Tools chosen: kalshi_direct kalshi_related polymarket_related claude_news claude_news gdelt_news wikipedia code_execution

Sub-questions (Fermi decomposition)

Planner reasoning

This is a long-dated Kalshi policy/tech question; the market price is the primary anchor, but the base rate is heavily shaped by already-announced US programs (Army Janus Program, Project Pele, DoD Advanced Nuclear Power for Installations, the May 2025 executive orders requiring a reactor at a military installation by Sept 2028, and Eielson AFB microreactor). The key research is whether an existing program already pairs a microreactor with a data/AI compute center on a base, and how loosely 'starts the process' would resolve.

# Current state As of the market's latest data (~Aug/Sept 2026), the Army has selected 5 specific bases and vendors (Fort Bragg, Campbell, Hood, Benning, Drum) under the Janus Program and awarded up to $2.2B in contracts to build on-base microreactors (FY2027–FY2031 build/operate window), driven by EO 14299's Sept 30, 2028 statutory deadline. However, these are framed as general installation/grid-independence power, not explicitly "for a data center" — while separate, less-mature efforts (Fort Bliss AI data center, floating reactor concept) more directly pair reactors with AI compute on bases but remain unconfirmed/pre-contract. No NRC/DOE reactor authorization, EIS filing, or groundbreaking tied specifically to a data-center-linked reactor has yet occurred. # Timeline of key events - 2020 – Project Pele launched by DoD to build deployable microreactor for remote bases (confirmed, Wikipedia). - 2022-06 – BWXT awarded $300M contract for Pele's 1.5 MWe prototype reactor (confirmed, BWXT/claude_news). - 2025-05-23 – Trump signs EO 14299, directing Army-regulated reactor operational at a domestic base by Sept 30, 2028; mandates NEPA categorical exemptions; directs DOE to designate an AI data center for advanced reactor use within 90 days (confirmed, presidency.ucsb.edu). - 2025-10 – Army formally launches Janus Program as executive agent for EO 14299 compliance (confirmed, army.mil). - 2025-12-18 – FY2026 NDAA signed: Sec 318 creates DoD nuclear-deployment executive agent; Sec 319 establishes Advanced Nuclear Technologies Transition Working Group (≥3 pilot projects); Sec 321 creates 10-year Navy SMR pilot (confirmed, morganlewis.com). - 2026-02 – Pentagon explores 300 MW floating reactor concept with UK's Core Power, targeting 2028 deployment, explicitly linked to AI compute support (reported, armyrecognition.com). - 2026-04-19 – Air Force/DIU select Buckley AFB (CO) and Malmstrom AFB (MT) as potential reactor sites (confirmed, gazette.com). - 2026-07-21 – Oklo and X-Energy join Trump admin's $200M DOE-led program to speed reactors for AI data centers (DOE/national labs track, not DoD-direct) (confirmed, Bloomberg). - 2026-08-08 – Pentagon reported advancing AI data centers across military bases, with Fort Bliss as flagship; procurement requires off-grid power plan (reported, nuclear-news.net). - 2026-08-26 – Army announces Janus Program awards: 5 bases/vendors, up to $2.2B combined, build/operate FY2027–FY2031; Army expects 20+ microreactors DoD-wide eventually (confirmed, Fox Business/The Hill/Washington Times). # Event Will the US start the process of building a nuclear-powered data center on a military base before 2030? # Outcomes to forecast Yes / No # Kalshi market anchor **Current YES price: 54%** (up +8pts over 7 days, +5pts over 30 days; range 30–63% over 47 data points; avg daily volume only 41 contracts — thin market). Price has trended upward, consistent with news flow (Janus awards, EO deadlines, Fort Bliss reports). # Sub-question answers 1. **Kalshi price/volume history** — YES at 54%, rising steadily (+8% wk, +5% mo); low liquidity (~41 contracts/day) suggests moves are news-sensitive but not deeply informed. [kalshi_direct] 2. **DoD programs targeting reactors on bases** — Project Pele (BWXT, Idaho Nat'l Lab demo ~2028); Janus Program (5 bases: Bragg, Campbell, Hood, Benning, Drum; FY2027–31); Air Force/DIU site picks at Buckley and Malmstrom (Apr 2026); Navy SMR 10-yr pilot (NDAA Sec 321). [claude_news, gazette.com, morganlewis.com] 3. **Explicit data-center linkage** — EO 14299 explicitly calls for reactors for both military bases AND AI data centers, and directs DOE to designate an AI data center for advanced reactor use. Fort Bliss is reported as flagship AI data center site requiring off-grid power. Floating reactor concept explicitly cites AI compute support. But the 5 awarded Janus bases are framed as general installation power, not data-center-specific. [presidency.ucsb.edu, nuclear-news.net, armyrecognition.com] 4. **EO/NDAA deadlines** — EO 14299 (May 2025) sets Sept 30, 2028 deadline for Army-regulated reactor operational on a base; directs NEPA categorical exemptions. FY2026 NDAA (Dec 2025) codifies executive agent, working group, and Navy SMR pilot but sets no new hard "data center" deadline. [presidency.ucsb.edu, morganlewis.com] 5. **Concrete authorizations/filings** — As of Aug 2026: contracts awarded ($2.2B Janus, $300M Pele/BWXT), sites selected, but no NRC/DOE formal reactor authorization, EIS filing, or groundbreaking confirmed for any data-center-linked reactor specifically. [claude_news] 6. **Failure modes** — Analysts (Hogan Lovells) flag funding/appropriations risk (Congress unlikely to fully fund until 2026+), regulatory ambiguity (DOE alternative authorization vs. NRC), and general schedule slippage on "first-of-a-kind" reactor by 2028 as ambitious. HALEU fuel supply risk not directly addressed but is a known industry-wide bottleneck. [hoganlovells.com] # Key facts (high-confidence, factual) 1. [presidency.ucsb.edu] EO 14299 mandates Army-regulated reactor operational at a domestic base by Sept 30, 2028. 2. [Fox Business/The Hill] Army awarded ~$2.2B across 5 bases (Bragg, Campbell, Hood, Benning, Drum) for microreactors, build/operate FY2027–2031. 3. [nuclear-news.net] Fort Bliss reported as flagship site for Pentagon AI data center requiring non-grid power. 4. [morganlewis.com] FY2026 NDAA codifies nuclear deployment executive agent and Navy SMR pilot program. 5. [claude_news] No NRC/DOE authorization or groundbreaking confirmed yet for any reactor explicitly tied to a data center. # Cross-market signals - Kalshi related: KXFUSION-30 (nuclear fusion by 2030) trading at 31%, roughly flat — shows market distinguishes fusion hype from fission/military reactor progress. - Polymarket: No matching markets found. - Sportsbook implied: N/A (not applicable to this event type). # Analyst opinions and speculation - Hogan Lovells: 2028 deployment goal "ambitious," contingent on congressional appropriations unlikely before 2026; legal pathway (DOE alternative authorization vs NRC) unresolved. - code_execution scenario model: structural probability ~29.5% (base case), driven mainly by uncertainty in whether any reactor effort gets explicitly "data-center" framed rather than whether a reactor program proceeds at all (which is judged 55–85% likely). # Directional lean per outcome - **Yes**: Strong structural momentum — EO 14299 deadline (2028, before market close), 5 bases under contract, NDAA statutory backing, explicit EO language tying reactors to AI data centers, Fort Bliss/floating-reactor reports directly merging nuclear+AI-compute themes. Kalshi price rising on this news. - **No**: Current awarded projects (Janus 5-base program) are framed as general/installation power, not formally "data center" reactors; no NRC/DOE authorization, EIS, or groundbreaking yet exists; funding/regulatory/schedule risk is real; resolution may hinge on strict interpretation requiring explicit reactor-to-data-center pairing, which remains only in early/rumored stage (Fort Bliss, floating reactor). # Gaps / unknowns - No official Kalshi resolution criteria/rules text provided — ambiguous whether general installation-power reactors count as "data center" reactors. - No confirmed NRC/DOE filing or NEPA/EIS action for any base reactor as of latest data. - Fort Bliss AI data center reactor pairing and floating reactor concept remain reported, not contracted. # Calibration anchors - Kalshi current YES price: **54%** (anchor), up from ~46% a month ago on Janus-award news. - code_execution structural estimate: ~28–30% base case, implying market may be pricing in more certainty on "data-center" framing than warranted. - Comparable: KXFUSION-30 trading 31% flat — illustrates market discounts speculative tech timelines absent binding deadlines; KXDATACENTER-30 benefits from statutory EO deadline (2028), justifying higher price than fusion.

1. Both forecasters cluster near 50% by "splitting the difference" between the 54% Kalshi anchor and the 29.5% structural model, but neither explains why a naive average is the right method rather than weighting one estimate more heavily given its evidentiary basis—this risks anchoring bias, especially since Kalshi's volume is only ~41 contracts/day and explicitly flagged as "not deeply informed." 2. Neither forecast grapples deeply with the resolution-criteria ambiguity flagged in the brief ("No official Kalshi resolution criteria/rules text provided—ambiguous whether general installation-power reactors count as 'data center' reactors"); this is the single largest driver of variance between the 29.5% structural estimate and the 54% market price, yet both treat it as a minor caveat rather than the central crux. 3. Forecast 1 leans "slightly above" Kalshi citing "multiple independent paths to YES" (Janus, Fort Bliss, floating reactor, DOE/Oklo/X-energy track), but doesn't acknowledge that most of these paths are correlated (all stem from EO 14299) rather than truly independent, which would justify less probability boost than implied. 4. Neither forecast addresses the base-rate/reference-class question of "first-of-a-kind" DoD reactor programs historically slipping schedule (Project Pele itself, launched 2020, still only at prototype stage in 2026)—six years for a 1.5 MWe demo reactor suggests the 2028 deadline is optimistic and deserves more weight in the "No" case than either forecast gives it. 5. Both forecasts undervalue the "loose language" interpretation angle: if "start the process" only requires a permit application, site selection, or contract award (already true for Janus), the question could already be resolved YES today regardless of "data center" framing debates—neither forecast explores this alternate low bar that could push probability meaningfully higher than 50-58%. 6. Neither critiques the KXFUSION-30 comparison brief provides (31% flat, no binding deadline) as a useful calibration check—if fusion at 31% reflects speculative-tech discount and this market has a binding statutory deadline (EO 14299), the market's 54% may still be underpricing the "Yes" given the deadline is structurally different from fusion's open-ended timeline, an asymmetry neither forecaster explicitly weighs into their final number.

60% (44% confidence)

Yes: 60% No: 40%

The resolution bar — 'start the process of building' — is deliberately loose, and multiple concrete steps (site selection, $2.2B Janus awards with FY2027–31 build/operate windows, Air Force site picks, Fort Bliss flagship AI data center explicitly requiring off-grid power) are already in motion with a 3+ year runway before 2030. The 29.5% structural model appears to over-penalize the 'data-center framing' crux by implicitly requiring a formally labeled reactor-plus-data-center project, whereas EO 14299 itself directs DOE to designate an AI data center for advanced reactor use, and the DOE/Oklo/X-energy track provides an additional (though correlated) route. Against this, the critique's point about first-of-a-kind schedule slippage is real — Project Pele took six years to reach prototype — but slippage mainly threatens completion, not 'starting the process,' which is what this question asks. I therefore weight the Kalshi anchor (54%, rising on genuine news flow) more heavily than the structural model and lean modestly above it, while keeping meaningful mass on No for the scenario where resolution demands an explicit, contracted reactor-to-data-center pairing that never formally materializes.

Total pipeline time: 207.9s