Quantifying Amazon-Backed X-energy: HALEU, Steam, and Silicon
3/12/26
SEQH Capital Research
Quantifying Amazon-Backed X-energy: HALEU, Steam, and Silicon
Tear Sheet – March 2026
Why This Report Exists
X-energy is not “one more SMR,” but the nucleus of an Amazon-anchored Xe-100 fleet that tightly links HALEU fuel demand, high-temperature industrial steam, and AI data-center power. This report rebuilds the fleet pipeline, models HALEU and SWU needs under three deployment paths, quantifies “compute-per-reactor” capacity, and translates these physics into portfolio-actionable signals for uranium, enrichment, and advanced reactor names.
X-energy in One Glance
Most commercially advanced HTGR developer in the West, with an order book >11 GW (~144 Xe-100 modules) across the U.S. and UK.
Anchors: Amazon’s up-to-5 GW U.S. deployment option through 2039, a 50B dollar strategic alliance with KHNP and Doosan, and >2B dollars of private capital raised (including a 700M dollar Series D led by Jane Street).
Sits at the intersection of three mega-trends: HALEU fuel-cycle dominance, industrial decarbonization via high-temperature steam, and nuclear-powered AI campuses.
Key commercial pillars:
Dow Seadrift (TX) 4-pack (320 MWe / 800 MWt), ARDP anchor, targeting ~440k t CO₂e/year avoided at a 4,700-acre chemical complex.
Cascade Advanced Energy Facility (WA) for Amazon (Phase 1: 4 modules, 320 MWe), expandable to 960 MWe; first concrete instantiation of the Amazon partnership.
Centrica Hartlepool (UK), JDA for up to 6 GW; first site 12 modules (~960 MWe, >15B dollar lifetime value).
Fleet Scale: Base, Bull, Bear
X-energy fleet scenarios (U.S. + UK):
Bull: 11,000 MWe by 2040 (138 modules) if full 5 GW U.S. + 6 GW UK are realized and Korean supply chain ramps quickly.
Base: 6,920 MWe by 2040 (86 modules) with ~70% of Amazon’s 5 GW converting plus initial UK Hartlepool; supply chain limits to ~8 modules/year by 2037.
Bear: 3,520 MWe (44 modules) if deployment is largely Dow + Cascade and UK slips beyond 2037.
Core finding: Base Case (~6.9 GW by 2040) drives ~395 MT cumulative HALEU demand, a dominant slice of projected Western supply and a direct volume tailwind to Centrus (LEU), Urenco, and ASP Isotopes (ASPI).
HALEU & Fuel-Cycle Demand
Xe-100 fuel design: TRISO-X pebble fuel, ~220k pebbles/core, ~18k TRISO particles/pebble, UCO kernels at 15.5% enrichment, 160 GWd/tHM burnup, 750°C outlet temp, with online refueling enabling ~95% availability.
Per-module calibration (normalized to 19.75% HALEU):
First-core load: ~1.3 MT per module.
Reload: ~0.58 MT per module per year.
Scenario HALEU demand:
2035 annual: Bear 10.4 MT/yr, Base 24.2 MT/yr, Bull 39.8 MT/yr.
2040 annual: Bear 28.4 MT/yr, Base 53.0 MT/yr, Bull 86.6 MT/yr.
Cumulative 2026–2040: Bear 199 MT, Base 395 MT, Bull 624 MT, equal to ~3.7%, 7.4%, and 11.7% of INL’s 2050 base HALEU scenario (~5,350 MT).
Western HALEU supply remains tiny: Centrus had delivered only 920 kg to DOE by mid‑2025; even with Centrus, Urenco Capenhurst, and early LIS (ASPI), Western HALEU is unlikely to exceed 10–20 MT/yr before early 2030s.
NIA pegs HALEU production cost near 23,725 dollars/kgU at 19.75% enrichment, reflecting ~1,000 dollars/SWU HALEU vs ~150 dollars/SWU standard LEU.
AI Data Centers: Power, Racks, Compute-per-Reactor
Each Xe-100 module: 80 MWe electric. At AWS global PUE of 1.15, yields ~69.6 MW of IT load per module.
Rack capacity at PUE 1.15:
40 kW: 1,739 racks
60 kW: 1,159 racks (current common AI density)
80 kW: 870 racks
Using Trainium3 (~2.52 PFLOPs FP8 per chip, ~500 W) and 60 kW/rack (~100–120 accelerators):
A single Xe-100 powers ~1,159 AI racks, ~130k–139k accelerators.
Implied compute: ~290–350 ExaFLOPs FP8 per module.
A 1 GW AI campus would need roughly 13–14 Xe‑100 modules at PUE 1.15; Amazon’s entire 5 GW U.S. strip could backstop about five such mega‑campuses.
Industrial Heat, Steam, and Hydrogen
Steam conditions: 565°C, 16.5 MPa, well above LWR steam range and tuned for petrochemicals, hydrogen, pulp/paper, and chemical synthesis.
Dow Seadrift reference case: 4 modules (800 MWt / 320 MWe) avoid ~440k t CO₂e/year. Back‑solved per-module metrics:
~110k t CO₂e/year avoided per module.
~2.1M MMBtu/year natural gas displaced per module.
Hydrogen (HTSE) potential: scaling INL’s 600 MWt HTGR‑HTSE (1.75 kg/s H₂, ~40.4% efficiency) to 200 MWt:
~0.58 kg/s ≈ ~50 t/day H₂ per module.
At 3.04–3.72 dollars/kg, ~55–68M dollars H₂ revenue per module per year.
Economics, Valuation-Style Metrics, and Optionality
Implied HALEU throughput value (using 23,725 dollars/kgU):
Bear 199 MT: ~4.7B dollars.
Base 395 MT: ~9.4B dollars.
Bull 624 MT: ~14.8B dollars.
Per 80 MWe module project NPV (40‑year life, 95% CF, 5,000 dollars/kW NOAK, 25 dollars/MWh OPEX):
At 90 dollars/MWh PPA: ~89M dollars NPV @8% WACC (~27M @10%).
At 100 dollars/MWh: ~157M @8%, ~88M @10%, IRR ~12.4%.
At 110 dollars/MWh: ~225M @8%, ~149M @10%.
Amazon 5 GW strip: treat each 320 MWe four‑pack as a project call option. At 100 dollars/MWh and 10% WACC, each four‑pack adds ~352M dollars NPV; full 5 GW (15.6 four‑packs) implies ~5.5B dollars of platform option value.
HALEU Efficiency vs. Other Advanced Designs
INL HALEU scenarios (equivalent energy output to 2050):
Xe-100 all‑HTGR fleet: lowest cumulative HALEU (~3,450 MT).
Natrium SFR: ~7,175 MT HALEU (2x Xe‑100).
USNC MMR: ~9,000 MT estimated (very high per GWe due to 20‑year cores).
Xe‑100’s high burnup, online refueling, and 15.5% enrichment make it the most HALEU‑efficient advanced reactor per unit energy, a structural advantage when HALEU supply is the binding constraint.
Fuel Fabrication, Supply Chain, and Key Trades
TRISO‑X (X-energy subsidiary) holds first‑mover monopoly on Western HALEU pebble fabrication:
TX‑1 (Oak Ridge) 5 MTU/yr (~11 reactors), NRC Category II license (Feb 2026), first HALEU feedstock (Dec 2025).
TF3 expansion: 8 MTU/yr initially, scalable to 16 MTU/yr by early 2030s.
Full 21 MTU/yr supports ~46 Xe‑100 reactors; further capacity needed for full U.S.+UK fleet.
Key beneficiaries and themes:
Centrus (LEU): only operating U.S. HALEU enricher; each X-energy FID = multi‑year fuel contract.
ASPI/QLE: LIS-based HALEU route; large addressable market at current HALEU pricing.
CCJ, UUUU, broader miners: more HALEU demand = higher natural uranium throughput and price support.
BWXT, CEG, DOW: component, operations, and industrial-heat customers.
Key catalysts: Dow Seadrift NRC CP, Cascade Phase 1 FID + PPA, TX‑1 first fuel, Hartlepool approvals, X‑energy IPO filing, Centrus/Urenco HALEU ramp milestones, additional Amazon/X-energy FIDs.
Want the Full X-energy / Xe-100 Deep Dive?
[READ THE COMPLETE X-ENERGY REPORT]
The full report includes a quantitative, Amazon-anchored model you won’t find anywhere else:
Complete reconstruction of X-energy’s 11+ GW pipeline across Dow, Amazon Cascade, Hartlepool, and residual U.S. options
Three-path fleet deployment model (Bear/Base/Bull) with module counts, MW, and HALEU tonnage through 2040
Detailed HALEU and SWU demand modeling for Xe-100 vs. Natrium vs. USNC MMR, with INL-based efficiency rankings
AI campus power stack: PUE, rack densities, accelerators per module, and ExaFLOPs-per-reactor metrics using AWS Trainium3
Industrial heat and hydrogen economics from the Seadrift reference case and HTSE-coupled Xe-100 configurations
Project NPV and IRR grids by PPA price, WACC, and capex, plus an explicit valuation of Amazon’s 5 GW option strip
Fuel-fabrication capacity roadmap (TX‑1, TF3) and HALEU feedstock bottleneck analysis
Risk matrix covering FOAK cost overrun, HALEU supply, licensing, tech execution, and competitive displacement
Knock-on trade map across LEU, ASPI, CCJ, BWXT, UUUU, CEG, and DOW
X-energy’s Xe‑100 fleet is a single platform that ties HALEU, industrial steam, and AI compute into one quantifiable demand engine. This report shows you exactly how that engine translates into tons, MWh, and dollars.
Available exclusively to SEQH Capital Research paid subscribers. Upgrade to access →
