EN
Back to the archive

The archive · Climate & Energy · Strategic decision · 2016–2026

Kairos Power's molten-salt bet: Google orders 500 MW, TVA signs first Gen IV PPA

Fluoride-salt reactors via small iterative demos: first Gen IV permits since 2023, Google's 500 MW order, TVA's first advanced-reactor PPA.

Kairos Power

The betThat a series of small licensed reactors — Hermes, then Hermes 2 — is the fastest path to commercial molten-salt nuclear, and Google-scale offtake can fund it.Building

What the business is

Develops fluoride-salt-cooled, high-temperature nuclear reactors (KP-FHR) as small, iteratively deployed plants delivering clean, 24/7 power.

How it started

Founded in 2016 in Alameda, California by Mike Laufer and co-founders to commercialize a fluoride-salt-cooled, pebble-fueled high-temperature reactor. Instead of betting everything on one giant first plant, the team applied a rapid iterative, vertically integrated approach — build small, test, learn — and in December 2023 won the first US construction permit for a non-water-cooled reactor in more than 50 years for the Hermes demonstration plant.

What happened

In October 2024, Google signed a Master Plant Development Agreement for 500 MW from a fleet of Kairos plants by 2035, with the first by 2030 — the first US corporate agreement for multiple deployments of a single advanced reactor design. In November 2024 the NRC approved construction permits for Hermes 2, the first electricity-producing Generation IV plant permitted in the US. In August 2025, Kairos, Google and TVA announced a PPA under which Hermes 2 — re-engineered from 28 MW to 50 MW — will feed TVA's grid powering Google data centers in Tennessee and Alabama, the first US utility PPA for an advanced reactor; safety-related construction of Hermes began in 2025.

How it ended up

Still in build: Hermes is under construction in Oak Ridge, Tennessee, Hermes 2 is permitted with operations targeted around 2030, and the Google fleet of up to 500 MW is due by 2035.

Background

Kairos Power was founded in 2016 in Alameda, California to commercialize the KP-FHR: a reactor cooled by fluoride salt and fueled with ceramic pebbles, designed to run safely at low pressure. The bet was not just the reactor — it was the path to market. Instead of raising once to build a huge first-of-a-kind plant, Kairos builds a series of small reactors, each one teaching the next, with licensing, supply chain and construction de-risked in small steps.

The strategy produced a string of firsts. In December 2023 the NRC issued a construction permit for Hermes, the first non-water-cooled reactor approved for construction in the US in more than 50 years. In October 2024 Google signed a master agreement to buy 500 MW from a Kairos fleet by 2035 — the first corporate agreement for multiple deployments of a single advanced reactor design in the US — with first power targeted by 2030. A month later the NRC approved construction permits for Hermes 2, the first electricity-producing Generation IV plant permitted in the US.

In August 2025 the bet became a market test: Kairos, Google and the Tennessee Valley Authority announced a power purchase agreement under which Hermes 2, upscaled from 28 MW to 50 MW, will deliver 24/7 power to TVA's grid that serves Google data centers in Tennessee and Alabama. TVA became the first US utility to sign a PPA for an advanced Generation IV reactor, and Hermes 2 became the first deployment under the Google fleet agreement.

The company is still building: safety-related construction of Hermes is underway in Oak Ridge, and Hermes 2 operations are targeted around 2030, ahead of the 500 MW fleet by 2035. Whether molten-salt reactors can be built on cost and schedule remains unproven — that is precisely what the iterative approach is designed to find out.

What has to be true

  • Advanced reactors historically die on one giant, never-finished first plant; Kairos split the risk into small permitted steps (Hermes, Hermes 2).
  • The NRC's first non-water reactor permit in over 50 years gave Kairos a regulatory moat no US competitor had.
  • Google's milestone-based 500 MW agreement turned a startup's reactor program into a customer-backed delivery schedule.
  • The TVA PPA closed the loop: a utility buyer, an offtaker (Google) and a permitted design in the same deal.
  • Data-center load growth gave firm, clean power a market willing to pay for new supply — the demand that makes the bet bankable.

What can be applied

Iterative licensing compounds: permit a tiny demo, reuse the review for Hermes 2, and let milestone-based corporate offtake de-risk the fleet — instead of one giant first plant.

Aftermath

As of September 2026, Kairos Power is executing its iterative build-out in Oak Ridge, Tennessee: Hermes is under construction as the non-nuclear-to-nuclear demonstration step, Hermes 2 has construction permits and is scheduled to begin producing around 2030, and the Google fleet agreement targets up to 500 MW by 2035. The TVA PPA is the commercial template — a utility buying power from an advanced reactor for the first time, with Google taking the clean-energy attributes for its data centers. Cost and schedule delivery remain the open question the company's whole approach is meant to answer.

Sources

spotted an error? The archive wants to know.

Your turn

You just read one. Describe what you are building, and see who is betting on the same thing.

Free account · 3 free questions · no card

Related cases