New Reactor Siting Rules Are About to Reshape Where Data Centers Get Built

The energy problem facing hyperscale data centers isn't a secret anymore. Microsoft, Google, Amazon, and a growing list of colocations are burning through grid capacity faster than utilities can add it. Interconnection queues stretch years deep. Diesel backup generators aren't a solution — they're an embarrassment. And renewables, for all their cost advantages, still can't deliver the 24/7 baseload power that a 100-megawatt data center actually needs to run.

This is why a quiet regulatory shift in reactor siting deserves a lot more attention than it's getting.

New siting policies now allow small modular reactors (SMRs) and advanced reactor designs to be located adjacent to data center campuses and industrial parks — not just at utility-scale distances from population centers with years of site approval bureaucracy attached. This isn't an incremental update to nuclear permitting. It's a structural change in where nuclear power can physically exist in the built environment.

What the Siting Rules Actually Changed

Traditional nuclear siting requirements were built around light-water reactors the size of stadiums. The exclusion zones, emergency planning zones, and seismic requirements that governed them made sense for gigawatt-class plants designed in the 1960s and '70s. Smaller, passively safe reactor designs don't carry the same risk profile — but for years, they were still subject to substantially the same siting framework.

The updated approach creates pathways for advanced reactors to be evaluated on their actual safety characteristics rather than legacy assumptions. That means reactors with passive cooling systems, smaller thermal footprints, and inherently limited meltdown scenarios can now be sited much closer to end-users — including commercial and industrial facilities.

For infrastructure developers, this changes the fundamental calculus of project siting: instead of building transmission infrastructure to carry power from a distant reactor to a load, you build the reactor where the load already is.

The practical implication is significant. A 50-100 MW SMR co-located with a hyperscale data center campus eliminates the need for dedicated high-voltage transmission lines, reduces line losses, and gives the data center operator something they desperately want: direct contractual control over a dispatchable power source.

The Proximity Advantage Is Bigger Than It Looks

Distance in energy infrastructure is expensive in ways that don't show up cleanly on a single line item. There's the cost of transmission construction itself — which can run $1-3 million per mile for high-voltage lines depending on terrain and permitting. There's line loss, typically 5-8% over long distances. There's the interconnection study process, which now averages over four years in most ISO/RTO queues. And there's curtailment risk, which affects how financeable a project is.

Co-locating a reactor with a data center campus compresses all of those costs and delays simultaneously.

From a developer's perspective, the energy supply chain becomes dramatically simpler. A behind-the-meter nuclear arrangement — where the reactor feeds directly into the campus without touching the utility grid — sidesteps interconnection entirely. That's not a minor operational detail. In markets where interconnection timelines are measured in years and queue positions are bought and sold like commodities, bypassing the queue is worth more than most people realize.

There are also load-matching benefits that cut in the other direction. Data centers, particularly those running large AI inference workloads, have relatively stable baseline loads with predictable ramp patterns. That's an excellent match for nuclear generation, which runs best at steady output. Unlike pairing nuclear with a variable industrial load, a data center actually makes the reactor's economics cleaner.

Cost savings compound over the long run. Power purchase agreements for co-located nuclear generation, once the regulatory pathway is established, should price below what a data center pays for premium firm power from the grid — especially as carbon costs and grid congestion charges continue to rise.

Where This Is Already Happening

The co-location model isn't purely theoretical. Constellation Energy's deal with Microsoft to restart Three Mile Island Unit 1 — an 835 MW plant dormant since 2019 — was structured specifically to serve Microsoft's data center power needs in the PJM region. That deal, announced in 2023, put an immediate price signal on nuclear-to-data-center power: reportedly in the range of $100/MWh on a long-term basis, which Microsoft judged worth paying for clean, firm power.

That's a large legacy plant, not an SMR. But it established the commercial logic. If a hyperscale operator will pay a premium for nuclear power from a distant plant, the economics of a co-located reactor — with lower transmission costs and no interconnection exposure — look even more attractive.

On the SMR side, projects like TerraPower's Natrium reactor in Wyoming and X-energy's work with Dow Chemical at an industrial facility in Texas are early demonstrations of the industrial park co-location model. The Dow partnership is particularly instructive: a chemical manufacturer with a large, consistent thermal and electrical load has the same fundamental compatibility with nuclear generation that a data center does. Steady load, long-term horizon, tolerance for upfront capital in exchange for operational certainty.

Lessons from these early projects are still being written. Permitting timelines remain longer than proponents would like. First-of-a-kind costs are elevated. But the supply chain is developing, and the NRC's new licensing pathways — including Part 53, the framework for advanced reactor licensing — are beginning to give developers something they can actually work with.

What the Next Decade Looks Like

The convergence of reactor siting flexibility, data center energy demand, and clean energy mandates sets up a decade of serious infrastructure activity — but it won't unfold uniformly.

The near-term opportunity is in large industrial sites with existing power infrastructure, experienced development teams, and anchor tenants who can backstop long-term power purchase agreements. Greenfield data center campuses being planned today — many of them in the 500 MW to 1 GW range — are already incorporating nuclear as a potential power source in their 10-year energy roadmaps. That's not aspirational; that's procurement planning.

The developers who move fastest to understand the new siting framework, secure sites with appropriate characteristics, and structure long-term offtake agreements with creditworthy counterparties will have a meaningful first-mover advantage in a market that doesn't forgive slow movers.

On the technology side, watch for liquid metal-cooled and molten salt reactor designs to gain regulatory traction as their safety cases become better documented. These designs offer even tighter siting envelopes than early SMR concepts and could, in theory, be deployed inside heavily developed industrial zones where current SMR siting is still impractical.

Grid operators are also paying attention. Several ISOs have started discussing how co-located nuclear generation should be treated in capacity markets — whether it counts toward a data center's capacity obligations, how it interacts with interconnection rules, and whether it creates any reliability obligations. How those conversations resolve will have significant consequences for project economics.

One underappreciated factor: states matter enormously here. States with nuclear-friendly regulatory postures — Wyoming, Texas, Georgia, Virginia — are likely to move faster on both permitting and supportive legislation. Virginia is particularly interesting given its position as the world's largest data center market; if the state moves aggressively to facilitate reactor co-location, it could establish a template others follow.

The Opportunity Ahead for Developers

The siting rule changes won't build anything by themselves. What they do is open a door that was previously closed — and the projects that walk through it first will define how this market develops.

For infrastructure developers, the actionable move right now is site identification: finding parcels adjacent to planned or existing data center campuses that meet the physical and regulatory criteria for advanced reactor siting, and beginning the environmental and geotechnical baseline work that will be required anyway. That groundwork takes years. Starting it now, even before a specific reactor technology is selected, puts developers in position when the commercial and regulatory environment matures — which it will.

The data center industry's power problem isn't going away. Utilities are overwhelmed. Renewables can't solve baseload alone. Nuclear, now that it can sit where the load is, has a real answer. The question is who builds it first.

[INTERNAL LINK: reactor siting rules]

[INTERNAL LINK: data center energy demand]

[INTERNAL LINK: nuclear energy solutions]

For more insights on the evolving landscape of data centers and energy solutions, visit InfraSale Marketplace.