Google’s First Kairos Nuclear Project Is Under Construction—Power Waits Until 2030
Google’s nuclear strategy has advanced from a broad procurement commitment to a named project under construction. A May 2026 update from Kairos Power records an April 17 groundbreaking for the Hermes 2 Demonstration Plant in Oak Ridge, Tennessee, identifies it as the first commercial-scale reactor under the Google agreement and puts its planned output at up to 50 megawatts.
The central limitation has not changed: Hermes 2 is not generating electricity. Google’s project account schedules operations for 2030 and explains that the Tennessee Valley Authority will buy the plant’s electricity, while Google will procure the associated clean-energy attributes for data centers in Tennessee and Alabama.
The first project now has a site and a commercial structure
The original Google–Kairos partnership, unveiled in October 2024, set a path toward multiple advanced reactors with combined capacity of up to 500 MW. Hermes 2 turns the first portion of that ambition into a specific deployment in Oak Ridge, with a utility purchaser and an intended destination for the environmental attributes attached to its output.
This arrangement is different from placing a private reactor beside a Google data center. Electricity from Hermes 2 is intended to enter TVA’s regional grid, where generation and demand are managed across the wider system. Google’s role is that of a major energy customer supporting the project and acquiring its clean-energy attributes through the utility.
The distinction matters when assessing the scale of the commitment. The proposed 50 MW belongs to the first deployment, while 500 MW is the ceiling for the wider multi-project collaboration. The larger figure does not represent capacity already built, operating or necessarily covered by the construction work now underway in Oak Ridge.
Breaking ground reduces uncertainty but does not remove it
The groundbreaking provides physical evidence of progress that the initial partnership lacked. Hermes 2 is also meant to carry lessons from non-nuclear engineering units and the separate Hermes test reactor into a power-producing project, giving Kairos a staged route from hardware demonstrations toward commercial deployment.
That sequence also shows why the target remains exposed to execution risk. First-of-a-kind nuclear construction involves detailed engineering, component production, fuel supply, civil works, commissioning and regulatory approvals. Beginning site work does not establish that the integrated reactor and power-conversion systems have completed those stages or can meet the planned operating date.
Kairos is developing a fluoride salt-cooled, high-temperature reactor rather than extending a conventional light-water design. Its system uses TRISO fuel particles embedded in carbon-matrix pebbles and molten fluoride salt as coolant. The project therefore has two jobs: deliver electricity and demonstrate that this reactor technology can be built, licensed and operated as the basis for a repeatable fleet.
The regulatory record requires careful reading
The Nuclear Regulatory Commission’s current Hermes 2 record lists construction permits issued on November 21, 2024, for a test facility comprising two low-power reactors rated at 35 MW thermal each, with a shared steam-powered conversion system. The page was last updated on August 11, 2026.
That regulatory description does not use the same scope or unit as the commercial presentation of up to 50 MW of electricity. Thermal megawatts measure reactor heat, whereas electrical megawatts measure output after conversion, so the figures cannot be compared directly. The public materials also describe the project from different institutional perspectives: the regulator documents the permitted test-reactor facility, while the developer presents Hermes 2 as its first commercial-scale, power-producing deployment.
The construction permits are therefore important but should not be mistaken for proof that electricity delivery has been authorized or completed. Construction, commissioning and operation are separate milestones. The decisive regulatory question is whether the facility satisfies every requirement applicable to operation before the planned start date.
What the project can—and cannot—do for Google
Nuclear generation appeals to data-center operators because it is designed to run independently of daily sunlight and wind conditions. If Hermes 2 performs as intended, its firm output could complement variable renewable generation on a grid serving computing facilities that operate continuously.
It cannot address Google’s immediate electricity demand because the plant is still years from its scheduled operating date. Nor does one project settle whether the broader collaboration can reach fleet scale. Later deployments would need suitable sites, commercial agreements, supply chains, construction programs and regulatory reviews of their own.
The project is consequently more substantial than a corporate promise, but it is not yet an operating solution to AI-related power demand. Its present status is narrower and more useful: work has begun on the first named deployment, a utility purchase structure is in place, and future electricity has been assigned a role in Google’s regional energy procurement. Whether the model succeeds will depend on Hermes 2 reaching licensed operation and on Kairos reproducing that result beyond the first plant.
Also read:
Subscribe to our newsletter
Get the latest Web3, AI, and crypto news delivered straight to your inbox.