A DC Atlas Frontier
Module 3 of 6
The Vendor Landscape
Design certification is not a construction licence, and an MOU is not a contract. A tiered read of the six vendors with near-term credibility, and what the hyperscalers have actually signed.
Built on DC Atlas's asset-level data across ~8,000 tracked facilities.
In October 2024, Kairos Power and Google signed a binding agreement to deploy 500 megawatts of advanced nuclear capacity by 2035. Seven months earlier, NuScale Power had laid off 28 percent of its workforce after its only US project collapsed with subscriptions stuck at 26 percent of capacity. Both companies describe themselves as leaders in the small modular reactor industry. Both statements are technically accurate. The distance between them is the entire subject of this module.
The SMR landscape is crowded with announcements. Memoranda of understanding multiply weekly. Capacity figures denominated in gigawatts appear in press releases from companies that have never generated a single electron of commercial electricity. For the infrastructure investor, the PE fund partner, and the data centre operator trying to underwrite a 20 year power commitment, the only question that matters is not who has the most impressive slide deck but who can actually deliver a functioning reactor on a timeline that aligns with when the data centre needs power. This module provides the framework for answering that question, applied vendor by vendor to the six companies with genuine near term credibility.
How to Read an SMR Claim
Before examining any individual vendor, it is worth establishing the framework that separates credible progress from aspirational marketing. The SMR industry operates along a development pipeline with distinct stages, and the gap between each stage is measured not in months but in years and billions of dollars.
The first distinction is between design certification and a construction licence. NuScale received NRC design certification in January 2023, making it the first and still the only SMR to achieve that milestone in the United States. But design certification means the NRC has approved the reactor design in the abstract. It does not mean anyone has permission to build one at a specific site. A construction licence requires a site specific application, environmental review, safety evaluation, and public hearing process that typically adds three to five years to the timeline. Confusing the two is one of the most common errors in SMR investment analysis.
The second distinction is between a memorandum of understanding and a binding contract. An MOU is a statement of mutual interest. It creates no financial obligations, no delivery commitments, and no legal consequences for failure to perform. A binding power purchase agreement or equipment supply contract creates enforceable obligations with defined payment terms, delivery schedules, and termination provisions. When a vendor announces a "partnership" or "agreement" denominated in gigawatts, the first question is always whether anyone has signed a document that requires them to write a cheque.
The third distinction is between non nuclear construction and nuclear construction. When a company announces it has "broken ground" on a reactor project, it typically means site preparation, access roads, support buildings, and foundations. Nuclear construction, the pouring of safety related concrete for the reactor building itself, is a fundamentally different activity subject to NRC oversight and quality assurance requirements. The gap between breaking ground on a support building and pouring nuclear concrete can be two to three years.
These three distinctions, design versus site licence, MOU versus binding contract, and non nuclear versus nuclear construction, are the filter through which every claim in this module should be read.
Which SMR Vendors Are Actually Credible?
Based on regulatory milestones achieved, physical construction progress, and binding commercial commitments as of February 2026, the six vendors with genuine near term relevance occupy distinct positions on the execution spectrum.
| Vendor | Reactor | Output | Tier | Key Milestone |
|---|---|---|---|---|
| GE Hitachi / OPG | BWRX-300 | 300 MW | Tier 1 | Construction licence (Apr 2025), building underway |
| TerraPower | Natrium | 345 MW | Tier 1.5 | NRC safety review complete (Dec 2025) |
| Kairos Power | KP-FHR | 50 MW demo | Tier 2 | Nuclear concrete poured (May 2025), Google binding PPA |
| NuScale | VOYGR | 77 MW/module | Tier 2 | Only NRC-certified SMR, Romania FID approved (Feb 2026) |
| Oklo | Aurora | 75 MW | Tier 2.5 | NRC readiness assessment cleared (Jul 2025) |
| X-energy | Xe-100 | 80 MW | Tier 3 | Amazon backing, no construction permit yet |
GE Hitachi BWRX-300: The Front Runner
The BWRX-300 at Ontario Power Generation's Darlington site in Ontario, Canada holds the distinction of being the furthest advanced grid scale SMR construction project in the Western world. The Canadian Nuclear Safety Commission issued the licence to construct on April 4, 2025, the first time Canada or any Western regulator has approved construction of a grid scale small modular reactor. Construction began in May 2025 with commercial operation targeted for end of 2030.
The project benefits from an institutional credibility that no other SMR programme can match. OPG completed the refurbishment of the existing Darlington nuclear generating station in February 2026, delivered four months ahead of schedule and CAD 150 million under budget after a decade of execution. That track record, managing 1,920 fuel channel replacements and 3,840 feeder pipe replacements on time and under budget, provides direct evidence that the organisation executing the SMR project can deliver complex nuclear construction.
“OPG completed the Darlington refurbishment four months ahead of schedule and C$150 million under budget — the strongest institutional credibility signal in the SMR sector.”Ontario Power Generation, February 2026
The Darlington SMR project operates under an integrated project delivery model that is unprecedented in the sector. OPG leads as licence holder and operator. GE Hitachi supplies the reactor technology and key components, including the reactor pressure vessel for which a manufacturing contract was awarded in January 2025. SNC Lavalin serves as architect engineer. Aecon leads construction. This four party alliance is designed to reduce risk, streamline decisions, and prevent the cost and schedule overruns that have historically plagued nuclear construction.
The broader programme envisions up to four BWRX-300 units at Darlington, generating 1,200 megawatts when fully built out. Units two through four are projected for service between 2034 and 2036. The first unit budget is approximately CAD 6.1 billion for 300 megawatts. OPG projects a 33 percent cost reduction between the first and fourth units, which would bring the fourth unit to roughly CAD 4.1 billion, a trajectory consistent with the learning curve economics discussed in Module 2.
TerraPower Natrium: The Closest US Project
TerraPower's Natrium demonstration reactor at Kemmerer, Wyoming represents the most advanced physical construction of any advanced reactor project in the United States. The company broke ground on non nuclear construction in June 2024, making Kemmerer the first advanced reactor project in the country to move from design into construction. Non nuclear construction has continued since, with the Test and Fill Facility rising visibly at the 60 acre site and steel beams for the interior erection crane installed in 2025.
The NRC delivered a significant regulatory milestone on December 3, 2025, completing the final safety evaluation for the construction permit application one month ahead of an already accelerated schedule. TerraPower expects to receive its construction licence from the NRC in early 2026. The full projected timeline specifies first nuclear related concrete pour by 2027, fuel loading by 2030, and commercial operation by 2031.
The Natrium design is a 345 megawatt sodium cooled fast reactor. It is coupled with a molten salt energy storage system. That storage can boost output to 500 megawatts. The boost is sustained for more than five and a half hours when grid demand peaks. This built in energy storage capability is unique among the SMR designs under development and enables integration with high renewable penetration grids, a feature that no other reactor in this landscape offers.
The critical execution dependency for Kemmerer is the HALEU fuel supply chain. The Natrium reactor requires high assay low enriched uranium, which until 2024 was sourced almost exclusively from Russia. Centrus Energy announced in January 2026 a $900 million DOE task order to expand its Piketon, Ohio enrichment facility to commercial scale HALEU production, with first new capacity expected online in 2029, precisely aligning with TerraPower's 2030 fuel loading target. TerraPower has further de risked its fuel supply through agreements with ASP Isotopes for a South African HALEU facility, Framatome for a HALEU metallisation plant, and a fuel fabrication facility under development at Global Nuclear Fuel Americas in Wilmington, North Carolina.
Kairos Power: Google's Bet
Kairos Power occupies a unique position in the vendor landscape as the only company with both a binding commercial agreement from a hyperscaler and physical nuclear construction underway. The October 2024 Master Plant Development Agreement with Google is not an MOU. It is a binding agreement establishing power purchase agreements for 500 megawatts of advanced nuclear capacity by 2035, with initial deployment scheduled for 2030. Google's Senior Director of Energy and Climate described it as providing "a strong customer demand signal" through "milestone based accountability baked into the agreement."
The construction progress is real. On May 7, 2025, Kairos poured the first nuclear safety related concrete for the Hermes Low Power Demonstration Reactor in Oak Ridge, Tennessee. This involved installation of 51 drilled piers, each six feet in diameter and extending 40 feet below ground, to anchor the reactor building foundation to bedrock. Hermes holds the distinction of being the first non water cooled reactor approved for construction in the United States in more than 50 years, with the NRC construction permit issued in December 2023.
In August 2025, Kairos, Google, and the Tennessee Valley Authority announced a power purchase agreement for the Hermes 2 plant, uprated from 28 to 50 megawatts, with TVA becoming the first US utility to sign a PPA for electricity from a Generation IV reactor. The fact that TVA, one of the country's largest and most experienced nuclear operators, independently validated the economics and technology through a binding purchase agreement provides credibility that no amount of corporate announcements can substitute.
Kairos uses a fluoride salt cooled high temperature reactor design with TRISO particle fuel, operating at low pressure with inherent passive safety. The company has pursued aggressive vertical integration, fabricating its first reactor vessel in house at its Albuquerque manufacturing campus in January 2025, with a target of 80 percent of component costs derived from raw materials or commercial off the shelf parts. This manufacturing strategy is designed to compress costs and timelines by controlling the supply chain directly rather than depending on nuclear grade component suppliers.
NuScale: Certified but Chastened
NuScale Power holds a distinction that no other SMR vendor can claim. It is the first and only company to receive full NRC design certification for a small modular reactor in the United States, achieved in January 2023 for the 50 megawatt configuration and extended to the uprated 77 megawatt design in May 2025. This regulatory achievement is real, significant, and insufficient on its own to guarantee commercial success.
The CFPP collapse is the cautionary tale that every SMR investor must understand. The Carbon Free Power Project, a partnership with the Utah Associated Municipal Power Systems to build six 77 megawatt modules at Idaho National Laboratory, was NuScale's flagship US deployment. The project's cost estimate escalated from $3 billion to $9.3 billion. The target electricity price rose from $55 per megawatt hour to $89. Member utilities withdrew steadily. Subscriptions stood at just 120 megawatts against a 462 megawatt target. That was 26 percent of the capacity needed to proceed. NuScale and UAMPS mutually terminated the project in November 2023. The stock fell 75 percent from its late 2022 highs. NuScale laid off 154 employees, 28 percent of the workforce.
The $55 and $89 figures are the project's stated target electricity prices, before and after its 2023 revision. Both are published; no later per-MWh target was published before termination.
The honest post mortem reveals site specific factors that do not necessarily condemn the technology. The Idaho National Laboratory location created extraordinary transportation challenges for shipping factory fabricated modules. NuScale CEO John Hopkins acknowledged that "getting those modules up to that location" would have been "a real challenge" and that "another location in the southeast, if you can barge them in, it significantly lowers the cost." The lesson is not that NuScale modules are too expensive to build. It is that first of a kind projects at challenging sites with a fragmented customer base of 50 municipal utilities are structurally prone to cost escalation and subscription failure.
The recovery trajectory has been dramatic. In September 2025, NuScale's partner ENTRA1 Energy announced a landmark agreement with the Tennessee Valley Authority to deploy up to 6 gigawatts of SMR capacity across TVA's seven state service region, the largest SMR deployment programme in US history. As of September 30, 2025, NuScale held $753.8 million in cash and investments with zero debt, providing approximately five to six years of runway at current burn rates.
Romania provides the counterbalancing signal. In February 2026, Romanian state owned nuclear operator Nuclearelectrica approved the Final Investment Decision for the Doicesti SMR project at a former coal plant site. The plant will use six NuScale 77 megawatt modules. Those modules generate 462 megawatts in total. The project carries a total investment of $6 to $7 billion. US government commitments cover approximately $4 billion of that, through Exim Bank and the Development Finance Corporation. First module commercial operation is targeted for 2033. This is NuScale's most advanced path to commercial electricity generation and represents genuine international de risking of the technology.
Oklo: The Pivot
Oklo's trajectory is the most complex narrative in the SMR landscape. The company's original combined licence application for its Aurora reactor was denied by the NRC in January 2022, with the regulator citing "significant information gaps" in the safety analysis and noting that Oklo had "repeatedly failed to submit the information needed" to complete the review. This denial, while without prejudice, was a credibility damaging outcome in regulatory circles.
Rather than immediately resubmitting through the traditional NRC pathway, Oklo executed a strategic pivot to the Department of Energy's newly established Reactor Pilot Programme, created through executive orders signed in May 2025. In July 2025, the NRC completed a pre application readiness assessment for Oklo's Aurora project with "no significant gaps identified that would hinder acceptance of the application," a marked improvement from the 2022 denial. The DOE subsequently selected Oklo for its advanced nuclear fuel line pilot projects programme in September 2025 and approved the preliminary safety analysis for Oklo's Aurora fuel fabrication facility in early 2026.
The investment case for Oklo rests on its market capitalisation relative to its development stage. As of early 2026, Oklo trades at approximately $10.57 billion in market capitalisation on zero commercial revenue and no operating reactors. This valuation prices in substantial future success. Meta announced agreements with Oklo for up to 1.2 gigawatts of Aurora capacity in Pike County, Ohio as part of its January 2026 nuclear procurement. Amazon has also backed Oklo's deployment plans. The hyperscaler demand signal is genuine. The question is whether Oklo can navigate either the NRC licensing process or the DOE Reactor Pilot Programme fast enough to meet the timelines these customers require.
The Hyperscaler Scorecard
The most reliable signal in the SMR landscape is not what vendors claim about themselves but what the world's largest technology companies are willing to commit in binding dollars. The hyperscaler nuclear commitments announced between 2024 and early 2026 represent the largest private sector pivot toward nuclear energy since the technology was commercialised.
| Company | Partner | Capacity | Structure | Value / Duration |
|---|---|---|---|---|
| Amazon | Talen Energy | 1,920 MW | Binding PPA (front-of-meter) | $18B / 17 years |
| Amazon | X-energy | 5 GW (SMR pipeline) | Investment + development | Multi-billion backing |
| Amazon | Energy Northwest | SMR Phase 1 | Development funding | Undisclosed |
| Microsoft | Constellation | 835 MW (TMI restart) | Binding PPA | 20 years, $1B DOE loan |
| Kairos Power | 500 MW | Binding Master PDA | By 2035 | |
| Elementl Power | 600+ MW per site | Site preparation | 3+ sites | |
| Meta | Vistra | 2.1 GW (existing Ohio nuclear) | PPA | Undisclosed |
| Meta | TerraPower | 2.8 GW (Natrium) | Development agreement | Up to 8 reactors |
| Meta | Oklo | 1.2 GW (Aurora) | Development agreement | Pike County, Ohio |
| Meta | Constellation | 1.1 GW (Clinton plant) | Binding PPA | 20 years |
The Amazon and Talen Energy story deserves particular attention because it illustrates exactly how the regulatory landscape is evolving. The original arrangement was a behind the meter colocation deal, with Amazon's data centre drawing power directly from the Susquehanna nuclear plant. FERC rejected this model twice, in November 2024 and again on rehearing in April 2025, finding that it shifted transmission costs onto other ratepayers. Rather than wait for a Fifth Circuit appeal, Talen and Amazon restructured the deal entirely into a front of meter retail supply arrangement. The result was an $18 billion, 17 year power purchase agreement with deliveries ramping from 840 to 1,200 megawatts by 2029 and 1,680 to 1,920 megawatts by 2032.
“This agreement creates a platform to expand across the Talen portfolio.”Mac McFarland, CEO, Talen Energy
FERC's subsequent December 2025 colocation order, approved unanimously 5 to 0, established the first coherent national framework for data centre colocation at power plants. The order directed PJM to create three new transmission service options for colocated loads and reformed behind the meter generation rules. Commissioner Rosner's concurrence captured the regulatory logic: "If a new large load wants to connect directly to a power plant and operate in a way that lowers grid costs, we should let it."
The Illinois Opening
The regulatory landscape shifted decisively on January 8, 2026, when Illinois Governor JB Pritzker signed Senate Bill 25, the Clean and Reliable Grid Affordability Act. The legislation lifted Illinois's 30 year moratorium on new nuclear development, effective June 1, 2026, permitting both conventional large scale reactors and advanced SMRs for the first time since the mid 1990s.
The timing is not incidental. Illinois operates 11 commercial reactors across six sites, the largest nuclear fleet of any US state, generating 11,864 net megawatts. Constellation Energy operates all of them. The state possesses the workforce, the grid infrastructure, the regulatory expertise, and the cooling water access that SMR deployment requires. What it lacked, until January 2026, was legal permission to build new reactors.
The market response has been immediate. Meta signed a 20 year PPA with Constellation for the entire output of the Clinton Clean Energy Center in Illinois, between 1,092 and 1,121 megawatts after planned uprates. Nano Nuclear Energy signed an MOU with the University of Illinois Urbana Champaign to construct and operate the first Kronos micro modular reactor on a US university campus. Argonne National Laboratory projects that data centre load in Illinois will reach 18 terawatt hours by 2030, doubling to 36 by 2039 and 72 by 2050.
The broader Midwest is following Illinois's lead. Michigan is pursuing Holtec SMR development at Palisades with $300 million in state funding and a $1.52 billion federal loan guarantee. Indiana partnered with First American Nuclear for a 240 megawatt liquid metal fast reactor. Wisconsin advanced legislation ranking nuclear as the state's second priority energy resource after efficiency, above both renewables and fossil fuels. Ohio is emerging as a hub for Meta's TerraPower Natrium and Oklo Aurora deployments.
When Can You Actually Get Nuclear Power?
The vendor landscape as of February 2026 produces a clear hierarchy for anyone trying to underwrite a nuclear powered data centre with a specific operational date.
| If You Need Power By | Your Realistic Options | Key Risk |
|---|---|---|
| 2028 to 2029 | Existing plant PPA (Talen/Susquehanna model, Constellation/Clinton model) | FERC colocation rules still evolving |
| 2030 to 2031 | BWRX-300 (if OPG model replicated), TerraPower Natrium | First-of-a-kind construction risk |
| 2032 to 2034 | NuScale (Romania or TVA), Kairos commercial fleet | Binding contracts not yet in place for all |
| 2035+ | Full SMR fleet deployment, Oklo Aurora at scale | Technology and regulatory pathway risk |
The vendor landscape is not a horse race. It is a portfolio of options at different stages of maturity, suited to different risk tolerances and timeline requirements. The operator who needs power in 2029 will pursue an existing plant PPA. The operator planning a campus for 2032 will evaluate NuScale or Kairos. The sovereign wealth fund making a 20 year bet on the sector will want exposure across the tier spectrum. Module 4 builds the economic model that determines which of these options makes financial sense, and the answer is less obvious than you might expect.
Questions this module answers
- Which SMR vendor is furthest along?
- GE Hitachi's BWRX-300 at Ontario Power Generation's Darlington site is the furthest advanced grid scale SMR construction project in the Western world. The Canadian Nuclear Safety Commission issued the licence to construct on April 4, 2025, construction began in May 2025, and commercial operation is targeted for end of 2030.
- Is NuScale's design certification the same as permission to build?
- No. NuScale received NRC design certification in January 2023 — the first and still the only SMR to achieve it — but that approves the reactor design in the abstract, not construction at any specific site. A construction licence requires a site specific application, environmental review, safety evaluation and public hearing process that typically adds three to five years to the timeline.
- Is NuScale's 6 GW agreement with the TVA a binding contract?
- No. The TVA/ENTRA1 agreement to deploy up to 6 gigawatts of SMR capacity is a non-binding memorandum of understanding, not a firm power purchase agreement. NuScale's SEC filings confirm that binding PPA execution and equipment supply contracts are separate, not-yet-achieved milestones.
- Why did NuScale's Carbon Free Power Project fail?
- Cost escalation destroyed subscriber demand. The cost estimate rose from $3 billion to $9.3 billion and the target electricity price rose from $55 per megawatt hour to $89, leaving subscriptions at just 120 megawatts of the 462 megawatt target — 26 percent of what was needed to proceed — and NuScale and UAMPS mutually terminated the project in November 2023.
- What is the largest corporate nuclear commitment to date?
- Amazon's power purchase agreement with Talen Energy: an $18 billion, 17 year binding contract for up to 1,920 megawatts, with deliveries ramping from 840 to 1,200 megawatts by 2029 and 1,680 to 1,920 megawatts by 2032. It was restructured into a front of meter retail supply arrangement after FERC rejected the original behind the meter colocation model.