01 — MethodologyHow to read these numbersThis guide benchmarks the installed capital cost of new US nuclear capacity in dollars per kilowatt of nameplate capacity — the number owners, lenders, and hyperscaler offtakers use to compare an SMR program against gas, renewables-plus-storage, or a plant restart. Because no US small modular reactor has completed construction, the dataset is necessarily a mix: one audited actual (Vogtle 3 and 4), one fully developed estimate that collapsed on contact with procurement reality (the NuScale/UAMPS Carbon Free Power Project), published budgets for programs now in construction or licensing, and DOE’s modeled cost curves.Sources and vintages: Vogtle figures come from Georgia PSC filings and industry reporting through 2024; the CFPP figure is the January 2023 NuScale/Fluor Class 3 estimate analyzed by IEEFA; the Darlington program budget is Ontario Power Generation’s 2025 disclosure (Canadian dollars, all-in); Natrium figures are from TerraPower and DOE (2024—2026); and the FOAK/NOAK curve is from DOE’s Pathways to Commercial Liftoff: Advanced Nuclear report (March 2023 — the September 2024 update does not restate those two figures). All figures retrieved August 2026 and rounded.One basis warning before the numbers: overnight cost excludes financing and escalation, while all-in cost includes them — the same plant can differ 30% or more between the two framings. Darlington reports all-in in Canadian dollars; DOE’s curves are overnight in US dollars; federal cost share lowers an owner’s net outlay without lowering the build cost. We label the basis on every figure.What these figures are — and are notThese figures are a market map, not a quote. They mix three kinds of numbers: audited actuals from the one completed project (Vogtle), estimates published before cancellation (the NuScale CFPP), and budgets for plants now in construction or licensing (Darlington, Natrium, Clinch River). Cost bases differ — some are overnight capital costs, others are all-in figures that include financing, inflation, and shared site infrastructure — and one program reports in Canadian dollars. What they are good for: setting a realistic planning band for owners, lenders, and EPCs, and understanding why first units run two to four times DOE’s long-run targets. What they are not: evidence that any US SMR has been built to budget — none has been built at all, and every SMR figure here is an estimate or a budget, not an actual.02 — At a glanceThe benchmarks at a glanceThe whole guide in one screen. Each row jumps to the full section.01Vogtle 3 and 4 (completed 2023—24)~$35B all-in for 2.2 GW — roughly $15,000 to $16,500/kW, the only US actualActual02NuScale/UAMPS CFPP (cancelled Nov 2023)$9.3B estimate for 462 MW — $20,139/kW at cancellation, up 75% from $5.3BCancelled03OPG Darlington BWRX-300 (4 units)C$20.9B program — ~C$17,400/kW all-in; first unit C$6.1B plus C$1.6B shared infraBuilding04TerraPower Natrium (Kemmerer, WY)Up to $4B for 345 MW — ~$11,600/kW, with roughly $2B in DOE cost shareBuilding05DOE Liftoff cost curve~$6,200/kW well-executed FOAK falling to ~$3,600/kW NOAK after 10 to 20 buildsTargetThe anchor figures: completed US nuclear cost about $15,000/kW, the first fully priced US SMR was cancelled near $20,000/kW, committed North American programs cluster between roughly $11,600 and C$17,400/kW all-in, and DOE’s serial-build target is $3,600/kW — a 5x spread the next decade of first units will resolve one direction or the other.
03 — BenchmarksCost per kW: the 2026 benchmark setSix numbers frame the market. Read them as a maturity curve: what nuclear construction actually cost, what the first US SMR was priced at, what committed programs have budgeted, and where DOE says the industry has to land for the order books to make sense.$15,000+Vogtle 3 and 4 actual, per kWRoughly $35B all-in for two AP1000s (about 2.2 GW), per Georgia PSC filings and industry reporting — about $15,000 to $16,500/kW. The only US reactors built from scratch in decades, and the only modern new-build benchmark.$20,139NuScale CFPP estimate at cancellation$9.3B for 462 MW (six 77 MW modules) in the January 2023 Class 3 estimate (UAMPS, via IEEFA), up 75% from $5.3B; the project was cancelled in November 2023 before construction began.C$17,400Darlington BWRX-300 program, per kWOPG’s C$20.9B budget for four 300 MW units — all-in, including financing and inflation; roughly US$12,500/kW at mid-2026 rates. First unit plus shared infrastructure: C$7.7B (OPG, 2025).$11,600Natrium budget, per kWTerraPower’s Kemmerer plant: up to $4B for 345 MW, with roughly $2B of DOE ARDP cost share; NRC construction permit issued March 2026, nuclear construction began April 2026 (TerraPower, DOE).$6,200DOE FOAK benchmark, per kWDOE’s overnight-cost estimate for a well-executed first-of-a-kind nuclear build, from Pathways to Commercial Liftoff: Advanced Nuclear (March 2023; the September 2024 update does not restate this figure).$3,600DOE NOAK target, per kWWhere overnight costs need to land after roughly 10 to 20 deployments of a standardized design for nuclear to compete broadly, per the same DOE Liftoff analysis.Sources: Georgia PSC filings and POWER/industry reporting (Vogtle 3 and 4); UAMPS Class 3 estimate via IEEFA, Jan 2023 (CFPP); Ontario Power Generation budget disclosure, 2025 (Darlington); TerraPower and US DOE (Natrium); US DOE Pathways to Commercial Liftoff: Advanced Nuclear, Mar 2023 and Sept 2024 update (FOAK/NOAK, workforce); US NRC and TVA (Clinch River) (retrieved Aug 2026)FOAK vs NOAK: the cost curve the whole thesis depends onFirst-of-a-kind to nth-of-a-kind compression is the entire SMR investment case: DOE models overnight costs falling from about $6,200/kW on a well-executed first build to about $3,600/kW after 10 to 20 deployments of the same design — a 42% reduction from completed engineering, supply-chain maturity, and crew learning. The load-bearing word is well-executed: Vogtle and the CFPP both priced far above the FOAK line, which is why every current program leans on standardization, multi-unit siting, and federal cost share to hold it. 04 — Cost driversWhat moves the number: six drivers behind the $3,600-to-$20,000 spreadThe spread between DOE’s target and the CFPP’s cancellation price is not noise — it decomposes into a handful of drivers an owner can actually underwrite.+75%Estimate drift before steelThe CFPP estimate rose from $5.3B to $9.3B in roughly two years on commodity inflation and interest rates — before major construction. That is the classic FOAK risk profile (UAMPS/IEEFA).+30-60%All-in vs overnight basisFinancing, escalation, and owner costs stack on top of overnight capital. Darlington’s C$20.9B explicitly includes interest and inflation across a decade-long build (OPG).-42%Serial-build learningDOE models compression from ~$6,200 to ~$3,600/kW after 10 to 20 deployments of a standardized design — the learning rate only accrues if the design stops changing.+$1-3kCraft labor and reworkVogtle’s overrun was driven partly by labor availability, productivity, and nuclear-grade rework. Scarce nuclear-qualified weld and QC talent shows up directly in $/kW.-25-50%Cost share and creditsDOE is cost-sharing roughly $2B at Natrium and selected TVA for $400M for Clinch River (half of an $800M package shared with Holtec); IRA credits improve the owner’s net economics without lowering the build cost itself.4 unitsMulti-unit sitingDarlington spreads C$1.6B of shared infrastructure — cooling tunnels, admin, switchyard — across four units. A single-unit program carries the same overhead alone (OPG).Net: treat any single $/kW quote as a point on a distribution, not a price. The controllable variance — labor productivity, rework rates, schedule discipline — concentrates in the workforce, which is why craft strategy belongs in the pro forma, not the appendix. 05 — VariationThe program-by-program picture, August 2026Only published figures appear below; where a developer has not disclosed capex, we say so rather than guess. Every US entry is pre-completion — these are budgets and milestones, not actuals.GE Hitachi BWRX-300 — Darlington (OPG)C$20.9B / 4 unitsFirst unit C$6.1B plus C$1.6B shared infrastructure; ~C$17,400/kW all-in across the 1,200 MW program; unit 1 targets 2030 grid connection.GE Hitachi BWRX-300 — TVA Clinch River$400M DOE awardNRC staff recommended a construction permit in June 2026 — the first for a commercial grid-scale SMR in the US; TVA has not published a final EPC cost for the 300 MW unit.TerraPower Natrium — Kemmerer, WYUp to $4B / 345 MW~$11,600/kW with roughly $2B DOE cost share; nuclear construction began April 2026 with Bechtel and a planned ~1,600-strong site workforce.X-energy Xe-100 — Dow Seadrift, TX4 x 80 MWConstruction permit under NRC review since 2025; DOE ARDP cost share up to ~$2.15B after an additional award in August 2026; Amazon holds options on 5+ GW of Xe-100 capacity by 2039.Kairos Power Hermes 2 — Oak Ridge, TN50 MW demoBroke ground April 2026; first deployment under Google’s 500 MW order book, with TVA as the purchasing utility; operations targeted for 2030.NuScale VOYGR — CFPP (cancelled)$9.3B est.Six 77 MW modules, 462 MW; the $20,139/kW estimate preceded the November 2023 cancellation — still the market’s clearest FOAK price signal.The honest read: one North American SMR program has a hard, government-approved budget (Darlington), one US program has a stated cap with federal cost share (Natrium), and the rest are pre-construction. Expect first US unit costs to print somewhere between DOE’s $6,200/kW FOAK line and Darlington’s all-in number — and treat anything quoted below $5,000/kW for a first unit with skepticism. 06 — TrendThe trend: what actually pulls costs downThe biggest change since 2023 is who is paying. Google has contracted for 500 MW of Kairos capacity by 2035, with TVA buying Hermes 2 output for data centers in Tennessee and Alabama; Amazon invested directly in X-energy and holds options on more than 5 GW of Xe-100 capacity by 2039; and Microsoft’s 20-year PPA behind the Crane (Three Mile Island) restart showed hyperscalers will pay a premium for firm carbon-free power. Contracted offtake converts first-of-a-kind risk into financeable revenue — the single biggest lever on cost of capital, which is itself a large share of all-in $/kW.Capital is following the contracts: X-energy closed a $700M Series D and raised roughly $1B in its 2026 IPO, TerraPower’s $4B program pairs ~$2B of DOE cost share with private capital from Gates, Nvidia’s venture arm, and HD Hyundai, and TVA was selected in December 2025 for a $400M DOE award — half of an $800M package shared with Holtec — to pull Clinch River’s schedule left. Layered on that, the structural cost reducers are boring and proven: freeze the design, repeat it, share infrastructure across multiple units, and keep the same crews building the same thing.The caveat: the curve only bends if first units hold schedule. Every month of slip adds interest and escalation to the all-in figure, and slip on nuclear projects has historically come from labor productivity, weld rework, and quality documentation — not from reactor physics. That puts workforce execution, not technology, at the center of whether the $6,200-to-$3,600 trajectory materializes. 07 — WorkforceThe workforce variable inside every $/kW figureEvery number above embeds a labor assumption. DOE estimates the US needs roughly 375,000 additional workers to build and operate 200 GW of new nuclear by 2050, against an industry workforce of about 100,000 today — and construction craft is the sharpest constraint. Nuclear-qualified welders, pipefitters, electricians, and QC inspectors work under NQA-1 quality culture, where the documentation is as load-bearing as the weld itself; Vogtle showed exactly what happens to $/kW when that talent is scarce and rework compounds.That scarcity is a planning problem owners can get ahead of. iRecruit’s ground is the construction talent that builds heavy industrial and power projects — project managers, superintendents, QA/QC leadership, and the craft pipelines behind them — and we are assembling candidate density in nuclear-adjacent skill sets ahead of the build wave, so that when programs like Clinch River and Seadrift mobilize, the market map already exists. Owners who treat workforce as critical-path procurement rather than a staffing afterthought will be the ones who print a $/kW their board can defend.For the hiring side of this market, see the Nuclear Construction practice. 08 — FAQFrequently asked questions What did Vogtle 3 and 4 actually cost per kW?+Total project cost is widely reported at roughly $35B for about 2.2 GW of new capacity — approximately $15,000 to $16,500 per kW all-in, per Georgia PSC filings and industry reporting. It remains the only completed US new-nuclear datapoint.Has any US small modular reactor completed construction?+No. TerraPower’s Natrium began nuclear construction in April 2026, Kairos broke ground on Hermes 2 the same month, and TVA’s Clinch River unit received an NRC staff construction-permit recommendation in June 2026. Every SMR cost-per-kW figure is therefore an estimate or a budget, not an actual.Why does the cancelled NuScale project still anchor SMR cost discussions?+The Carbon Free Power Project was the only US SMR to publish a fully developed price: $9.3B for 462 MW, or $20,139/kW, up 75% in about two years (UAMPS/IEEFA). Its November 2023 cancellation is the clearest evidence of how far first-of-a-kind estimates can drift before construction even starts.What is the difference between FOAK and NOAK cost?+FOAK (first-of-a-kind) is the cost of the initial build of a design, carrying incomplete engineering, immature supply chains, and green crews; NOAK (nth-of-a-kind) is the settled cost after repeated builds. DOE’s Liftoff analysis puts a well-executed FOAK near $6,200/kW overnight and the NOAK target near $3,600/kW after 10 to 20 deployments.How does workforce scarcity show up in cost per kW?+Through productivity, rework, and schedule slip. Nuclear-grade welding and NQA-1 quality work require certifications few craft workers hold, and DOE projects a need for roughly 375,000 additional nuclear workers by 2050 — so programs that secure qualified craft and QC leadership early protect both the schedule and the all-in $/kW. Talent ahead of the build waveThe SMR decade will be won by owners who secure nuclear-qualified craft early. iRecruit is assembling candidate density in the trades that build heavy power projects — map the talent market before your program needs it.Scope a search →Candidates — register with us →
