Six rings, fully mapped.

Factory-fabricated modules, passive safety cases, and licensing pathways where FOAK economics are priced honestly.

SMRs are not one product category. Licensing cadence, forgings, and construction method determine bankability more than nameplate capacity. We diligence regulatory milestones, supply-chain depth, and whether FOAK cost is modelled as central case — not as a footnote.

Declined — Reactor vendors whose only evidence is a rendering; programmes without a named competent authority review path.

Molten salt, high-temperature gas, and sodium-cooled designs with published safety analysis and materials programmes.

Gen-IV is interesting where chemistry and materials science are the moat — not where the moat is a slide comparing thermal efficiency to PWR. We read corrosion chemistry, tritium handling, and fuel qualification as engineering ledgers.

Declined — Concept reactors with no hot-cell or loop-test history; any design whose civilian case depends on weapons infrastructure.

Teams whose neutronics, tritium inventory, and first-wall lifetime are measured — not narrated.

Fusion belongs in mandate only where the physics team has published diagnostics and the corporate structure is explicitly civil-only. Power balance, breeding blanket performance, and materials replacement intervals are the first three diligence files we open.

Declined — Fusion companies differentiated by fundraising narrative; dual-use adjacency without governance separation.

Robotics, digital twins, and waste-characterisation tools that shorten outage and decommission schedules.

The installed base of civil reactors is large and ageing. Inspection robotics, radiation mapping, segmentation tooling, and digital-twin maintenance reduce dose and downtime. These businesses sell into every jurisdiction regardless of which new-build design wins.

Declined — Consultancies without proprietary sensors or software; services whose revenue requires defence clearance.

Pellet production, cladding, and dry-storage cask engineering for civil programmes.

Fuel-cycle technology is investable where the customer base is civil utilities and research reactors, with export controls and governance documented. We read safeguards design and material accountancy as seriously as yield.

Declined — Enrichment or handling technologies without published civil-only use case; any dual-use ambiguity left unresolved.

Spectroscopy, cementation, and volume-reduction systems that change repository economics.

Waste is a chemistry and logistics problem. Characterisation that reduces category uncertainty, and treatment that reduces volume, directly changes long-term liability. We prefer hardware with field deployments over simulation-only platforms.

Declined — Deep-geologic repository advocacy without operating technology; promises of transmutation without physics programme.

Area monitors, personal dosimetry networks, and source localisation for plants and industrial sites.

Monitoring is the feedback loop for every other safety system. Semiconductor and scintillator advances are opening lower-power, higher-resolution networks. We diligence calibration traceability and regulatory acceptance in target jurisdictions.

Declined — Dashboards that visualise third-party sensors without proprietary detection physics.

Passive heat removal, filtered venting, and hydrogen management for beyond-design events.

Post-Fukushima retrofits created a durable market for severe-accident mitigation hardware. New-build designs embed many functions passively; the retrofit layer remains large for decades. We invest in components with type-tested evidence, not policy slides.

Declined — Autonomous lethal systems; any product line marketed for area denial or munitions.

Turbines, heat exchangers, steam cycles, and control systems common to every firm generator.

Balance of plant is where a generating asset actually wins or loses on availability. The same steam cycle, feedwater chemistry, and control hardware serve nuclear, geothermal, and thermal plant, so a qualified supplier sells across the fleet regardless of which primary technology is chosen.

Declined — Retrofit consultancies without proprietary hardware; efficiency claims with no metered baseline.

Switchgear, transformers, HVDC components, and cable systems where lead times gate deployment.

Grid hardware is now the binding constraint on connecting new generation. Transformer and switchgear lead times run to years. We invest in manufacturing capacity and component technology that shortens interconnection, not in software that visualises the queue.

Declined — Interconnection-queue dashboards; developers without equipment or manufacturing ownership.

Instrumentation and protection systems that hold frequency and voltage as inertia falls.

As synchronous inertia leaves the system, protection and fast-response control become safety-critical engineering rather than optimisation. We read type-test evidence and grid-code compliance in target jurisdictions the same way we read a licensing basis.

Declined — Trading or forecasting platforms with no protection or control hardware of their own.

Waste-heat recovery, process electrification, and recuperation with measured payback.

Industrial heat is the largest under-addressed energy demand. We back recovery and electrification hardware where the saving is metered at a real site with a named offtaker — efficiency is an engineering result, not a marketing category.

Declined — Offset-linked efficiency models; savings evidenced only by simulation.