Tools chosen: kalshi_direct claude_news claude_news kalshi_related polymarket_related wikipedia

Sub-questions (Fermi decomposition)

Planner reasoning

This is a long-horizon (2030) question about an emerging tech/policy initiative combining nuclear power, data centers, and military bases. The key drivers are recent DoD/DoE announcements and pilot programs (e.g., microreactor programs like Project Pele) and any specific co-located data center plans. Market price anchors the consensus, and news search will reveal whether any concrete process has already started.

# Event Will the US start the process of building a nuclear-powered data center on a military base before 2030? # Outcomes to forecast - **Yes**: US starts the process of building a nuclear-powered data center on a military base before Jan 1, 2030 - **No**: Does not happen before Jan 1, 2030 # Kalshi market anchor **Current YES price: 54%** | 7-day change: -2% | 30-day change: -4% | Avg daily volume: 56 contracts | Historical range: 26%–64% *Mild recent drift downward; low volume = thin market susceptible to mispricing* # Sub-question answers 1. **Has DoD/US agency announced or initiated a nuclear-powered data center project on a military base?** — Yes, partially. Army entered exclusive negotiations (March 2026) with Carlyle/CyrusOne to build hyperscale data centers on Army posts; Army's Janus Program simultaneously targets nuclear microreactors at nine Army bases. The Air Force announced nuclear data center co-location consideration at Eielson AFB and JBER Alaska. [Stars & Stripes, DefenseOne, Air Force announcements] 2. **Status of DoD microreactor programs (Project Pele, Eielson)?** — Project Pele: fuel delivered to INL (Nov 2025), construction underway, electricity expected ~2028. Eielson AFB: selected as preferred site for DoD microreactor pilot, target ~2028. Neither is explicitly labeled "data center" but Eielson is explicitly linked. [DOE, BWXT, DefenseOne] 3. **EOs/DOE-DoD partnerships linking AI data centers to nuclear on federal/military land?** — EO 14299 (May 2025) mandates nuclear reactor on military base by Sept 30, 2028. July 2025 EO directs DoD to lease military land for data centers. DOE issued RFAs (Sept–Oct 2025) for nuclear-powered AI data centers at federal sites (INL, Oak Ridge, Paducah, Savannah River). [White House, DOE, Morgan Lewis] 4. **What does 'start the process' require and is it feasible before 2030?** — The resolution standard is low ("start the process"), not completion. Given Army groundbreaking target of 2027 for microreactors, explicit data center co-location plans on military land, and EO mandates, "starting the process" via site selection, contracts, or groundbreaking is highly plausible well before 2030. NRC licensing hurdles are reduced on military bases per congressional commentary. [DefenseOne, Breaking Defense] 5. **Private-sector deals signaling momentum?** — Carlyle/CyrusOne negotiating for Army base data centers (March 2026); tech firms (Google, Microsoft) pursuing SMR deals broadly; Air Force ANPI selected 8 vendors including Oklo, Kairos, Westinghouse for base reactors by 2030. [SMR Wikipedia, Stars & Stripes, Air Force] # Key facts (high-confidence, factual) 1. [White House, May 2025] EO 14299 legally mandates nuclear reactor operation on a military base by Sept 30, 2028 2. [DefenseOne, Oct 2025] Army's Janus Program targets groundbreaking on microreactor at US base by 2027; nine sites identified 3. [Stars & Stripes, March 2026] Army in exclusive negotiations with Carlyle/CyrusOne for hyperscale data centers on Army posts 4. [Air Force, April 2026] ANPI selected vendors for reactors at multiple DAF bases; Joint Base San Antonio paired with Antares Nuclear 5. [DefenseOne, June 2026] Army planning integrated data center + power generation campus on base 6. [DOE, Sept–Oct 2025] RFAs issued for nuclear-powered AI data centers at DOE federal sites; construction intent by end 2025 7. [C&EN, Oct 2025] Government shutdown caused brief delays; draft RFP for Janus was pending "weeks" away as of late 2025 # Cross-market signals - **Kalshi related**: Only market found is this one (KXDATACENTER-30, 54%). No directly comparable markets. - **Polymarket**: No matching markets found. - **Sportsbook implied**: N/A # Analyst opinions and speculation - Congressional Rep. Harrigan explicitly noted military base nuclear for data centers is feasible under relaxed NRC jurisdiction [Breaking Defense] - Pentagon evaluating 300 MW floating nuclear plant concept for military base by 2028 [Army Recognition] - Late-2025 commentary: no US data center yet powered by SMR; infrastructure still being built [ASCE] - Concern noted that DoD "breakthrough nuclear moment risks slipping away" if not locked in quickly [War on the Rocks] # Directional lean per outcome - **Yes (strong lean)**: EO mandate with hard 2028 deadline; groundbreaking target 2027; active vendor selection; explicit Army data center + nuclear campus planning; resolution bar is low ("start the process" not "complete"); multiple parallel programs (Janus, ANPI, Eielson, Project Pele) - **No**: Threshold ambiguity — must be *nuclear-powered data center specifically*, not just a nuclear reactor OR a data center separately on the same base; government shutdown delays; no SMR yet operational in US; fusion of both elements (nuclear + data center, same base) not yet formally contracted # Gaps / unknowns - Whether "start the process" requires both nuclear and data center to be explicitly co-designed, or if parallel programs on same base count - Exact contract language/status for Carlyle/CyrusOne Army deal and whether nuclear power is contractually tied - Whether Janus/ANPI groundbreakings will explicitly incorporate data center co-location in initial scope # Calibration anchors - **Kalshi current YES: 54%** (anchor, mild downward drift) - EO 14299 creates a legal mandate with 2028 deadline — unprecedented policy driver - Base rate for DoD programs meeting stated milestones: moderate (budget/political delays common) - Resolution threshold ("start the process") is notably low — site selection + contract likely sufficient - **Assessment**: Evidence supports YES probability meaningfully above 54%, likely 70–80%, given the explicit EO mandate, active groundbreaking timelines, and low resolution bar

1. Both forecasters under-weight the threshold ambiguity risk. The brief explicitly flags that the question may require nuclear and data center to be "explicitly co-designed" rather than parallel programs on the same base — yet neither forecaster seriously quantifies how likely it is that the resolution criterion catches a scenario where Janus groundbreaks a microreactor and Carlyle/CyrusOne groundbreaks a data center on the same Army post but under separate contracts with no explicit power-supply linkage. This ambiguity could easily sink a Yes resolution even with full programmatic success. 2. Neither forecaster adequately engages with the base rate for DoD programs meeting stated milestones. The brief notes "budget/political delays common," the C&EN fact confirms a government shutdown already caused delays, and "War on the Rocks" commentary warns the nuclear momentum "risks slipping away" — yet both forecasters treat the 2027 Janus groundbreaking target as near-certain input rather than an aspirational milestone subject to the historically poor on-time delivery record of DoD acquisition programs. 3. Both forecasters effectively ignore the operational gap between nuclear *reactor* programs (Project Pele, Eielson, Janus) and nuclear *data center* programs. The brief states no US data center is yet powered by an SMR, and EO 14299 mandates a nuclear reactor on base — not a nuclear-powered data center — leaving a meaningful gap between the legal mandate and the specific event being resolved. 4. The Kalshi anchor drift is under-analyzed. The market has moved from 64% to 54% over the recent period (–10 pp from historical high) with mild continued downward drift; both forecasters dismiss this signal too quickly by citing thin volume, without considering that informed traders may have updated on the ambiguity in co-design requirements or on procurement delays that are not fully reflected in the public evidence brief. 5. Neither forecaster considers the political/administrative tail risk adequately: a change in administration priorities, continued CRs or government shutdowns, or a DoD budget sequester could stall all of Janus, ANPI, and the Carlyle/CyrusOne negotiations simultaneously, and the brief's only mitigation is the EO mandate — but executive orders can be rescinded or deprioritized, particularly in a new fiscal environment.

66% (50% confidence)

Yes: 66% No: 34%

The evidence strongly supports a low resolution bar ('start the process') and multiple converging programs (EO 14299 mandate, Janus 2027 groundbreaking, Carlyle/CyrusOne negotiations, ANPI vendor selection), justifying a YES probability above the thin 54% Kalshi anchor. However, the devil's advocate raises legitimate concerns: threshold ambiguity over whether parallel nuclear and data-center programs count as one co-designed 'nuclear-powered data center,' DoD's poor milestone track record, and the recent downward market drift. I settle slightly below both forecasters' ~72-76% at 66% to account for the co-design ambiguity and procurement/political delay risks the market drift may be signaling.

Total pipeline time: 109.8s