Advanced Nuclear Funding Is Back—AI Makes It Deployable

A clear signal cut through December’s end-of-year noise: private capital is pouring back into advanced nuclear—and it’s flowing toward projects that can prove they work soon. By early Q3 2025, nuclear fission companies had raised $1.3B in equity funding, the highest annual total on record, and SMRs and microreactors captured roughly 75% of that funding.

Here’s what I think is driving the spike: investors aren’t suddenly feeling sentimental about big infrastructure. They’re responding to a more practical truth—the grid’s reliability problem is now a compute-and-data problem as much as it’s a steel-and-concrete problem. If you’re running an energy portfolio, building data centers, or modernizing utility operations, you don’t just need more generation. You need generation that can be planned, monitored, optimized, and integrated with a grid that’s getting more complex every quarter.

That’s where this fits in our “AI in Energy & Utilities” series. The funding headlines are about reactors, but the underlying story is about data intelligence: better modeling, better safety monitoring, better commissioning, better dispatch, and faster decisions from licensing through operations.

Why 2025’s nuclear investment surge looks different

Answer first: 2025 capital is concentrating around developers with near-term demonstrations, defined regulatory paths, and manufacturing plans—because investors now price “execution certainty” above big promises.

Market data points tell the story. In addition to the $1.3B raised by early Q3, deal volume accelerated: 28 equity transactions by October 2025, compared with an average of about 15 in prior years. That’s not just “more money.” It’s a broader investor base taking more shots.

But the important nuance: much of the money is going to companies moving from design to deployment. That shift matters because advanced nuclear is capital intensive, schedule sensitive, and deeply exposed to execution risk.

The new investor question: “Can you prove it—then build it?”

In advanced nuclear, “prove it” means:

• A credible demonstration plan (site, timeline, fuel pathway)
• Regulatory engagement that reduces licensing ambiguity
• A manufacturing approach that doesn’t treat every plant as a custom megaproject
• Early commercial validation (deposits, offtake interest, anchor customers)

And “build it” means a supply chain, quality system, instrumentation, commissioning discipline, and a plan to operate reliably.

AI doesn’t replace nuclear engineering. It does something equally valuable: it shrinks uncertainty across those steps by turning complex systems into measurable, monitorable, optimizable operations.

Three December fundraises—and what they really signal

Answer first: Radiant, Last Energy, and ARC Clean Technology raised capital because they’re aligning technology choices with manufacturability and regulator-ready execution—and AI will be central to making those deployments operationally bankable.

Radiant: microreactors for “always-on” power—plus a factory plan

Radiant announced $300M+ in new funding in mid-December 2025, following a $165M Series C in May 2025. The company is developing Kaleidos, a 1 MWe / 1.9 MWth helium-cooled microreactor using TRISO fuel with passive cooling. The commercialization storyline is unusually concrete:

• Fueled test planned for 2026 at Idaho National Laboratory’s DOME facility
• Customer commitments including an agreement (with deposits) for 20 units for data center use
• A new manufacturing facility—the R-50 factory in Oak Ridge, Tennessee—with construction expected to begin in early 2026

Where AI fits: microreactors live or die on repeatability. AI helps create that repeatability in three practical ways:

1. Manufacturing quality analytics: detecting process drift (welds, machining tolerances, sensor calibration) before it becomes a field issue.
2. Condition-based maintenance: learning normal signatures for pumps, valves, and heat exchangers so maintenance is planned rather than reactive.
3. Operational anomaly detection: monitoring multivariate signals to flag issues early—important for remote, defense, or disaster-response use cases.

A sentence investors understand: “If we can monitor it reliably, we can operate it predictably.” AI makes that believable.

Last Energy: standardized PWR thinking—built for schedule

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

The commercial design is the PWR-20, a 20 MWe pressurized water reactor designed for modularization and factory fabrication, using standard PWR fuel (<4.95% enriched, 17×17 assemblies) and drawing from the operating experience of 300+ PWRs worldwide.

Last Energy’s thesis is basically: use what the industry already knows works, then industrialize it. I’m generally bullish on that approach because it’s aligned with the biggest pain point in energy infrastructure: schedule certainty.

Where AI fits: standardization creates a huge advantage for machine learning because it produces comparable fleets and comparable data.

• Digital commissioning: AI-assisted test verification can reduce commissioning cycles by catching misconfigurations earlier.
• Predictive maintenance for modular plants: a fleet of near-identical units is ideal for fleet-wide reliability models.
• Grid integration and dispatch: smaller units need smarter dispatch decisions—especially in grids with high renewable penetration.

If you’re a utility or C&I energy buyer, what you want is not just a reactor. You want a repeatable operating model.

ARC Clean Technology: a fast reactor built on proven roots

ARC closed a Series B to accelerate commercialization of its ARC-100: a 100 MWe / 286 MWth sodium-cooled fast reactor, derived from EBR-II operating experience. ARC also completed Phase 2 of the Canadian Nuclear Safety Commission Vendor Design Review in 2025—a meaningful de-risking milestone because it signals no fundamental barriers identified at that stage.

ARC-100’s design targets include a 20-year core life without refueling and a 60-year plant operating life.

Where AI fits: fast reactors and industrial heat applications raise the bar for operational insight.

• Advanced sensor fusion can improve visibility into thermal-hydraulic behavior and equipment health.
• AI-driven performance optimization can support heat-and-power co-optimization (electricity plus industrial heat, hydrogen, or data centers).
• Operational decision support becomes valuable when plants serve multiple outputs and contractual commitments.

For ARC and similar platforms, AI is a practical tool for operational assurance—the kind that reduces perceived risk for host sites and offtakers.

AI’s real job in advanced nuclear: reduce uncertainty across the lifecycle

Answer first: AI creates value in advanced nuclear by lowering schedule risk, improving reliability, and making grid integration more predictable—three things that directly affect financing and customer adoption.

A lot of “AI in nuclear” talk gets abstract fast. Here’s the grounded version—what utilities, IPPs, and large energy buyers actually care about.

1) Design-to-license: faster iteration, clearer evidence

Advanced nuclear programs generate mountains of analysis: safety cases, thermal models, component qualification packages, operational procedures. AI can help teams:

• Summarize and trace requirements across documents (reducing gaps)
• Detect inconsistencies in engineering documentation
• Prioritize verification activities based on risk signals

This isn’t about letting an algorithm “approve” anything. It’s about reducing human rework and tightening the chain of evidence.

2) Construction and manufacturing: quality at production speed

Factory-built nuclear only works if the factory behaves like a modern high-quality manufacturer, not a one-off project shop.

AI supports:

• Vision inspection for surface defects and assembly verification
• Process monitoring to detect drift before parts fail acceptance
• Supplier risk scoring based on delivery performance and quality escapes

If you’re planning multiple units per year, quality intelligence becomes as important as steel.

3) Operations: reliability and safety monitoring that scales

Microreactors and SMRs are often pitched for remote sites, campuses, and industrial facilities. That increases the importance of:

• Remote monitoring
• Early anomaly detection
• Fleet-wide learning (what one unit learns can help others)

AI-enabled condition monitoring can reduce forced outage rates—one of the most expensive and reputation-damaging outcomes for early deployments.

4) Grid integration: dispatch that matches a renewables-heavy reality

This is the bridge many teams miss: advanced nuclear won’t be deployed into a static grid. It will be deployed into a grid with:

• higher renewable variability
• more congestion and constraints
• more distributed resources
• more volatile price signals

AI in energy and utilities already supports load forecasting, congestion forecasting, and optimal dispatch. Adding SMRs to that environment means you need stronger tools for:

• unit commitment under uncertainty
• probabilistic forecasting (weather → renewables → net load)
• reliability planning (N-1 contingencies, restoration scenarios)

Put simply: capital gets reactors built, but data gets them used well.

What utilities and energy buyers should do in 2026 (even if you’re not buying a reactor)

Answer first: treat advanced nuclear like a data-and-operations program, not just a generation procurement—and start building the AI-ready foundation now.

Whether you’re a utility, muni, co-op, industrial, or data center operator, you can prepare without making a giant bet.

A practical readiness checklist

1. Define the operating model you’d require

   • Who monitors the plant, how often, and with what tools?
   • What SLAs exist for availability and response time?

2. Get your grid analytics in order

   • Can you run scenario-based studies for adding 20 MW, 100 MW, or multiple small units?
   • Do you have forecasting workflows that incorporate uncertainty, not just point forecasts?

3. Plan your data pipeline early

   • What telemetry will you receive?
   • How will you store it, secure it, and make it usable for reliability analytics?

4. Decide how you’ll validate AI outputs

   • What’s advisory vs. automated?
   • What human sign-offs are required?

5. Make cybersecurity non-negotiable

   • Remote monitoring and fleet learning increase the attack surface
   • Build a clear separation between monitoring analytics and control systems

If you’re already investing in AI for grid optimization or predictive maintenance, you’re not starting from zero. You’re extending what you already do into a new generation asset class.

The lead indicator to watch: demonstrations that produce operating data

Answer first: the next 12–24 months will separate “funded concepts” from “deployable products,” and operating data from demonstrations will be the credibility currency.

Radiant’s 2026 demonstration plan, Last Energy’s 2026 criticality target, and ARC’s regulatory progress are all pointing toward the same inflection point: proof that these systems can run predictably and be supported like products.

From an AI in Energy & Utilities perspective, that’s the moment when teams should pay attention—not just to the reactor physics, but to the operational data streams:

• What signals are monitored and at what cadence?
• How do operators respond to anomalies?
• What maintenance patterns emerge?
• How does the unit perform under real grid conditions?

Advanced nuclear is re-entering the spotlight because reliability has become a board-level concern again—especially with data center growth, electrification, and tighter reserve margins in many regions. Capital is necessary. Operational intelligence is decisive.

If you’re building your 2026 roadmap for AI in energy and utilities, here’s the forward-looking question to sit with: When the first wave of SMR and microreactor demos start producing real operating data, will your organization be ready to turn that data into dispatch, reliability, and planning decisions?