FIELD CONSOLE / AI x Energy x Nuclear
● 17 SEPT 2026
STRATEGIC INTELLIGENCE · DUAL-USE ENERGY · AI × ENERGY SOVEREIGNTY

MOBILE NUCLEAR MICROREACTORS

A field note focused on factory-built, transportable microreactors in roughly the 1–20 MWe class; adjacent larger SMRs appear only for context. Many leading concepts target multi-year cores. One substrate, two strategic imperatives: reduce the fuel-logistics tail of remote / semi-fixed defence installations and unlock AI compute where grid capacity arrives too slowly.

Focus range
1–20 MWe
larger adjacent SMRs shown for context
Core life target
5–8 yrs
common design target; vendor-specific
Commercial window
2028–30
first deployments / electrons; site install ≠ total lead time
Fuel bottleneck
HALEU*
common constraint; fuel form varies by design
STATUS CHECK · 17 SEPT 2026. Westinghouse eVinci completed hot zero-power criticality testing on 3 Sept (critical at 663°C; peak core temperature 1,011°C). The U.S. Army's Janus program announced five vendors paired with five initial installations on 26 Aug, with more than 20 total microreactors expected across defence installations. These are physics, procurement and deployment milestones — not evidence of a mature commercial fleet. DOE / eVinci ↗ · U.S. Army / Janus ↗
01 / PHYSICAL FOUNDATIONS

Reactor Physics & Taxonomy

Three families reduce conventional nuclear-plant complexity in different ways: high-temperature gas cooling, passive heat-pipe transport, or compact integral water systems. Safety case, coolant, fuel and transport assumptions are design-specific. Pick an architecture to inspect the illustrative envelope.

Operating envelope — HTGR
Core temperature
Fuel life (years)
Passive safety index
Power density
Air-cooling feasibility
🛡 TRISO — multi-layer particle fuel with high-temperature retention 🔥 Heat pipes — passive capillary transport ⚖ Temp feedback — rising fuel temperature reduces reactivity 📦 Transportable — some designs target road / rail / air / barge logistics
02 / COMPETITIVE MAP

Global Industrial Landscape

US venture swarm anchored by DoD procurement vs. Rosatom's operating fleet vs. China's commercial SMRs. Filter by bloc.

03 / DEFENCE.IS RATIONALE

Dual-Use Resilience & the Porcupine Strategy

Hardening society's nervous system — power, comms, water, compute — against paralysis. The near-term defence lane is semi-fixed installations, remote garrisons and critical-base resilience, not a battlefield reactor. A microreactor can reduce dependence on repeated fuel deliveries and on a single external grid connection.

🚚 Shrink the fuel tail

A multi-year core can replace repeated diesel deliveries for persistent loads. Exact avoided fuel volume and convoy count depend on site load, generator efficiency and logistics.

⚡ Megawatt mission systems

High-energy lasers, AESA radar, counter-drone systems and edge AI can create demanding power-quality and transient-load requirements — often pushing designs toward storage and hybrid microgrids.

🛡 Sovereign grid cells

Cyber, sabotage and EMP targets centralized grids. Islanded microreactors keep comms, water and financial nodes alive — difficult to divide, harder to disable.

Mission energy demand · 5 MWe · 5 yrs
0
MWh at continuous 5 MWe · fuel-equivalent and tanker trips are site-specific
Microreactor alternative
multi-year core
many concepts target 5–8 years between refuelling; transport and servicing model varies by vendor
Read the civic-resilience handbook → defence.is↗
04 / DRAM.GOLD RATIONALE

AI Compute & Token Generation

The bottleneck evolved: GPUs → HBM → primary power. DRAM is the reserve asset, tokens are the unit of account, and continuous energy is the extraction mechanism. For some remote, sovereign or queue-constrained deployments, an islanded reactor + BESS can be valued against time-to-power, not only against grid LCOE.

Phase I · 2022–24 · accelerators Phase II · 2024–25 · memory bandwidth Phase III · 2025– → energy & interconnection
Inference spike vs nuclear baseload · 24 h
spike intensityL3
GPU inference load reactor baseload 100% BESS ± buffer
Illustrative orchestration sketch only — not a reactor-dynamics model. Y-axis: MW; X-axis: hours.
Power source ≠ control layer → Compute Flex / Power to Intelligence↗ The full token-economics thesis → dram.gold↗
05 / ECONOMICS & HALEU

FOAK → NOAK Learning Curve

First units are expensive and published estimates vary widely. The comparison is not apples-to-apples: microreactor bands are model-dependent, remote diesel is often fully burdened, and central-grid prices usually exclude years of interconnection delay. For AI infrastructure, the decision variable can be MW delivered by date, not only $/MWh.

FOAK estimates
$150–300+/MWh
NOAK targets
$60–140/MWh
Burdened diesel
$250–600/MWh
Central grid
$40–80/MWh
READ THE BANDS CAREFULLY. These are directional ranges, not a single harmonised dataset. FOAK/NOAK values vary by design and financing assumption; remote diesel is often quoted on a fully burdened basis; central-grid prices usually exclude queue delay and site-enablement cost.
⚠ HALEU supply — a binding constraint for many designs. DOE has projected that more than 40 metric tons of HALEU may be needed by 2030 across advanced-reactor deployment. Commercial Western supply remains constrained, so fuel availability can become a schedule and bankability issue alongside reactor performance. Not every design uses the same fuel form or enrichment.
NRC · design / site licensing pathways 10 CFR 70 · special nuclear material / fuel handling 10 CFR 71 · transport casks COCO / PPA structures IRA production tax credit
Where this fits the defence-tech →↗