Kairos Power broke ground on its Hermes 2 demonstration plant in Oak Ridge, Tennessee, on April 17, 2026, starting construction on a pair of reactors designed to feed up to 50 megawatts of electricity directly into the Tennessee Valley Authority grid for Google data-center operations. The project sits on a regulatory foundation laid in late 2024, when the Nuclear Regulatory Commission issued two construction permits in a single action, and it now tests whether a non-light-water reactor can move from permit to power delivery fast enough to meet surging demand from hyperscale computing.
Data-center demand is reshaping the NRC permit pipeline
The NRC granted construction permits CPTR-7 and CPTR-8 for the Hermes 2 facility on November 21, 2024, according to the Federal Register notice published eight days later. Those permits authorize two fluoride-salt-cooled, high-temperature reactor units at the same Oak Ridge campus where the earlier Hermes non-power test reactor already holds its own permit, CPTR-6. The speed of the review stands out: the agency completed a final environmental assessment and issued a finding of no significant impact before signing off on construction, compressing what has historically been a multi-year process for reactor licensing.
That pace reflects a concrete commercial pull. Kairos Power’s own announcement states that the Hermes 2 units will supply up to 50 MW to the TVA grid specifically to support Google data-center decarbonization. Traditional utility planning cycles, where a public power authority forecasts residential and industrial load growth over decades, did not produce this project. A single corporate buyer with an urgent appetite for carbon-free electricity did. The NRC’s willingness to process the Hermes 2 application on a timeline that aligns with data-center build schedules, rather than the slower cadence of conventional baseload additions, signals that regulators are treating this category of demand as its own driver for advanced reactor demonstrations.
The hypothesis that data-center load is functioning as a distinct accelerant for non-light-water reactor permitting holds up against the public record, but it also has limits. Nothing in the NRC’s formal documents explicitly names data centers as a reason for the review pace. The agency’s stated rationale rests on safety findings, environmental review, and the applicant meeting regulatory criteria. The commercial motivation lives in the Kairos and Google announcements, not in the permit text itself. That gap matters because it means the speed could be attributed to the project’s technical simplicity or the NRC’s growing familiarity with the Hermes design family, rather than to any deliberate policy tilt toward tech-sector electricity buyers.
Permits, environmental reviews, and the Oak Ridge site record
The regulatory paper trail for Hermes 2 builds on years of work at the Oak Ridge location. The NRC’s earlier permit for the Hermes non-power test reactor, documented under ADAMS accession ML23338A258, established the site’s suitability for Kairos Power’s molten-salt technology. That first reactor does not generate electricity, but its environmental and safety reviews created a baseline that the Hermes 2 application could reference. The final environmental impact statement for Hermes, cataloged in NUREG‑2263, evaluated alternatives including a no-action scenario and an alternate site in Idaho before concluding that the Oak Ridge location was acceptable.
For Hermes 2, the NRC took a lighter environmental path. Staff produced an environmental assessment rather than a full environmental impact statement, and the resulting finding of no significant impact cleared the way for the construction permits. The two permits carry ADAMS accession numbers ML24324A021 and ML24324A022, tying them to the broader docket that includes the Commission’s mandatory hearing memorandum and the safety evaluation package. Each of these documents is publicly accessible through the NRC’s electronic records system, giving outside engineers, competitors, and watchdog groups a detailed look at the agency’s reasoning.
The decision to use an environmental assessment instead of a full impact statement for a power-producing reactor is itself notable. It suggests the NRC concluded that the Hermes 2 units, at 50 MW combined, fall below the threshold of environmental consequence that would trigger the longer review. For Google and Kairos, that distinction saved months. For communities near Oak Ridge, it means the public comment window and the depth of environmental analysis were narrower than what a full EIS would have required.
Open questions between the permit and the power line
Several pieces of the story remain outside the public record. No executed power purchase agreement between Kairos Power, Google, and TVA appears in any NRC docket or Federal Register filing. The 50 MW figure that anchors the project comes from public statements, not from a tariff schedule or a filed contract. That leaves open questions about how the output will be priced, how curtailment risk will be allocated between TVA and Google, and what happens if the reactors underperform or face extended outages.
Financing is another unresolved element. Construction permits authorize Kairos to build, but they do not guarantee cost recovery. Traditional nuclear plants have relied on rate-base treatment, federal loan guarantees, or long-term offtake contracts to secure capital. In the Hermes 2 case, the demonstration label signals that Kairos is treating the facility as a bridge between its non-power test reactor and a future commercial fleet. Whether that bridge is funded primarily by private equity, strategic investors, or structured agreements with customers like Google has not been detailed in regulatory filings.
Technical risk also looms between the permit and the power line. The Hermes 2 reactors use a fluoride-salt coolant at high temperature, a configuration that differs substantially from the light-water reactors that dominate the U.S. fleet. The NRC’s safety evaluation concludes that the design meets existing regulatory requirements for the construction phase, but commissioning and operation will test how the novel coolant behaves under real-world transients, maintenance cycles, and grid interactions. Any unexpected behavior could slow the path to commercial operation, regardless of how quickly the permits were issued.
Even the grid connection involves uncertainties. TVA must integrate a relatively small but highly specialized power source into a system that is balancing conventional generation, growing renewable portfolios, and rising data-center loads. If Hermes 2 is dispatched primarily to serve Google’s facilities, questions arise about how its output is treated in system planning models and how costs are allocated across TVA’s broader customer base. Those decisions will shape whether advanced reactors are seen as bespoke solutions for tech companies or as shared assets that benefit the entire region.
A test case for advanced reactors in the AI era
Despite those open questions, Hermes 2 stands as an early test case for how advanced reactors might scale in an era defined by artificial intelligence and cloud computing. Data centers are emerging as anchor customers with both the financial capacity and the climate commitments to underwrite first-of-a-kind projects. By aligning its demonstration schedule with a specific corporate load, Kairos is attempting to de-risk the demand side of the equation that has historically plagued nuclear projects.
The regulatory story at Oak Ridge will therefore resonate beyond Tennessee. If Hermes 2 moves from construction permit to synchronized power within a timeframe that matches data-center build cycles, it will strengthen the argument that non-light-water reactors can be deployed as modular infrastructure for digital economies. If delays mount, or if the project encounters unanticipated regulatory hurdles despite its streamlined environmental review, it will reinforce skepticism about nuclear’s ability to respond to fast-moving market signals.
For now, the record shows a regulator willing to adapt processes, a developer leveraging prior site work to accelerate approvals, and a technology buyer prepared to tie its decarbonization goals to a first-of-a-kind power plant. How those elements interact over the next several years will determine whether Hermes 2 is remembered as a one-off experiment or as the template for pairing advanced nuclear reactors with the next wave of energy-hungry computing infrastructure.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
