Nuclear-Adjacent Data Center Sites: A Developer's Guide to Power-Certain Site Selection
Why proximity to operating or restarting nuclear plants has become an explicit acquisition criterion for data center developers in 2026.
Data center site selection used to start with land. In 2026, it starts with power. And for developers underwriting projects where a single energization slip can cost tens of millions in delayed revenue, nuclear adjacency has moved from a nice-to-have to an explicit screening criterion.
AWS is exploring a campus adjacent to the Calvert Cliffs Nuclear Power Plant in Maryland, where 1.7 GW of Constellation Energy capacity sits within interconnection reach. Microsoft's deal at Three Mile Island in Pennsylvania set the pattern. The logic is consistent: nuclear delivers what data centers actually need -- dispatchable, carbon-free baseload with capacity factors above 90%, on timelines that do not depend on greenfield transmission development.
FERC's December 2025 colocation ruling created the regulatory framework that makes this viable at scale. The order gives data center developers three new transmission service options and establishes clear rules for behind-the-meter generation arrangements at power plant sites. Before the ruling, nuclear colocation operated in an ambiguous regulatory space that lenders would not finance. Now it has a defined pathway.
Why Nuclear Adjacency Commands a Premium
The two-tier land market is real and measurable. Sites within 20 miles of operating or restarting nuclear plants with clear interconnection paths are pricing at a lease rate premium of 15-25% over comparably located grid-dependent sites. The reason is power certainty, and power certainty is what separates financeable data center projects from ones that stall at construction lending.
Lenders evaluating construction loans increasingly require demonstrated power access before committing. A site with a signed utility service agreement tied to nuclear output satisfies that requirement in a way that an interconnection queue position with a 3-4 year delivery estimate does not.
For hyperscalers, the calculation is the same. The big four -- Amazon, Google, Meta, Microsoft -- have each signed PPAs with utilities that own and operate nuclear plants. Combined, those arrangements target approximately 6.9 GW of nuclear energy capacity by the early 2030s (Carnegie Endowment, June 2026). Independent developers building capacity that hyperscalers will ultimately lease need to understand the sites those tenants will prefer.
What Nuclear-Adjacent Site Screening Actually Covers
Nuclear adjacency is not a single variable. Developers should evaluate five distinct dimensions.
Operating license status. Existing plants with remaining license terms represent near-term power certainty. Plants under license renewal review carry moderate timeline risk. Restarts (Palisades, TMI Unit 1) require separate underwriting around restart timelines, which range from 2-5 years depending on NRC review status, physical readiness, and workforce reconstitution.
Interconnection path. Adjacent land does not mean connected land. The critical question is whether a nuclear-adjacent site can reach the plant's output without crossing a constrained transmission segment. Sites on the same transmission path as nuclear output avoid the interconnection queue entirely in behind-the-meter colocation structures. Sites that require new transmission connection through congested regional networks lose most of the timing advantage.
Capacity headroom at the generator. Nuclear plants operating near capacity have limited incremental power available for direct service. Developers need to verify what MW the generator can commit contractually above its existing utility obligations. For PPA structures (rather than direct colocation), this translates to understanding the utility's ability to contract nuclear output to a specific load rather than into the general wholesale market.
State regulatory jurisdiction. Retail energy sales remain under state jurisdiction regardless of FERC's colocation ruling. States where the utility is investor-owned and operates under cost-of-service regulation present different commercial structures than restructured competitive markets. In PJM territory, tariff revisions due through early 2026 are establishing specific rates, terms, and technical requirements that define what colocation arrangements actually cost.
Cooling water risk. Nuclear plants are often sited near large water bodies for cooling. That proximity benefits data center cooling design but introduces shared water risk. In drought-stressed regions, a nuclear plant reducing output due to elevated river temperatures creates a power availability risk that is correlated with cooling difficulty. Screen this before treating a nuclear-adjacent site as fully de-risked.
The FERC Colocation Framework in Practice
The December 2025 FERC ruling created three transmission service options for data centers co-located with generation:
Behind-the-meter service -- the load takes power directly from the generator without flowing through the grid, subject to requirements on excess power treatment when generator output exceeds load
Network integration transmission service -- the load participates as a normal network customer but with contractual priority toward the adjacent generator's output
Point-to-point service -- a direct contractual path between the generator and the load with defined capacity reservation
For data center developers, Option 1 (BTM) provides the highest power certainty but requires physical proximity and NRC safety review if the arrangement involves nuclear-plant-side infrastructure. Option 2 is more accessible for sites within the transmission zone but does not guarantee power delivery when the grid is constrained. Option 3 is the structure most lenders prefer because it creates a defined, bankable power delivery obligation.
Existing BTM arrangements have until December 2028 to come into compliance with the new rules. Developers structuring new arrangements should design for compliance from day one.
AI's Role in Nuclear-Adjacent Screening
The analytical complexity of nuclear-adjacent site screening makes it a strong candidate for AI-assisted diligence. What AI can handle:
Layering transmission topology, interconnection queue position, and NRC license data across a candidate site list to identify where nuclear-adjacent power access is technically actionable
Screening NRC public filings for license renewal status, restart applications, and operating history
Modeling transmission path congestion between candidate sites and nuclear plant interconnection points using FERC eQueueFull data
Cross-referencing cooling water source with USGS drought monitor and nuclear plant historical output reduction events
What requires human judgment and direct relationships:
PPA negotiation with utilities and nuclear plant operators, which depends on counterparty relationships and commercial positioning
Assessment of restart timeline realism, which requires engineering judgment about physical plant condition and workforce availability
Community and local political dynamics around nuclear plant sites, which vary significantly and cannot be modeled from public data alone
Structuring BTM arrangements to satisfy both FERC requirements and NRC safety review standards
Developer Action Items
Nuclear-adjacent site acquisition is a longer-horizon play than conventional data center land banking. The sites that become available near operating plants are limited, and competition from hyperscalers with dedicated capital is real.
Monitor NRC filings. The NRC's license application pipeline reveals future reactor sites 3-5 years before construction begins. Early movers have time to acquire adjacent land before developer competition drives premiums above current 15-25% levels.
Track utility IRPs and rate filings. Utilities with large nuclear fleets are filing integrated resource plans that signal how they intend to serve large loads over 10-20 year horizons. Understanding those plans is essential for evaluating whether a nuclear-adjacent site can actually be served at the scale a modern data center campus requires.
Underwrite with power structure transparency. Do not treat nuclear adjacency as a single power certainty score. The structure -- BTM, PPA, utility service agreement -- determines bankability. Get the structure to a term sheet before underwriting at a premium.