AI + Advanced Nuclear: What 2025 Funding Says Next

Nuclear fission startups pulled in $1.3 billion in equity funding by early Q3 2025—the highest annual total on record, and nearly 40% of all fission equity since 2020. That’s not “nice to see.” It’s a signal: investors believe advanced nuclear is shifting from slide decks to poured concrete.

Three December closes made that signal louder: Radiant (more than $300M), Last Energy (more than $100M, oversubscribed), and ARC Clean Technology (Series B closed). Each is aimed at a near-term milestone—manufacturing scale-up, a pilot reaching criticality, or regulatory progress toward a first-of-a-kind deployment.

For utilities and large energy users, this matters for one reason: advanced reactors don’t become valuable when they’re announced—they become valuable when they’re operated. And operation is exactly where the “AI in Energy & Utilities” story intersects with microreactors and SMRs. If you’re serious about firm clean power, the hard part isn’t only reactor physics. It’s integration, reliability, maintenance, and regulatory-grade evidence. That’s AI’s home turf.

Why 2025 advanced nuclear funding is different

Answer first: The 2025 funding spike is different because it’s tied to defined demonstrations, manufacturing plans, and clearer licensing pathways, not just R&D optimism.

Net Zero Insights counted 28 equity transactions by October 2025, up from a historical average of about 15 per year. Even more telling: SMRs and microreactors represented ~75% of total fission funding. Investors are leaning toward reactors that can be manufactured in repeatable ways and sold into clear early markets—data centers, defense, remote industry, and industrial heat.

What I like about this moment is that it’s more operationally grounded than past nuclear hype cycles. We’re seeing:

• Factory-first language (build in a controlled facility; ship modules)
• Concrete schedules (tests in 2026, early deployments by 2028)
• Market pull (deposits, leases, defined sites)
• Regulatory sequencing (vendor design reviews, DOE pilot programs)

That’s where utilities should pay attention. Not because every timeline will hit, but because the commercialization playbook is getting clearer.

A quick reality check on venture funding and nuclear timelines

Answer first: A late-stage round doesn’t mean “risk-free,” but it does mean investors see enough maturity to fund expensive next steps like licensing, supply chain, and first builds.

Venture funding stages matter because they correlate with what a company can credibly execute:

• Series A: prove the plan and early customer pull
• Series B: scale teams, manufacturing readiness, and execution capacity
• Series C and beyond: expand deployment, markets, and supply chain depth

Advanced nuclear is capital-intensive. That makes milestone-driven financing a practical proxy for maturity. If a company raises a large round and immediately ties it to a test campaign, a factory, or a specific licensing path, that’s more meaningful than a generic “growth” announcement.

Three funding rounds, three commercialization paths

Answer first: Radiant, Last Energy, and ARC are taking different routes to the same destination—repeatable deployments that can sell into high-value reliability markets.

Radiant: portable microreactors built for repeatable manufacturing

Radiant closed more than $300 million and says it will use the funds to scale commercialization as it prepares to break ground in early 2026 on its R-50 factory in Oak Ridge, Tennessee.

Its Kaleidos design is a 1 MWe helium-cooled microreactor using TRISO fuel and passive cooling. Radiant is targeting a fueled test in 2026 at Idaho National Laboratory’s DOME facility, and it’s eyeing initial customer deployments starting in 2028.

Here’s the operationally relevant part: Radiant has emphasized customer commitments, including an agreement with a data center company to purchase 20 microreactors, plus defense-related agreements that position the technology for early deployments where reliability has a high price tag.

AI tie-in: Microreactors sold into data centers and critical infrastructure will live or die on uptime, fast fault detection, and evidence-rich operations. AI can support:

• Condition-based monitoring of balance-of-plant components (pumps, valves, heat exchangers)
• Anomaly detection for sensor drift and early warning signals
• Digital twin approaches that connect manufacturing QA data to operational performance

If you’ve ever run a complex asset fleet, you know the big win isn’t “automation.” It’s reducing unplanned downtime while proving to regulators and customers that your controls and maintenance are disciplined.

Last Energy: a familiar PWR architecture with a DOE pilot pathway

Last Energy announced an oversubscribed Series C of more than $100 million to complete its PWR-5 pilot at the Texas A&M–RELLIS Campus under the DOE’s Reactor Pilot Program, targeting criticality in 2026.

Last Energy’s commercial product is the PWR-20, a 20 MWe pressurized water reactor that uses standard PWR fuel (<4.95% enriched) and leans on widely understood PWR operating experience. It’s also designed for modularization: factory-built modules, compact footprint, and a stated goal of delivery in less than 24 months.

Whether you love that timeline or not, this is a straightforward bet: reduce novelty, reduce construction scope, and standardize deployment.

AI tie-in: Traditional PWR operations already generate huge volumes of operational data. The difference with modular fleets is scale: if you deploy dozens of similar units, AI becomes the practical way to manage consistency.

• Fleet learning: one unit’s early-life issues become every unit’s preventive actions
• Predictive maintenance scheduling: align outages with customer load profiles and grid constraints
• Automated reporting: faster generation of compliance-ready documentation for internal governance and external stakeholders

The best operators will treat AI as a discipline, not a software add-on: curated data pipelines, strict model governance, and human-in-the-loop procedures.

ARC Clean Technology: sodium fast reactor maturity plus Canadian licensing signals

ARC Clean Technology closed a Series B to accelerate commercialization of its 100 MWe ARC-100 sodium-cooled fast reactor derived from EBR-II heritage.

In Canada, ARC completed Phase 2 of the Canadian Nuclear Safety Commission’s Vendor Design Review in July 2025, supporting a proposed demonstration at Point Lepreau. Phase 2 is meaningful because it signals regulators see no fundamental barriers to licensing at that stage.

ARC’s design targets include a 20-year core life without refueling and a 60-year plant operating life.

AI tie-in: Fast reactors and high-temperature systems can benefit from AI-driven monitoring because:

• Temperature, flow, and materials behavior often require tight operational envelopes
• Early detection of off-normal signatures protects availability and extends component life
• Digital operational evidence helps bridge the trust gap for new reactor classes

For utilities considering non-light-water reactors, the commercial question is simple: Can you operate it predictably, maintain it efficiently, and prove performance in a regulator-friendly way? AI can strengthen all three.

The missing link: grid integration is where AI earns its keep

Answer first: SMRs and microreactors don’t automatically solve reliability—they shift the reliability problem into integration, dispatch, and asset management, where AI performs best.

Advanced nuclear is often discussed as “firm clean power.” True. But the grid doesn’t buy slogans; it buys services:

• Capacity and resource adequacy
• Frequency and voltage support
• Ramping and load-following behavior
• Black start and resilience capabilities (in some cases)

A microreactor behind the meter at a data center has different priorities than a 100 MWe unit supporting a regional grid. Either way, AI helps translate physical capability into grid value.

Practical AI use-cases utilities can start now

Answer first: You don’t need to wait for the first SMR to connect to start building AI capabilities that will make SMR integration cheaper and safer.

1. Interconnection and dispatch simulation

   • Train forecasting models on load and price patterns
   • Use optimization to evaluate when to run nuclear heat/power vs. import/export

2. Predictive maintenance for balance-of-plant

   • Start with rotating equipment, cooling systems, and electrical components
   • Build failure-mode libraries and baseline signatures now

3. Asset performance management (APM) with model governance

   • Treat models like safety-critical software: versioning, testing, audit trails
   • Document why models recommend actions (traceability matters)

4. Regulatory-grade data and reporting pipelines

   • Automated evidence collection reduces manual burden
   • Consistent data lineage improves credibility in oversight settings

If you’re running a utility AI program, this is a good time to align your roadmap with likely SMR adoption patterns: behind-the-meter reliability first, then fleet deployments, then broader grid service stacking.

What utilities and large energy users should do in 2026

Answer first: The next 12 months are for building readiness—commercial, technical, and organizational—so you’re not starting from zero when demonstrations start producing real operating data.

Here’s a pragmatic checklist I’d use if I were advising a utility, IPP, or large C&I buyer:

• Define your “firm clean power” use case: capacity replacement, data center growth, industrial heat, remote reliability, or resilience.
• Decide what you’ll standardize: interconnection studies, cybersecurity controls, SCADA integration patterns, maintenance workflows.
• Build an AI-ready data foundation: historian strategy, sensor standards, tagging conventions, and quality rules.
• Establish model governance early: who approves models, how they’re tested, and how outputs are logged.
• Plan for fleet operations, not one-off assets: even one pilot should be set up like it’s unit 1 of 20.

If you do only one thing: treat operational data as a product. Because for SMRs and microreactors, operational credibility becomes a sales engine.

A useful stance for 2026: “Prove it in operations, or it doesn’t count.” The companies that win will be the ones that can prove reliability at scale.

The bet investors are making—and the bet utilities should make

Advanced nuclear funding in 2025 is a bet that repeatable manufacturing plus clearer licensing plus real customer demand can finally compress the timeline between invention and deployment.

Utilities and energy-intensive customers should make a parallel bet: AI-driven grid optimization and predictive maintenance will be the difference between a successful first-of-a-kind unit and an investable fleet. Not because AI replaces engineering, but because it makes engineering operational at scale—through better monitoring, better scheduling, and better proof.

If your organization is already investing in AI for demand forecasting, outage prediction, and grid optimization, you’re closer to SMR readiness than you might think. The question is whether you’re building those capabilities with nuclear-grade discipline—data quality, auditability, and governance.

What would change in your planning if you assumed that by the late 2020s, the constraint isn’t “Do SMRs exist?”—but “Can we integrate and operate them efficiently enough to justify a fleet?”