Deep Fission deep-well SMR for Edged data centers: 1.6 km reactors and 2 GW ambition

Clean electricity at scale is hard to come by, planning approvals take time, and technology hubs are running out of tolerance for delays.

A fresh proposal from the United States aims to relieve that strain: place compact nuclear reactors more than a mile underground and connect them directly to new campuses. The case for it rests on geology, modern drilling techniques, and the demand for firm power at a predictable long-term price.

Why bury reactors 1.6 km down

Deep Fission, a U.S. startup, says it can deploy small nuclear units into 30-inch (76 cm) boreholes drilled to about 1.6 km. Endeavour Energy, the firm behind Edged data centres, has agreed a plan targeting up to 2 GW for AI-ready sites. Together, they position deep-well nuclear as a low-carbon, dispatchable option that avoids the land-take, lengthy schedules, and grid-connection friction that often derail large above-ground schemes.

"Two promised advantages stand out: a smaller surface footprint and a stronger safety envelope delivered by the rock itself."

The two big advantages

The first claimed benefit is footprint and cost. With most of the reactor system underground, what remains at the surface could be limited to a relatively small pad, a substation and supporting equipment. The companies say that could shorten build programmes and cut pricey civil engineering work, including large containment structures. They are also aiming for a delivered cost of €0.05 to €0.07 per kWh-an eye-catching figure for operators dealing with escalating utility tariffs.

The second is safety. At 1.6 km depth, surrounding geology is intended to serve as a passive barrier: it can attenuate radiation, provide resilience against external shocks, and increase the time available for operators to respond if something goes wrong. The approach is also presented as reducing the likelihood of airborne release and making physical interference more difficult.

"Rock becomes a permanent shield. No giant dome. No skyline-changing tower."

How the deep-well reactor would work

In practice, the concept is similar to a downhole heat source paired with a sealed primary loop. A narrow shaft is drilled, the reactor module is lowered into place, and heat exchangers are linked to surface equipment that can run turbines or supply high-efficiency generators. The borehole provides much of the shielding, while engineered casings are intended to control pressure, temperature and fluid management. The proposal also relies on remote monitoring and swapping modular components to streamline maintenance intervals.

The demand picture makes the attraction clear. The International Energy Agency estimates data centres consumed about 1.3% of global electricity in 2023, equivalent to roughly 260 to 360 TWh. AI training workloads can run for extended periods, inference scales rapidly, and local grids frequently cannot expand quickly enough. Locating generation alongside compute therefore looks pragmatic, and nuclear power offers the high-availability profile hyperscalers tend to prioritise.

Attribute	Surface smr	Deep-well smr
Surface land use	Dozens of acres (roughly tens of hectares) with visible structures	Small pad and substation
Shielding	Engineered containment buildings	Geologic barrier plus casing
Siting politics	Intense community scrutiny	Lower visual impact, fewer neighbors
Cooling approach	Often needs large water systems	Closed-loop systems, careful groundwater isolation
Security posture	Perimeter-heavy, above ground	Hard to access, below grade
Maintenance	On-site crews, larger components	Modular service, constrained access

What it could mean for ai-scale data centers

If licensing and financing hurdles are cleared, Endeavour intends to supply Edged locations with up to 2 GW of nuclear capacity. That volume could support multiple campuses with a stable, long-duration price profile. Colocation providers might then package services around assured power availability, rather than waiting on substation upgrades or competing for limited queue capacity in overloaded markets.

"Stable power at the fence line changes site selection and speed-to-market for new compute."

The market signal grows louder

Large technology firms are already experimenting with nuclear-linked procurement. Google has a framework agreement to purchase electricity from a small modular reactor developer. Elsewhere, cloud and semiconductor companies are funding advanced nuclear startups or committing to early offtake arrangements. The repeated message is that clean, local and reliable supply can be more valuable than exposure to volatile wholesale pricing-especially when GPU clusters cost billions and cannot deliver value without power.

Questions that regulators will ask

It is an ambitious concept, but it still must address the usual nuclear regulatory issues-plus additional ones tied to drilling and subsurface conditions.

• Licensing pathway: How do agencies treat deep-well units under existing reactor rules?
• Seismic and subsurface risk: What happens under strong ground motion or fault movement at depth?
• Groundwater protection: How do casings, liners, and seals prevent any interaction with aquifers?
• Emergency planning: What does an offsite plan look like when the core sits under rock?
• Decommissioning: How do you retrieve or entomb the module after its service life?
• Fuel and waste: What fuel form is used, and how do you handle spent assemblies?

Deep Fission argues that geology narrows the credible accident pathways. That assertion will be tested through modelling, experimental evidence and independent review. The sector also carries the legacy of public trust deficits, so rigorous measurement, open reporting and plain-language explanations will be as important as the technical design.

Costs, timelines, and real-world hurdles

A delivered price of €0.05 to €0.07 per kWh would be compelling. However, it depends on repeatable drilling performance, standardised modules and financing that remains predictable. Even with on-site generation, grid interconnection still plays a role for export and backfeed, although campus microgrids could cover most day-to-day operation. Compared with conventional plants, delivery could be faster-provided permits, supply chains and drilling teams align.

There are still material risks. Underground construction can produce surprises. Maintaining casing integrity for decades will require conservative engineering assumptions. Servicing equipment at depth means dependable remote tools. Any contact with groundwater would put public confidence at risk. During consultations and hearings, clear messaging on sampling, monitoring and multiple barrier layers will carry significant weight.

What this means for cities and states

Places trying to attract AI-scale industrial development are encountering a power bottleneck. Solar and wind can be low-cost, but they are not continuous. Batteries can bridge hours rather than days. Gas can cover peaks but increases emissions. A compact nuclear unit close to the load addresses the duty-cycle challenge and also avoids contentious transmission projects that can hold developments up for years.

"Put power under the parking lot, not 200 km away behind a contested transmission line."

Extra context that helps frame the bet

Small modular reactors span many sizes and technical approaches. Deep-well variants sit at the micro end of the spectrum, where individual units deliver tens to hundreds of megawatts. That output range aligns more naturally with data centre clusters than with supplying an entire city. It also suits phased expansion: deploy compute capacity, install another module, and repeat.

Cooling deserves careful scrutiny. With a sealed primary loop, heat can be transferred to a secondary loop and rejected via dry coolers, hybrid towers or water-based systems. Sites facing water constraints will favour air-cooled or hybrid configurations. Developers may also recover low-grade heat for nearby buildings, greenhouses or absorption chillers, improving overall site efficiency.

A practical way to track whether the concept is progressing is to look for test wells, pre-application engagement with regulators and supply contracts for fuel and drilling services. When those appear, schedules start to shift from investor slides into executable project plans. Data centre operators work to roadmaps-and increasingly, power must have one too.