SMRs and Advanced Reactors: What's Actually Being Built

The blue diamonds on this map are the next generation, and they're best understood as answers to two different questions about today's light-water giants.

The first question is economic: what if the problem is size? Gigawatt plants are bespoke megaprojects — Vogtle 3&4 cost ~$35 billion and ran years late — and megaproject risk, not physics, is what nearly killed Western nuclear construction. Small modular reactors (SMRs) bet that reactors of 50–300 MWe, factory-fabricated in series and shipped to site, can trade economies of scale for economies of replication: lower per-unit capital risk, faster builds, learning curves like aircraft manufacturing instead of cathedral construction. The bet is unproven — it lives or dies on volume — but it's why nearly every project in the pipeline panel is measured in hundreds rather than thousands of megawatts. Some SMRs stay technologically conservative: GE Vernova's BWRX-300 (a shrunk, simplified BWR; Clinch River, TN) and Holtec's SMR-300 (a compact PWR, sited at the restarted Palisades plant) keep water cooling precisely so they can lean on seventy years of light-water experience and fuel supply.

The second question is technical: what if the problem is water? Water moderates beautifully but boils at 100°C, forcing the enormous pressures (and pressure-rated everything) of the LWR, and capping steam temperatures at modest thermal efficiency. The advanced coolants each buy something water can't offer:

Sodium (TerraPower's Natrium in Kemmerer, Wyoming — the most advanced new build in the country, under construction on a retiring coal-plant site; Oklo's Aurora microreactors at Idaho National Laboratory) runs near atmospheric pressure at ~500°C, with fast neutrons that can eventually burn the transuranics in spent fuel. Natrium's signature is on the grid side: a molten-salt heat-storage tank lets the 345 MWe reactor flex its electrical output up to 500 MWe for hours — a baseload machine that can also chase the evening peak in a renewables-heavy grid.

Helium (X-energy's Xe-100, sited with Dow at Seadrift, Texas and with Energy Northwest in Washington under Amazon backing) is chemically inert and runs to 750°C+, unlocking industrial process heat — which is why its flagship customer is a chemical plant, not a utility. Its TRISO fuel — uranium kernels individually wrapped in layered ceramic shells, billions per core, each shell its own pressure vessel — is often called the most robust fuel form ever engineered; it simply cannot melt at reactor-achievable temperatures.

Molten salt (Kairos Power's Hermes demonstrations in Oak Ridge, TN, backed by Google) dissolves the coolant question differently: a fluoride salt that runs hot at atmospheric pressure with enormous thermal margin, paired in Kairos's case with TRISO pebbles. Kairos's iterative hardware-demo approach — build small, learn, repeat — is the most Silicon-Valley development style in the field.

Three system-level realities frame all of it. Fuel: most of these designs need HALEU (the 5–20% enrichment tier from the fuel-cycle page), whose supply chain is only now being born — a genuine schedule risk. Licensing: the NRC has now certified one SMR design (NuScale's, the first ever) and issued the first non-LWR construction permits in half a century (Kairos, TerraPower's permit progress), but every first-of-a-kind review is slow by definition; the regulation page covers the new pathways meant to speed it. Demand: what's genuinely different about this wave versus past false dawns is the customer list — Google, Amazon, Dow, Meta — visible one layer over in the deals data. Previous advanced-reactor eras pitched governments; this one has signed offtakers, and several projects (the violet markers on the demand layer) are being co-developed with the data centers they'll power.

Sober expectations, stated plainly: as of mid-2026 no SMR generates commercial power in the US; Natrium and the first demos target the late 2020s; series production, if the bet pays, is a 2030s story. The pipeline panel ranks every project by how far through that gauntlet it has actually gotten — stage labels, not press releases.