Deep Fission bets a mile of Kansas bedrock can replace a containment dome

Two boreholes are in play, and they answer different questions. The data-collection well already drilled reaches 1830 metres, about 221 metres deeper than the mile the design assumes ^[5][19]. The proof-of-concept hole coming next is 762 metres, roughly 47 per cent of that mile ^[5][18]. New Scientist reports that the following phase puts the reactor canister, heat exchanger and other components into the borehole, with a nuclear demonstration targeted for 2027 ^[6], without specifying which hole receives them. Assembling the string at half depth would show whether the pieces fit together; only full depth exercises the pressure the design leans on ^[10].

The premise is that a mile of rock does the containment work a heavy concrete dome does at the surface, and that the water column above supplies what a pressuriser supplies in a surface plant ^[3][10]. What is missing is the size of the prize. Domes account for a significant portion of conventional plant cost and build time ^[3], which gives a direction without a magnitude, and Deep Fission's chief operating officer, Michael Brasel, is careful in the same way: removing engineered active emergency core cooling with pumps and backup power, he says, "could translate into real capital and operating savings" ^[11]. That figure has not been converted into dollars per kilowatt outside the company.

The trade is stated plainly by the engineers New Scientist asked. Leslie Dewan of Neutronic Designs says the design uses the environment to perform functions normally engineered into a plant, and that the loss of easy access makes inspection and maintenance both harder and more important ^[12]. Todd Allen at the University of Michigan frames it as a margin problem: even a reactor built for minimal maintenance cannot be prepared for every eventuality, and what worries him is the event nobody anticipated ^[13]. Brasel says the reactor would not be raised for routine maintenance, but can be raised if needed ^[14].

The cost case also depends on repetition. At 15 megawatts a unit ^[7], the company's 150-megawatt site is ten holes and its gigawatt-scale site is about 67 ^[20]. Deep drilling is routine oil and gas work, and the surface heat exchanger is standard geothermal equipment ^[8], so those unit costs have references in the world. The assembly does not. Dewan's summary is that the hard part is "proving that all of these individually familiar technologies will work together in this unfamiliar configuration" ^[9].

No reactor has operated at this depth. The nearest precedent, Norway's Halden research reactor, sat in a mountainside cavern about 100 metres down, also for containment ^[15], which is around a sixteenth of the target depth ^[21]. My read is that the unresolved variable is the borehole system rather than the reactor: a pressurised water reactor on low-enriched fuel ^[7] is well characterised, while how casing, vessel and surrounding geology behave over decades of hot operation is not ^[16]. A 2027 demonstration ^[6] would speak to the first of those and say very little about the second.