Tools chosen: kalshi_direct kalshi_related polymarket_related claude_news gdelt_news wikipedia congress_bills code_execution

Sub-questions (Fermi decomposition)

Planner reasoning

This is a long-dated (2030) Kalshi question about a fairly specific but arguably low-bar event ('starts the process'), so the market price plus resolution-criteria ambiguity dominate. Key research is on the DoD Army Janus program / ANPI microreactor awards, the June 2025 executive orders requiring a reactor at a military installation by Sept 2028, and any announced data-center-plus-reactor pairings on military installations.

# Current state Kalshi YES trades at 53% (Aug 2026), reflecting genuine uncertainty about whether DoD's active microreactor programs will produce an explicit "nuclear-powered data center on a military base" before 2030. No project has yet been publicly announced merging a reactor and a data center into one explicit contract/groundbreaking, but component pieces (RFP for AI data centers at Eielson AFB, which already has a preferred-site reactor designation) are converging. # Timeline of key events - 2020: DoD launches Project Pele to build deployable microreactors for military bases (confirmed, Wikipedia). - 2024 (June, per prior reporting): Eielson AFB named preferred site for Oklo-built microreactor; DLA selects Oklo to design/build/own/operate it (reported, govtech.com). - 2025-06: Executive Order 14299 signed, mandating an Army-regulated nuclear reactor operational at a domestic military installation by 2028-09-30 (confirmed, army.mil). - 2025-10: Army announces Janus Program for next-gen nuclear energy at defense installations (confirmed, army.mil). - 2025-11: Army selects nine candidate installations (incl. Fort Bragg, Fort Hood, Fort Drum, JB Lewis-McChord, Redstone Arsenal) for microreactor siting; DIU issues Area of Interest for vendors (confirmed, army.mil/thenationaldesk). - 2026-02-15: Containerized nuclear reactor hardware physically moved to Hill AFB, Utah (confirmed, executivegov.com). - 2026-04: ANPI program narrows from 8 to 3 vendor finalists; Buckley SFB (CO) and Malmstrom AFB (MT) advanced as sites, targeting deployment by 2030 (reported, thenextweb.com/gazette.com). - 2026-05-08: Air Force issues Solicitation AFCEC-26-R-0006, an RFP for AI Data Center Development at Joint Base Elmendorf-Richardson, Eielson AFB, and Clear Space Force Station — first explicit pairing of AI data center procurement with a base that already hosts an active microreactor site designation (confirmed, highergov.com). - 2026-06-18: Army official confirms Janus milestone-based OTA award structure under EO 14299 (reported, executivegov.com). - 2026-08: Eielson AFB schedules public Microreactor Project Town Hall (confirmed, eielson.af.mil). # Event Will the US start the process of building a nuclear-powered data center on a military base before 2030 (KXDATACENTER-30)? # Outcomes to forecast Yes / No # Kalshi market anchor YES = 53% (current). 7-day change: flat (+0%). 30-day change: -2%. Avg daily volume: ~50 contracts (thin market). Historical range 26%-75% over 114 data points — high volatility reflecting news-driven repricing on program announcements. # Sub-question answers 1. **Kalshi price/history** — Currently 53%, down slightly from 30 days ago; has ranged 26%-75% since listing, indicating sensitivity to program news (kalshi_direct). 2. **What counts as "starting the process"** — No explicit rulebook criteria provided; ambiguous whether site selection/RFP/contract award vs. actual groundbreaking counts (research silent; code_execution sensitivity analysis shows ~19-point swing between strict vs. loose reading). 3. **Janus/ANPI status** — Both active: Janus selected 9 Army sites (Nov 2025) and moved hardware to Hill AFB (Feb 2026); ANPI narrowed to 3 vendors for Buckley SFB and Malmstrom AFB (Apr 2026). No site yet explicitly pairs a reactor with an AI data center, though Eielson's RFP (May 2026) links AI data center development to a base with an existing reactor site designation (claude_news). 4. **EO 14299 and data centers** — The EO mandates an operational Army-regulated reactor at a domestic installation by 2028-09-30; it does not explicitly reference data centers/AI compute (army.mil), though secondary reporting frames the program as a "proving ground for reactors that could also power AI data centres" (thenextweb.com). 5. **Specific bases** — Eielson AFB (Alaska): furthest along, Oklo-designated preferred site, target operational ~2028, now also subject of AI data center RFP. Buckley SFB (CO) and Malmstrom AFB (MT): ANPI finalists. Nine Army sites (Fort Bragg, Fort Hood, Fort Drum, JB Lewis-McChord, Redstone Arsenal, others) under Janus consideration. Hill AFB (UT): received physical reactor hardware Feb 2026. 6. **DOE federal-site AI data centers** — No research found on Idaho NL/Oak Ridge/Savannah River/Paducah/Portsmouth AI data center projects; gap in research (GDELT rate-limited, no other source coverage). 7. **Pilot timelines/cancellation rates** — Project Pele (est. 2020) has not yet produced an operating reactor; no cancellation reported but multi-year delays are typical for DoD microreactor pilots (Wikipedia). No explicit failure-rate data found. # Key facts (high-confidence, factual) 1. [army.mil] EO 14299 legally mandates an operational military-base reactor by 2028-09-30. 2. [army.mil] Janus Program launched Oct 2025; 9 candidate sites named Nov 2025. 3. [thenextweb.com] ANPI narrowed to 3 vendors; Buckley SFB and Malmstrom AFB targeted by 2030. 4. [highergov.com] AI Data Center RFP (AFCEC-26-R-0006) issued May 2026 for Eielson AFB and nearby installations — first direct data-center/military-base procurement link found. 5. [executivegov.com] Physical reactor hardware transported to Hill AFB, Feb 2026. # Cross-market signals - Kalshi related: KXFUSION-30 (nuclear fusion by 2030) priced at 31%, showing lower confidence in a harder physics milestone vs. this policy-driven build-out question. - Polymarket: No matching markets found (0/100 scanned). - Sportsbook implied: N/A (not applicable to this event type). # Analyst opinions and speculation - claude_news synthesis: "highly likely" the process formally begins before 2030 given RFP/reactor-site overlap at Eielson, but no single project yet explicitly merges both elements. - code_execution Fermi model: central estimate P(YES) ≈ 0.40-0.45, range 0.30-0.55, driven mainly by (a) program schedule slippage risk, (b) whether any reactor gets explicitly billed "for a data center" vs. general base power, (c) resolution strictness. # Directional lean per outcome - **Yes**: Multiple concurrently advancing programs (Janus, ANPI, Eielson/Oklo), statutory 2028 deadline, physical hardware movement already occurring, and a direct AI-data-center RFP overlapping a reactor-designated base (Eielson) all support an early "start of process" before 2030. - **No**: No project yet explicitly labeled "nuclear-powered data center"; most DoD reactor efforts are framed as general base/microgrid power, not data-center-specific; DoD pilot programs (e.g., Project Pele) have historically slipped years past initial targets; resolution ambiguity could favor a strict reading requiring explicit reactor-data center pairing. # Gaps / unknowns - No official Kalshi resolution criteria/rulebook language on what counts as "starting the process." - No DOE federal-site AI data center research found (Idaho NL, Oak Ridge, etc.) — unclear if these could satisfy "military base" criterion or are irrelevant. - No confirmed instance of a reactor project explicitly billed as powering a "data center" (vs. general grid/microgrid power) as of Aug 2026. - GDELT news search was rate-limited, reducing corroboration of claude_news claims from independent outlets. # Calibration anchors - Kalshi current YES price: 53% (primary anchor). - Fermi/model central estimate: 40-45%, range 30-55%. - Comparable Kalshi market (fusion by 2030): 31%, illustrating market discount for technical/definitional uncertainty in similar "before 2030" science-policy questions.

1. Both forecasters center near the Kalshi price (57%/48%) but neither explains why they should be anchored to a thin, volatile 50-contract market with a 26-75% historical range—this market has shown large news-driven swings, so "anchoring" here may just be inheriting noise rather than genuine signal, and neither forecast discusses discounting for thinness explicitly in their probability derivation. 2. Neither forecaster meaningfully engages with the Fermi/code_execution model's 40-45% central estimate as a competing reference class; Forecast 1 acknowledges it only to override it upward, and Forecast 2 doesn't mention it at all despite the brief flagging it as a key calibration anchor alongside Kalshi. 3. Both treat "starting the process" as likely to resolve YES under a loose reading (RFP/site selection counts), but the brief's sensitivity analysis shows "~19-point swing between strict vs. loose reading" — neither forecast quantifies how much probability mass they're assigning to each resolution interpretation, making their point estimates hard to audit despite this being the single largest source of uncertainty identified. 4. The comparable KXFUSION-30 market priced at 31% for a similarly ambiguous "before 2030" science-policy milestone is cited as a cross-market signal in the brief but is not addressed by either forecaster; this comparable suggests markets systematically discount "before 2030" technical/definitional milestones below 50%, which would argue for a lower probability than either forecast gives. 5. Neither forecaster explores the scenario where DOE federal-site AI data centers (Idaho NL, Oak Ridge, etc.) could satisfy the "military base" or resolution criteria via an entirely different pathway than DoD microreactor programs — this is flagged as a research gap in the brief but dropped rather than treated as tail-risk upside for YES. 6. Both forecasts underweight the historical base rate of DoD pilot slippage (Project Pele since 2020, still no operating reactor after ~6 years) relative to the aggressive convergence narrative; given EO 14299's 2028 deadline is itself a mandate rather than a demonstrated outcome, the forecasts may be over-crediting policy announcements as evidence of actual progress rather than applying a stronger discount for the "announcement-to-execution" gap typical in this domain.

53% (50% confidence)

Yes: 53% No: 47%

The Kalshi anchor sits at 53% and has been flat, and while the market is thin and volatile, no piece of evidence in the brief clearly justifies a large departure from it. The bullish case is real: EO 14299's statutory 2028 reactor deadline, Janus site selections, ANPI vendor narrowing for Buckley/Malmstrom, reactor hardware delivered to Hill AFB, and the May 2026 AFCEC AI Data Center RFP at Eielson (already an Oklo microreactor preferred site) — with 3.5 years of runway and 'start the process' being a low bar under most readings. Offsetting this, no project is yet explicitly billed as a reactor powering a data center, DoD microreactor pilots slip chronically (Pele since 2020 with no operating unit), and the comparable KXFUSION-30 at 31% plus the Fermi 40-45% estimate suggest markets/models discount ambiguous 'before 2030' milestones. Balancing the ~19-point strict-vs-loose resolution swing (I put roughly 60% weight on a loose reading where site selection/RFP/contract award suffices), I land essentially at the market: 53% Yes, between the two forecasts and only modestly above the Fermi model.

Total pipeline time: 304.2s