreage and utility context for that scale. It also brings demolition permits, groundwater monitoring, ash ponds, legacy contamination, stack removal, community scrutiny and environmental agency oversight.
If the underwriting ignores that, the project is not power-led. It is wishful thinking.
The site criteria that matter
The first criterion is switchyard reuse and import capacity. Developers need to verify voltage, breaker positions, protection schemes, ownership, transfer capability and whether the utility treats the project as a new large load or a continuation of an existing energy site. The answer should come from utility engagement, not a map.
The second is environmental liability. Coal sites require Phase I and Phase II work, ash pond assessment, groundwater data, demolition scope, asbestos and lead review, stormwater analysis and clear allocation of remediation responsibility. Brownfield incentives can help. They do not remove schedule risk.
The third is water and cooling. Many coal plants sit near meaningful water infrastructure, but withdrawal rights, discharge limits, community sensitivity and cooling technology still decide usability. For AI workloads, the site plan has to preserve optionality for air cooling, direct-to-chip liquid cooling or hybrid systems.
The fourth is gas, emissions and bridge power. Gas generation can solve schedule risk and introduce fuel, air quality, noise, emissions and future carbon exposure. A power solution that cannot be permitted is not a solution.
The fifth is demolition and constructability. Retired generation sites contain stacks, turbines, coal-handling systems, rail spurs, foundations, cooling structures and buried utilities. Demolition sequencing belongs on the critical path.
The sixth is community acceptance. A retired plant may be locally accepted as industrial, but a data center is still a new political event. The community case has to be specific: jobs, tax base, infrastructure reuse, remediation and power cost protection.
Where AI helps the underwriting
AI is useful on retired coal sites because the data room is fragmented. The relevant facts sit across utility filings, environmental permits, demolition records, tax parcels, grid maps, air permits, water rights, zoning records and local meeting minutes.
A good workflow uses AI for four tasks:
assemble the site evidence base
extract obligations from permits and consent orders
compare electrical assumptions against utility documents
surface conflicts between environmental, construction and energization schedules
Human experts still decide whether the site works. Engineers, environmental counsel, utility specialists and local operators carry the judgment. AI reduces the chance that a critical document is missed.
The test is not brownfield versus greenfield
Retired coal plant data centers are not automatically better than greenfield sites. They are different. They trade land assembly and power adjacency advantages for brownfield, permitting and community complexity.
The right question is not, 'does the site have power history?' The right question is, 'can that history be converted into a bankable power path for the next load?'
If the answer is yes, the site can work. If the answer is vague, it is a remediation project with a data center story attached.