The bet: a reactor company that ships hardware
Most advanced-nuclear startups sell a design. Kairos Power is trying to sell a way of building — and that distinction is the whole thesis. Founded in 2016 to commercialize the fluoride salt-cooled, high-temperature reactor (KP-FHR), Kairos concluded early that the industry's core problem is not reactor physics but the cost, schedule, and first-of-a-kind risk of nuclear construction. Its entire program is organized around driving those down through repetition rather than around a single hero project.
That framing has aged well. As hyperscalers scramble for firm, carbon-free power to feed AI data centers, the scarce commodity is not a clever reactor concept — there are dozens — but a developer that can actually pour concrete on a licensed nuclear site and hit dates. Kairos is one of the very few that has done so: it holds the first U.S. construction permits issued for non-light-water reactors in over half a century, and in 2024–2026 it moved two reactors in Oak Ridge from paper into physical build. In an industry long on renderings and short on rebar, Kairos's differentiation is that it is building.
Origins and the iterative model
Kairos was co-founded by Mike Laufer and Ed Blandford, drawing on fluoride-salt reactor research that ran through UC Berkeley, MIT, and the University of Wisconsin. Rather than jump straight to a commercial plant, the company adopted a hardware-rich, iterative development loop borrowed more from aerospace and manufacturing than from traditional nuclear: build progressively more complete non-nuclear prototypes, learn from each, and fold the lessons into the next.
The physical expression of that philosophy is the Engineering Test Unit (ETU) program — full-scale, electrically heated, non-nuclear mockups of the Hermes reactor that let Kairos rehearse salt handling, pumping, integration, and construction methods long before any nuclear license is at stake. ETU 1.0 logged more than 2,000 hours of pumped-salt operation and produced the company's first high-purity FLiBe coolant; by mid-2025 Kairos was installing the reactor vessel for ETU 3.0 on its Oak Ridge campus, next door to the reactors it is licensing. Each ETU is deliberately cheaper to get wrong than a licensed reactor, which is the point.
Inside the KP-FHR
The KP-FHR pairs two mature-but-underused ingredients: TRISO pebble fuel — tiny uranium kernels wrapped in ceramic coatings rated to very high temperatures — and FLiBe, a molten mixture of lithium fluoride and beryllium fluoride, as coolant. Crucially, FLiBe stays liquid across a wide temperature band at atmospheric pressure, so unlike a conventional water-cooled reactor the KP-FHR runs hot without running at high pressure. That removes the massive, expensive pressure boundary that dominates light-water plant design and cost.
The safety story follows from the chemistry. Low-pressure operation means there is no stored energy to flash coolant into steam and drive a pressurized release; the TRISO fuel retains fission products to temperatures well above normal operation; and the salt's high boiling point gives large thermal margins. Kairos designs the reactor so decay heat can be removed passively, without operator action or offsite power. The commercial architecture is modular — Kairos describes plants built from paired ~75 MWe units that scale from roughly 150 MWe up toward 900+ MWe — so the same standardized reactor hardware serves a small data-center load or a large grid customer by adding units rather than redesigning the plant.
Hermes 1 and Hermes 2: the demonstration ladder
Kairos is climbing a deliberate two-rung demonstration ladder in Oak Ridge, Tennessee. The first rung is Hermes 1, a 35 MWth low-power test reactor that generates no electricity; its job is to prove the reactor, the FLiBe coolant system, the TRISO fuel, and the licensing and operating model at real nuclear scale. The NRC issued its construction permit in December 2023, and Kairos began nuclear safety-related construction in 2025.
The second rung, Hermes 2, converts that learning into power. It received NRC construction permits in November 2024 — making it the first electricity-producing Gen IV reactor project permitted in the United States — and Kairos broke ground on April 17, 2026. Re-envisioned as a single-reactor plant of up to 50 MWe, Hermes 2 is designed to feed the Tennessee Valley Authority grid, with first power targeted around 2030. Schedule has not been frictionless: in May 2026 the NRC granted Kairos a 28-month extension on the Hermes 1 construction-completion deadline (to April 30, 2029), and the company now expects to finish building the test reactor in 2028 — a first-of-a-kind slip that is worth watching but does not, on its own, undercut the strategy.
Money, moat, and the Google order book
Kairos funds itself unusually. It has avoided the splashy venture rounds common among its peers, leaning instead on a milestone-based U.S. Department of Energy cooperative agreement worth up to $303 million under the Advanced Reactor Demonstration Program, together with undisclosed private investment and, increasingly, strategic customer capital. The company remains firmly pre-revenue: it has no operating reactors and sells no power yet.
The commercial pivot arrived in October 2024, when Kairos signed a master plant development agreement with Google to build, and operate, a U.S. fleet of advanced reactors totaling up to 500 MW by 2035. In August 2025 that framework produced its first concrete offtake — a power purchase agreement among Kairos, Google, and TVA under which Hermes 2 will deliver up to 50 MW to the grid to help supply Google's Tennessee and Alabama data centers. That order book is the closest thing Kairos has to a moat: it converts a technology program into a repeatable, financeable pipeline, and it makes Kairos one of the reference cases for the emerging hyperscaler-nuclear deal structure. The company reinforces the industrial side of the story by standing up ISO 9001:2015-certified manufacturing at both Alameda, California and Albuquerque, New Mexico, positioning it for serial reactor-hardware production rather than one-off builds.
What to watch
The near-term scoreboard is concrete and dated. First, fuel: Kairos finalized a DOE HALEU supply contract and must manufacture the initial TRISO fuel pebbles — drawing on Los Alamos capability — needed for Hermes 1 startup. Second, the test reactor itself: completing construction (now expected in 2028) and bringing Hermes 1 online is the single most important proof point for the entire KP-FHR thesis. Third, first power: getting Hermes 2 into operation around 2030 would validate the leap from demonstration to commercial electricity sales under the Google agreement.
The bull case is that Kairos's build-to-learn discipline lets it convert the Google framework into a genuine fleet on cost and schedule, at exactly the moment data-center demand makes firm clean power extraordinarily valuable. The bear case lives in the same timeline: nuclear construction is unforgiving, the Hermes 1 deadline has already slipped once, and the commercial units still depend on fuel supply, HALEU availability, and licensing that has not yet been exercised at scale. For a company whose entire premise is that it can build reactors predictably, the coming Hermes milestones are where that premise gets tested for real.