Deep Fission Deep-Well Reactors for AI Data Centres

Clean megawatts are in short supply, planning permissions move slowly, and patience is wearing thin across technology hubs.

Against that backdrop, a striking proposal from the United States has emerged: place compact nuclear reactors more than a mile below ground, then connect them directly to newly built campuses. The proposition relies on geology, drilling expertise and the need for dependable electricity at a predictable cost.

Why place reactors 1.6 km underground

Deep Fission, a US start-up, proposes lowering small nuclear units into 30-inch (76 cm) boreholes drilled to around 1.6 km. Endeavour Energy, which is behind Edged data centres, has agreed to pursue up to 2 GW for its AI-ready locations. The companies present the approach as a clean, dispatchable power source that avoids the land requirements, lengthy delivery schedules and grid-connection problems that can unsettle major surface projects.

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

The two main benefits

The first benefit concerns footprint and cost. A deep-well reactor would sit almost entirely underground. Above ground, the site would require only a relatively small pad, substation and supporting equipment. The firms say this could shorten build times and cut costly civil engineering work, including the vast containment structures needed for conventional plants. They are also aiming for a delivered cost of €0.05 to €0.07 per kWh, an attractive prospect for operators facing increasing electricity tariffs.

The second advantage is safety. At a depth of 1.6 km, surrounding geology becomes a passive protective barrier. It contains radiation, cushions the installation from external events and gives operators more time to react should a problem occur. The design is intended to lower the chance of an airborne release while making physical interference more difficult.

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

How the Deep Fission deep-well reactor could operate

The proposed system resembles a downhole heat source using a sealed primary loop. Drilling teams would create a narrow borehole, lower the reactor module into it and link heat exchangers to surface equipment that runs turbines or high-efficiency generators. The borehole provides shielding, while engineered casings control pressure, heat and fluids. Remote monitoring and replaceable modules are intended to make maintenance periods easier to manage.

The attraction becomes clearer when considering demand. The International Energy Agency estimates that data centres consumed about 1.3% of global electricity in 2023, equal to approximately 260 to 360 TWh. AI training workloads run for long periods, inference requirements are expanding, and local grids frequently do not have sufficient capacity. Locating generation alongside computing infrastructure therefore appears logical, while nuclear offers the availability profile sought by hyperscalers.

Attribute	Surface SMR	Deep-well SMR
Surface land use	Dozens of acres with visible structures	Small pad and substation
Shielding	Engineered containment buildings	Geologic barrier plus casing
Siting politics	Intense community scrutiny	Lower visual impact, fewer neighbours
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 centres

Endeavour intends to supply Edged sites with up to 2 GW of nuclear capacity, provided the technology passes licensing and financing milestones. At that scale, it could support several campuses with a fixed electricity price for decades. Colocation providers could build their offering around assured power rather than seeking substation upgrades or waiting for places in connection queues within overloaded regions.

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

The market signal is becoming clearer

Major technology companies have begun exploring nuclear-supported power contracts. Google has a framework agreement to purchase electricity from a small modular reactor developer. Other cloud and chip companies are investing in advanced nuclear start-ups or entering early offtake agreements. The underlying argument is consistent: clean, local and reliable electricity is preferable to fluctuating wholesale prices when GPU clusters cost billions and cannot operate without power.

Questions regulators will raise

The proposal is ambitious, but it must still address conventional nuclear issues as well as new questions related to geology and drilling.

• Licensing pathway: How will agencies classify deep-well units under current reactor regulations?
• Seismic and subsurface risk: What would occur during major ground movement or fault displacement at depth?
• Groundwater protection: How would casings, liners and seals stop any interaction with aquifers?
• Emergency planning: What form would an off-site plan take when the reactor core is beneath rock?
• Decommissioning: How would the module be recovered or entombed once its operating life ends?
• Fuel and waste: Which fuel type would be used, and how would spent assemblies be managed?

Deep Fission argues that geology limits potential accident routes. Regulators will test that assertion through modelling, test evidence and independent assessment. The nuclear sector has experienced gaps in public confidence. Rigorous measurement, open reporting and clear explanations will be as important as the engineering itself.

Costs, timelines and practical obstacles

The proposed price of €0.05 to €0.07 per kWh is appealing, but it depends on repeatable drilling, standardised modules and reliable finance. Grid interconnection remains necessary for backfeed and excess generation, although campus microgrids could manage most day-to-day operations. Delivery could be quicker than for a traditional plant if permissions, supply chains and drilling teams are all in place.

There are still significant risks. Underground projects can bring unexpected complications. Maintaining casing integrity over decades requires cautious engineering. Servicing equipment at depth depends on capable remote tools. Any interaction with groundwater could damage public support. During consultation hearings, clear information on sampling, monitoring and protective barriers will matter greatly.

What this means for cities and states

Areas seeking AI factories face increasing pressure on electricity supply. Solar and wind can provide low-cost energy, but not continuous output. Batteries can assist for several hours rather than several days. Gas can cover peaks but produces emissions. A compact nuclear module installed near the demand source addresses the duty-cycle challenge. It may also avoid prolonged disputes over transmission infrastructure, which can delay schemes for years.

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

Additional context for assessing the proposal

Small modular reactors include a broad range of designs and capacities. Deep-well concepts sit at the micro end of that range, with individual units supplying tens to hundreds of megawatts. This capacity better suits a data-centre cluster than an entire city. The arrangement also supports phased construction: add computing capacity, install another module, then repeat.

Cooling strategy also warrants close attention. A sealed primary loop could transfer heat to a secondary loop, which would reject it through dry coolers, hybrid towers or water-based systems. Water-stressed locations are likely to favour air-cooled or hybrid systems. Developers could also use low-grade heat for nearby buildings, greenhouses or absorption chillers, improving total site efficiency.

One practical indicator of progress is to watch for test boreholes, regulator pre-application submissions and agreements to supply fuel and drilling services. If these emerge, the project moves from a pitch deck towards a delivery plan. The data-centre sector operates through roadmaps. Electricity now needs one as well.