Deep Fission bets a mile-deep borehole can replace a nuclear containment building

Pre-revenue developer of the Gravity Reactor, a 15 MW pressurized-water unit meant to be lowered on cables into a 30-inch, one-mile-deep borehole and paired with a surface steam generator, aimed first at AI datacenters.

Depth is the cheapest safety system: a submarine-grade PWR in a mile-deep borehole needs no containment building and can sell power at 5–7¢/kWh — half new-nuclear cost.

$4M (Aug 2024); an unfinished $15M seed attempt (Apr 2025); ~$30M net from a $3/share reverse merger (Sept 2025); ~$33M net from a June 2026 IPO priced at $16/share; $93.0M cash at 2026-06-30. 8VC funds hold over 5%.

Physicist Richard Muller (UC Berkeley emeritus; MacArthur Prize, Waterman Award, Breakthrough Prize; 34 years in JASON) and his daughter Elizabeth Muller had already founded nuclear-waste startup Deep Isolation in 2016. In mid-2023 they inverted the idea: if boreholes work for waste, they work for reactors. Deep Fission incorporated on 2023-07-17 with Richard as CEO; Elizabeth — formerly CEO of Deep Isolation and an OECD policy advisor — took over as CEO in March 2024.

NRC engagement began in March 2024. In January 2025 datacenter developer Endeavour agreed to host 2 GW of the company's reactors and invested undisclosed terms. In August 2025 DOE selected Deep Fission for its Reactor Pilot Program, a streamlined path outside standard NRC licensing. After the private seed stalled, the company merged with shell Surfside Acquisition on 2025-09-05, netting ~$30M at $3/share on the OTCQB. In June 2026 it completed a traditional IPO at $16/share (~$32.9M net) — a prospectus that disclosed restated financials, material weaknesses, and substantial doubt about going concern.

Boreholes carry regulatory and operational precedent from Deep Isolation's waste work, so the founders bet on geology they knew; the PWR inside is deliberately conventional.

AI datacenters created a rare customer willing to sign 2 GW before hardware existed, giving the startup an order book to fund engineering instead of waiting for utility procurement cycles.

The DOE Reactor Pilot Program offered a permitting path outside the multi-year NRC queue, trading regulatory depth for speed — the same trade the borehole makes underground.

Going public twice was a financing strategy: micro-cap markets priced risk differently than private investors who would not close a $15M seed.

When private capital hesitates, public micro-cap markets may say yes — twice. A pre-revenue nuclear startup listed at $3/share, then IPO'd at $16, going-concern warning printed in the prospectus.

As of 2026-06-30 the company holds $93.0M in cash after its June IPO, with proceeds earmarked for engineering the first pilot nuclear reactor. It is one of ten DOE Reactor Pilot Program selections and intends its first reactor online by mid-2026-era targets set in 2025; TechCrunch reported the goal as starting the first reactor by July 2026. The prospectus puts the remaining bill at ~$67M to a test reactor and ~$138M to a licensed commercial unit, and management states existing resources will not fund operations through the next twelve months — so another raise is expected soon.