This group sells nuclear power-generating capacity — the machines that split atoms (nuclear fission) to make heat, and use that heat to make electricity. Three kinds of company sit here. First, SMR developers (SMR = Small Modular Reactor: a factory-built reactor in the roughly 50–300 MW range, smaller than the 1,000+ MW giants built piece-by-piece on-site today): NuScale (SMR) and Oklo (OKLO) are the pure-play designers, both largely pre-revenue (little or no current sales). Second, builders and component makers who already earn cash from nuclear: BWX Technologies (BWXT) makes US Navy submarine/carrier reactors and reactor parts; GE Vernova (GEV) builds the BWRX-300 SMR plus the steam turbines and generators that turn reactor heat into electricity. Third, operators who own running reactors: Constellation Energy (CEG), which runs the largest US nuclear fleet and sells the electricity. The product is in demand because AI data centers need enormous, always-on, carbon-free power, and nuclear is one of the densest sources of exactly that.
Demand for firm, carbon-free power (firm = available on demand around the clock, not weather-dependent) is rising while new nuclear supply grows slowly this decade. The binding limit is physical and slow to fix: there is no new US reactor in years, the forging/welding/fuel supply chain has shrunk, and SMRs are still pre-commercial. The cash actually being earned today sits mostly in the diversified builders and the reactor operators, not the SMR start-ups, whose value is almost entirely future-dated. What you pay today for each kind of cash, and whether deployment timelines arrive, are facts the reader weighs — this sheet does not judge them.
The product is electrical generating capacity from nuclear fission, sold either as megawatts (MW) of capacity (the maximum power a plant can produce) or megawatt-hours (MWh) of energy actually delivered. A large reactor is ~1,000+ MW; an SMR is roughly 50–300 MW. The thing a buyer ultimately cares about is firm, 24/7 power: a reactor running near 90%+ of the hours in a year is worth far more than the same nameplate rating in intermittent solar or wind, because data centers cannot pause.
The companies make money in distinctly different ways, and it matters for an owner:
So within one group you have pre-revenue option value (SMR, OKLO), lighter-asset manufacturing cash (BWXT), large diversified equipment cash (GEV), and operating-asset cash (CEG).
Contracted/known demand today: hyperscalers are already signing for nuclear power. The scan notes Microsoft, Google, Amazon, and Oracle have signed nuclear PPAs or announced nuclear-powered data-center plans, and that SMRs are "the technology most frequently cited in data center power announcements," with Kairos Power, X-energy, and NuScale signing agreements with parties such as Google and Dow. Constellation's restart of a Three Mile Island unit under a long Microsoft contract is the clearest example of a hyperscaler paying directly for firm nuclear output.
Forward demand (forecast): US data-center electricity load is the swing factor. The scan's power work puts US data-center demand at roughly 25 GW today, rising to about 50–70 GW by 2030 est., against a backdrop where total US electricity demand had been roughly flat for around 15 years before AI broke the trend. Every incremental gigawatt of always-on AI load is a candidate for firm, carbon-free supply, and nuclear is one of the densest such sources. These figures are forecasts, not contracted volumes.
The AGI lens: given AGI is arriving, compute demand does not plateau — it compounds, and physical-AI (robotics and autonomous systems, i.e. AI acting in the physical world) adds further electrical load on top of model training and inference (inference = running a trained model to answer queries). The constraint on scaling intelligence shifts from chips to power, which makes firm baseload (the steady around-the-clock supply that runs constantly) the strategic bottleneck of the build-out, and positions nuclear as a supply with a strong combination of density, uptime, and zero carbon. This is the structural reason hyperscalers are reaching years into the future for reactor capacity rather than buying spot power (power bought as-needed at current market prices). This is a forward-looking thesis, not a contracted fact.
Who the buyers are: (1) hyperscalers and large AI labs buying firm power via PPAs or, increasingly, equity/offtake in reactor projects (offtake = a commitment to buy the future output); (2) the US government/DOE (Department of Energy), which funds advanced-reactor deployment and (for BWXT) buys naval reactors regardless of the AI cycle; (3) utilities and industrial users (e.g. Dow) extending plant life or co-locating SMRs next to their operations.
✓ VERIFIED — the following figures were confirmed from primary sources after initial publication:
Current supply is highly constrained. The scan is blunt: "No new commercial reactor has been completed in the US since Vogtle (2023–2024)." SMRs are "still pre-commercial" — NuScale received NRC (Nuclear Regulatory Commission, the US licensing authority) design certification but canceled its first project on cost overruns, a concrete reminder that paper agreements are not delivered megawatts.
The main bottleneck is physical and slow, not financial. Per the scan: "The nuclear supply chain (forgings, qualified welders, ASME-certified components) has atrophied over decades. Rebuilding it will take years." (ASME = American Society of Mechanical Engineers, whose certification is required for nuclear-grade pressure parts.) Large reactor forgings (massive shaped steel parts), pressure-grade welds, and ASME-certified components have few qualified suppliers worldwide; trained nuclear welders and engineers are scarce after decades of low US build activity; and fuel — especially the higher-enriched HALEU fuel (High-Assay Low-Enriched Uranium, a more concentrated reactor fuel) that several advanced SMRs need — is a separate constrained input largely tied up in limited enrichment capacity. On top of physical inputs, NRC licensing itself is multi-year. None of these are solved by simply spending more money quickly.
Expansion plans (forecast): SMR developers target first deployments in the late 2020s to early 2030s; builders like GEV and component makers like BWXT are scaling factories and qualifying supply chains; the DOE is funding advanced-reactor programs to seed capacity. But the lead times mean meaningful new commercial nuclear supply in volume is largely a 2030s event, not a 2026–2028 one. These dates are developer targets and forecasts, not delivered capacity.
Market-share structure (who controls supply): supply is concentrated and specialized. BWXT supplies essentially all US naval nuclear manufacturing and is a leading nuclear-component supplier — a position built on hard-to-replace certifications and decades of relationships. GE Vernova is one of very few makers of both an SMR design (BWRX-300) and the large turbine-generator "power island" (the equipment that converts reactor heat into grid electricity), giving it a broad manufacturing position. Curtiss-Wright (CW, in the scan but not a headline ticker here) supplies coolant pumps and instrumentation "to every US nuclear plant." The SMR designers (SMR, OKLO) hold design IP and first-mover positioning but do not yet control any operating supply. Operating reactor supply itself is owned by utilities/IPPs (independent power producers — companies that generate electricity to sell rather than serving captive customers), with Constellation holding the largest US nuclear fleet. Specific market-share percentages are not in the provided files and are therefore not stated here.
Put together, the product is structurally short this decade (demand running ahead of supply): firm carbon-free power demand is rising on an AI/AGI curve while net new nuclear supply is roughly flat until the 2030s. The evidence of shortage is qualitative but consistent — hyperscalers signing direct PPAs and even funding reactor restarts/projects (buyers do not pre-buy years out for something in surplus), the willingness to pay premium prices for firm power noted in the related IPP/utility scans, multi-year licensing and supply-chain lead times, and SMR capacity being "spoken for" via agreements long before it can be built.
The counterweight to keep in view: the shortage is real, but the nuclear-specific relief is slow on both sides. Because reactors take years, much of the near-term AI power gap is being filled by gas, grid upgrades, and existing nuclear restarts/uprates (uprate = boosting an existing reactor's output) — not by new SMRs. So the SMR portion of this group is exposed to a gap that bites hardest in the 2030s, while the operator/builder portion is exposed to the gap that exists right now. Which timing matters is a fact the reader weighs.
| Dimension | Demand side | Supply side | Net read |
|---|---|---|---|
| Near term (now–2028) | Rising fast (AI load) signed PPAs | ~Flat new reactors; restarts/uprates only | Short — met by existing nuclear, gas, grid |
| Medium term (2028–2032) | Compounding (AGI compute + physical AI) forecast | First SMRs deploy; supply chain reflating forecast | Still short if forecasts hold |
| Could flip to oversupply when | AI load growth slows / efficiency gains | SMR mass-production + fuel scale-up succeed | Earliest a 2030s question; not visible now |
Source: 500-stocks Energy & Power scan, Nuclear (Reactors/SMR/Fuel) and IPP/Utility sub-sections. Timeline columns are neutral arithmetic on the scan's lead-time statements plus labelled forecasts.
| Company (ticker) | What it makes / does | Exposure to nuclear | Rough size | Position / edge |
|---|---|---|---|---|
| NuScale (SMR) | SMR designer; NRC-certified design | Pure-play; ~all (pre-revenue) | Small-cap est. | First NRC-certified US SMR; canceled first project — paper, not delivered |
| Oklo (OKLO) | SMR developer; build-own-operate, sell power | Pure-play; ~all (pre-revenue) | Small/mid-cap est. | Owner-operator model + signed customer pipeline; no operating reactor yet |
| BWX Technologies (BWXT) | US naval reactors; reactor components; SMR parts | Diversified; majority nuclear, much from Navy | Mid/large-cap; real revenue est. | Supplies essentially all US naval nuclear; certifications = hard-to-replace position; cash today |
| GE Vernova (GEV) | BWRX-300 SMR + turbines/generators + grid gear | Diversified; nuclear is a slice of a large power company | Large-cap; tens of $B revenue est. | Rare maker of both an SMR design and the turbine "power island" |
| Constellation Energy (CEG) | Operates largest US nuclear fleet; sells power | Operator; nuclear is core of generation | Large-cap; tens of $B revenue est. | Hard-to-replace running fleet; hyperscaler PPAs = recurring cash now |
| Curtiss-Wright (CW)* | Coolant pumps, instrumentation | Diversified; nuclear is a minority slice | Mid/large-cap est. | Supplies "every US nuclear plant"; aerospace/defense also large |
Source: 500-stocks scan key-companies list and notes. *CW listed in the scan's nuclear sub-section though not a headline ticker for this group. Sizes marked est. are general-knowledge order-of-magnitude, not live-verified.
An owner is buying very different cash shapes inside one label, so a single multiple (a price-to-revenue or price-to-earnings ratio) is misleading. In plain money terms:
Money-in / money-out shape of the group: nuclear is capital-heavy by nature, but the burden falls unevenly. SMR and OKLO are cash-consuming today — capex (capital expenditure: money spent building long-lived assets) and development spend with no offsetting revenue. BWXT and GEV are capex-heavy but cash-generating — they fund factories and qualification but get paid as they deliver. CEG is the most clearly owner-cash-generating now: the reactors are already built and largely paid for, so incremental high-priced power flows toward free cash flow (cash left after running and maintaining the business). The group therefore spans the full spectrum from pure cash-burn optionality to established free-cash-flow assets, and the "price of exposure" question is really which point on that spectrum you are buying. Where each name sits is a fact for the reader to weigh.
This is a pointer to where company-level work is most informative, not a recommendation on any name.
Used:
Hard facts (grounded in the scan or well-known filings): SMR and OKLO are pure-play SMR developers and largely pre-revenue; BWXT makes US naval reactors and nuclear components; GEV builds the BWRX-300 SMR and turbine islands; CEG operates the largest US nuclear fleet; no new US commercial reactor since Vogtle; the stated supply bottleneck is forgings/welders/ASME components/fuel and multi-year licensing.
Approximate and NOT live-verified (general knowledge, treat as order-of-magnitude): the GW demand figures and 2030 trajectory, the uranium demand reference, any TAM/growth/market-share percentage, all market-cap/revenue size bands marked est., and SMR deployment dates. Live web retrieval was unavailable when this was built; confirm every number against current filings and live quotes before acting.