Nuclear-Backed AI Power: FRMI vs Pure-Play Nuclear Fuel and SMRs
3/28/26
SEQH CAPITAL RESEARCH
NUCLEAR‑BACKED AI POWER – TEAR SHEET PREVIEW
KEY TAKEAWAYS
AI data center power demand is on track to become the largest incremental power shock in U.S. history, with global DC consumption projected to rise from 415 TWh in 2024 to 945 TWh by 2030 (+175%).
Nuclear is rapidly becoming the preferred baseload option for hyperscalers, with $50B+ of announced commitments across Meta/Vistra/TerraPower/Oklo, Microsoft/Constellation (TMI‑1 restart), Amazon/Talen, and Google/Kairos.
SEQH’s “Nuclear‑for‑AI Capital Stack” frames investable exposure across five layers: Fuel (CCJ, LEU), Utilities/Fleet (CEG, VST, NRG), SMR Technology (OKLO, SMR, NNE), Behind‑the‑Meter (FRMI), and DC Infrastructure (DLR, EQIX), with fuel and utilities offering the best risk‑adjusted profiles.
FRMI’s 17 GW Project Matador (Amarillo, TX) is the most aggressive nuclear‑for‑AI bet, but an 81% drawdown since IPO underscores extreme execution, governance, and structure risk.
Cameco (CCJ) is the only positive performer in the coverage universe since FRMI’s IPO (+24.5%), confirming the market’s preference for proven cashflows and commodity leverage over speculative nuclear development stories.
AI POWER DEMAND & WHY NUCLEAR WINS
U.S. data centers consumed 183 TWh in 2024 (~4% of U.S. electricity) and are projected to triple to ~12% by 2028; grid interconnection queues of 3–7 years and a ~45 GW U.S. shortfall highlight a structural supply problem.
Nuclear’s 90%+ capacity factors, zero direct carbon, and high energy density make it uniquely suited to multi‑GW AI campuses; matching a single 1 GW nuclear unit would require 5,000–8,000 acres of solar or ~60,000+ acres of wind plus storage.
Meta, Microsoft, Amazon, and Google are already locking in nuclear PPAs and fleet deals (e.g., Meta’s 4 GW+ SMR program, Microsoft–CEG TMI‑1 restart, Amazon–Talen $20B+ campus, Google–Kairos Hermes/commercial), institutionalizing nuclear within AI power procurement.
THE NUCLEAR‑FOR‑AI CAPITAL STACK
Layer 1 – Fuel (CCJ, LEU): Lowest‑risk leverage; Cameco controls ~18% of global uranium supply, while Centrus is the only U.S. HALEU enricher, benefiting from any reactor build‑out regardless of technology.
Layer 2 – Utilities/Fleet (CEG, VST, NRG): Existing nuclear and thermal fleets with direct data‑center PPAs; CEG’s ~21 GW nuclear fleet plus Microsoft PPA and Vistra’s Meta partnership anchor this layer.
Layer 3 – SMR Technology (OKLO, SMR, NNE): Pre‑revenue, high‑beta technology options with 5–10 year timelines; materially levered to policy and AI power shock scenarios but heavily penalized in slowdowns.
Layer 4 – Behind‑the‑Meter (FRMI): A 17 GW private “HyperGrid” campus model that bypasses the grid; maximum AI exposure but extreme risk around capital intensity, REIT qualification, governance, and nuclear delivery.
Layer 5 – DC Infrastructure (DLR, EQIX): AI‑driven data center REITs capturing tenant demand growth without taking nuclear tech or fuel risk.
FRMI & RELATIVE POSITIONING
Project Matador plans: 2 GW gas secured by 2026, 6 GW permitted gas/mobile, 4 GW of AP1000 nuclear (2032–2036), 2 GW SMR, for 17 GW total; more than $700M in financing committed but tens of billions ultimately required.
REIT structure, zero revenue to date, no public DC tenant announcements, and CEO fraud allegations drive market skepticism; shares are down ~81% from IPO (~$32.53 to $6.18).
On EV/GW, CEG’s operating capacity is valued at ~$3.1B per GW (and ~$4.7B per nuclear GW) versus FRMI’s ~$0.22B per planned GW (~$0.9B per planned nuclear GW), reflecting the discount investors apply to unbuilt, uncontracted assets.
WHERE RISK‑ADJUSTED RETURNS LOOK BEST
Fuel (CCJ, LEU) and Utilities (CEG, VST) dominate the probability‑weighted scenario work: positive skew in 65–75% of cases, with only the AI slowdown scenario meaningfully negative.
FRMI and SMR developers (OKLO, SMR, NNE) show the widest dispersion – highest upside in AI power shock / pro‑nuclear regimes, but severe downside in any AI growth or policy disappointment; appropriate only for investors sizing them as high‑volatility options.
WHAT PAID MEMBERS GET IN THE FULL REPORT
Upgrade to unlock the full institutional version, including:
Detailed build‑out of the five‑layer Nuclear‑for‑AI Capital Stack, with company‑by‑company role definitions, risk profiles, and revenue visibility assessments.
Complete comparable‑company tables (valuation, EV/GW, capacity mix) plus the return, volatility, and correlation work underpinning the risk‑adjusted rankings.
Full FRMI Project Matador timeline, capacity build schedule, financing structure, and governance/risk discussion, including REIT qualification analysis.
Scenario matrix (5 macro/AI paths) with probability weights and layer‑by‑layer impact, plus narrative guidance on how to position across CCJ, CEG, SMRs, FRMI, and DC REITs under each case.
Positioning section with explicit “core,” “satellite,” and “high‑conviction speculative” roles for each name across different risk tolerances and time horizons.
For the full capital‑stack framework, the return/volatility/correlation work behind our rankings, and the granular FRMI vs. CCJ/CEG trade‑off analysis, consider upgrading to the paid tier. That unlocks the complete Nuclear‑Backed AI Power report and future updates as AI power demand, nuclear policy, and FRMI’s execution trajectory evolve.
