← Back to scans

Will the US start the process of building a nuclear-powered data center on a military base before 2030?

KXDATACENTER-30 · Science and Technology · 2026-04-22
60%
Agent
55%
Market Price
+5.0%
Edge
45%
Confidence
Volume: 6,523
Spread: 7.0c
Days to resolution: 1349
Markets in event: 1
Final Rationale
My independent estimate of 67% relied on a liberal reading of 'start the process' that likely over-credits the ANPI program, which is fundamentally a base-power microreactor initiative with data centers as one possible tenant use case, not a dedicated nuclear-powered data center program. I accept Challenges 1, 2, and 4: the ANPI/Pele reference class suggests DoD nuclear timelines slip (Pele announced 2019, still not built), and my three sub-questions measure milestones of the same pathway rather than independent routes, so weighted-averaging inflates the estimate. I partially reject Challenge 5 since the evidence-driven perspective captures real momentum (Radiant's 2028 target, INL testing summer 2026, Trump EOs bypassing some NRC review on military bases). Settling at 60% reflects the genuine positive momentum while respecting resolution strictness; the modest premium over the shallow 55% market is justified by specific April 2026 ANPI site-selection evidence that low-volume traders may not have priced.
Price vs. prediction
How the market moved after the agent's call. Solid blue line is the actual market price; dashed line is the agent's static forecast at $t{=}0$.
$n = 58$ follow-ups
1. Planner Opus · Fermi decomposition + tool plan
Tools chosen: kalshi_data article_search web_search wikipedia_lookup congress_bills
Sub-questions (Fermi decomposition)
  1. Has the US government already announced or initiated a formal program to build a nuclear-powered data center on a military base as of April 2026? (w=0.30)
  2. Will the US government announce a concrete plan or contract (not just exploratory study) to build a nuclear-powered data center on a military base by end of 2027? (w=0.35)
  3. Will the regulatory, political, and technical barriers to nuclear-powered military data centers be sufficiently resolved to allow construction to begin before 2030? (w=0.35)
Planner reasoning
This question sits at the intersection of several fast-moving trends: the AI/data center buildout, the nuclear energy renaissance (SMRs, microreactors), and military energy security concerns. The key distinction is 'start the process of building' on a military base — this is a low threshold (groundbreaking or formal construction contract suffices, not commissioning). Reference classes: (1) Project Pele — the DoD's mobile nuclear microreactor program — reached prototype stage by 2024 but faced termination/delays; (2) Big Tech (Google, Amazon, Microsoft) signing nuclear PPAs for data centers, but none yet on military bases; (3) Historical DoD nuclear infrastructure timelines (nuclear propulsion, base reactors) suggest multi-year procurement cycles. Causal mechanisms favoring YES: AI energy demand creating pressure, Trump administration's pro-nuclear executive orders, DoD energy resilience mandates, co-location of data centers with military bases being explored. Mechanisms against: NRC/DOE licensing for microreactors takes years, NIMBY/environmental review, budget constraints, Project Pele cancellation signal. Temporal analysis: With 1,349 days (~3.7 years), to resolve YES by Jan 1, 2030: a program decision would need to be made in 2026-2027, contracts awarded in 2027-2028, and groundbreaking by late 2028-2029. This is tight but not impossible given the current political environment. The question asks only for 'start the process' which could mean a signed contract or construction permit, not completed facility. I'll assess current market pricing and news to calibrate base rates.
2. Research Parallel tool calls · legacy v2
ToolStatusTimeItemsSummary
kalshi_data OK 1.8s -
article_search OK 0.1s -
web_search OK 78.0s -
wikipedia_lookup OK 1.8s -
congress_bills OK 0.8s -
3. Evidence Synthesis Sonnet · 12 items · legacy v2
#StrengthCredibilityDirectionSourceClaimPriced In
1 STRONG 82 ↑ UP web_search The Department of the Air Force, via the Advanced Nuclear Power for Installations (ANPI) program, selected Buckley Space Force Base (CO) and Malmstrom Air Force Base (MT) as the first sites to receive nuclear microreactors, with future data center support explicitly mentioned as a use case (April 2026). No
2 MODERATE 72 ↑ UP web_search Radiant Nuclear has announced an agreement with the Defense Innovation Unit and the Air Force to deliver a microreactor to an American military base, with commercial availability targeted by 2028. No
3 MODERATE 80 ↑ UP web_search In February 2026, the Pentagon and DOE conducted the first-ever airlift of a nuclear microreactor (unfueled) from California to Utah, demonstrating rapid deployment capability for military and data center use. Yes
4 MODERATE 75 ↑ UP web_search The ANPI program explicitly references powering data centers (including AI) and space/cyber infrastructure as future mission requirements for the military bases selected, but no formal data center construction contract has been announced yet. No
5 MODERATE 70 ↑ UP web_search Radiant Nuclear's Kaleidos microreactor (1 MW) is planned for testing at Idaho National Laboratory's DOME facility in summer 2026, with the Air Force agreement signaling potential military base deployment by 2028. No
6 MODERATE 68 ↑ UP web_search The ANPI program is described as calling for deployment of 'advanced, contracted' nuclear power at military installations, suggesting a procurement/contract pathway is already underway as of April 2026. No
7 MODERATE 70 ↓ DOWN web_search No specific contract or groundbreaking for a nuclear-powered data center on a military base has been confirmed; current programs focus on microreactor deployment for general power resilience with data centers mentioned as future applications. Yes
8 MODERATE 65 ↑ UP web_search The Trump administration's pro-nuclear executive orders and AI energy demand are providing political tailwinds for expedited nuclear deployment at military installations, reducing some political barriers. Yes
9 MODERATE 72 ↑ UP wikipedia_lookup Nuclear microreactors (1–20 MWe) are designed specifically for isolated military bases and can be transported by road, rail, or air — technical feasibility for military deployment is established, though regulatory licensing remains a barrier. Yes
10 MODERATE 68 ↓ DOWN web_search Radiant's microreactor requires fueling from the Nevada National Security site and INL testing before deployment, indicating regulatory and fueling logistics steps remain before any military base construction begins. Yes
11 MODERATE 55 ↑ UP kalshi_data Kalshi market prices the question at 55% probability with modest trading volume (~65 contracts/day), suggesting moderate market confidence in YES resolution by 2030. Yes
12 WEAK 45 ↓ DOWN congress_bills No relevant Congressional legislation specifically targeting nuclear-powered military data centers was found in the 10 bills retrieved, suggesting legislative pathway is not yet formalized. Yes
Information Gaps
  • No confirmation of whether ANPI program contracts have been formally awarded or are merely in solicitation/planning phase — the web search was truncated and did not confirm contract award details.
  • No data on NRC or DOE licensing status for Radiant or other microreactor developers seeking to deploy on military bases specifically for data center use.
  • No evidence of any existing nuclear-powered data center on a military base anywhere in the world as a reference class comparator.
  • Unclear whether 'start the process of building' under the question's resolution criteria would be satisfied by the ANPI program's site selection and contracting, or requires physical groundbreaking.
  • No detail on budget appropriations for ANPI or Radiant's Air Force agreement — whether funded or only an MOU.
  • Project Pele's current status (previously signaled potential cancellation/delays) not addressed in the research — unclear if lessons learned affected ANPI design.
  • No information on whether Buckley or Malmstrom data center plans have a separate construction contract or timeline distinct from the microreactor deployment.
Key Uncertainties
  • Resolution criteria ambiguity: Does 'start the process of building' require groundbreaking, a signed construction contract, or merely a formal program decision with site selection? The ANPI program may already qualify.
  • Whether Radiant or another vendor successfully completes INL testing in 2026 on schedule, which gates the 2027-2028 military deployment window.
  • NRC/DOE licensing timeline for microreactors on military bases — this has historically been the longest lead-time item and could push construction past 2030.
  • Whether the data center co-location requirement becomes a formal component of ANPI contracts, or remains aspirational language in program documents.
  • Stability of Trump administration nuclear policy priorities and DoD budget allocations through 2029.
  • Whether 'nuclear-powered data center on a military base' requires the reactor to be purpose-built for the data center vs. a microreactor on-base that also powers a data center.
Market Context
Price: 0.55 · 7d: +0.02 · 30d: +0.00
Volume: LOW-MODERATE · Depth: SHALLOW
Reliability: LOW-MEDIUM — only ~65 contracts/day average volume; price has been stable over 30 days suggesting limited new information being priced in; the +2% 7-day move may reflect the April 2026 ANPI site selection news
4. Independent Forecaster Opus · No market price · legacy v2 Did NOT see market price
Independent estimate: 67% (48% confidence)
base rate
62%
50% conf
evidence driven
78%
55% conf
contrarian
62%
40% conf
Spread
16pp
High disagreement
sq1: Has the US government already announced or initiated a formal program to build a nuclear-powered data center on a military base as of April 2026?
72%
The ANPI program has selected Buckley and Malmstrom as sites with data center support explicitly mentioned as a use case. Radiant has an Air Force agreement. This arguably constitutes 'initiating a formal program' under a liberal reading. However, no formal data center construction contract is confirmed — data centers are mentioned as 'future mission requirements.' The question asks about a nuclear-powered data center specifically, not just a microreactor on base. Base rate for ambiguous 'process started' resolution tends to resolve YES when programs exist. ~72%.
sq2: Will the US government announce a concrete plan or contract (not just exploratory study) to build a nuclear-powered data center on a military base by end of 2027?
55%
Base rate for DoD major construction contracts being awarded within 2 years of program initiation is moderate (~40-50%). ANPI already in contracting pathway, Radiant targeting 2028 commercial availability, INL testing summer 2026. Political momentum is strong. However, DoD procurement frequently slips, and NRC licensing could delay. The question specifically requires nuclear-powered *data center* — the ANPI contracts appear to be for microreactors with data centers as a possible future tenant. Still, by end of 2027, it's quite plausible a combined contract gets announced given AI energy pressure. ~55%.
sq3: Will the regulatory, political, and technical barriers to nuclear-powered military data centers be sufficiently resolved to allow construction to begin before 2030?
50%
Historical microreactor licensing timelines have been 3-5+ years. Radiant testing at INL in 2026 is promising but fueling logistics and NRC/DOE approvals remain. Trump EOs and DoD authority on military bases (potentially bypassing some NRC review) create a faster pathway — military bases have historically been able to deploy nuclear assets under DoD authority. Project Pele's struggles are a cautionary signal. 'Start construction' is a lower bar than 'operational.' Groundbreaking by late 2029 is plausible but not certain. ~50%.
Show full reasoning trace
base rate: Reference classes: (1) DoD nuclear infrastructure programs historically take 5-10+ years from announcement to construction — base rate ~30-40% for hitting 3-4 year windows. (2) Microreactor deployment programs (Project Pele) have faced significant delays/cancellations. (3) However, ambiguous 'start the process' resolution criteria with active programs tend to resolve YES ~60-70% of the time.
evidence updates: Strong UP: ANPI program has selected specific military bases with data center use cases explicitly mentioned (+15% from base). Radiant Air Force agreement with 2028 target (+5%). February 2026 microreactor airlift demonstration shows operational momentum (+3%). DOWN: No confirmed construction contract specifically for data center (-5%). Licensing/fueling logistics remain (-3%). Project Pele precedent of delays (-3%).
combination method: Weighted average of three sub-questions, with adjustment for positive correlation and liberal interpretation of 'start the process.'
final: The combination of active ANPI program with site selection, explicit data center mentions, Radiant's contracted pathway, and a ~3.7-year window creates a moderately favorable case. The liberal resolution criterion ('start the process') likely captures formal program decisions + contract awards, which seem plausible by 2029. Main risks are resolution ambiguity about whether microreactor deployment that happens to power a data center counts, and DoD procurement delays. Settling at 0.62 with medium confidence.
ensemble: {'ensemble_count': 3, 'perspectives': {'base_rate': {'p_yes': 0.62, 'confidence': 0.5}, 'evidence_driven': {'p_yes': 0.78, 'confidence': 0.55}, 'contrarian': {'p_yes': 0.62, 'confidence': 0.4}}, 'spread': 0.16, 'individual_p_yes': {'base_rate': 0.62, 'evidence_driven': 0.78, 'contrarian': 0.62}}
5. Devil's Advocate Sonnet · legacy v2 Market price revealed
Reasoning Flaw Missing Info Asymmetric Evidence
Challenges
  1. Resolution criteria ambiguity is under-examined. The question asks about 'a nuclear-powered data center' — the ANPI program is primarily about powering military installations with microreactors, with data centers as ONE possible use case. Sq1's 72% may be too generous: 'data centers mentioned as future mission requirements' is quite different from 'initiating a program to build a nuclear-powered data center.' A strict resolver could easily say the ANPI is a microreactor program, not a data center program.
  2. Positive correlation between sub-questions is acknowledged but not properly handled. Sq1, sq2, sq3 are not independent — they largely describe the same underlying program trajectory. The combination method 'weighted average' is not mathematically principled for this question structure. The question resolves YES if ANY qualifying process starts before 2030, which is closer to a disjunction than a conjunction — but the three sub-questions aren't capturing different pathways, they're capturing different milestones of one pathway.
  3. The 'liberal interpretation' assumption is doing a lot of work (moving from ~50% to 67%) but isn't justified against specific resolution source wording. Superforecasters should anchor to actual resolution criteria, not assumed generosity.
  4. Project Pele precedent is mentioned but under-weighted. Pele was announced ~2019 with aggressive timelines and has faced significant delays/restructuring over 6+ years. That's the closest reference class and it suggests ANPI-to-construction in 3.7 years is optimistic.
  5. Ensemble shows evidence_driven at 78% pulling the mean up, but with a 16-point spread and acknowledged information gaps (contract award status, funding, licensing), the forecaster should lean toward the more conservative perspectives, not the weighted mean.
Suggested adjustment: -8pp
The forecast should be modestly lower due to (1) under-weighting of resolution criteria strictness — ANPI is a base power program, not a data center program, (2) Project Pele base rate suggesting DoD nuclear timelines slip, and (3) the 9-UP/3-DOWN evidence count includes several items that are redundant signals of the same underlying program rather than independent confirmations. A forecast around 0.58-0.60 better reflects these concerns. This is not about market conformity — the market happens to be at 55% but the adjustment is driven by the resolution-criteria and correlation issues.
Market comparison: Forecaster: 67.3%, Market: 55%. The 12-point divergence is partially justified by the forecaster's detailed evidence synthesis (ANPI site selection, Radiant agreement, February 2026 airlift), which the thin market may not have fully priced. However, the forecaster's liberal interpretation of 'start the process' and weighted averaging of correlated sub-questions likely inflates the estimate. Market at 55% with stable 30-day price despite recent ANPI news suggests informed traders see meaningful resolution risk. A forecast in the 58-62% range would better balance the genuine positive evidence against resolution ambiguity.
6. Aggregator Opus · final calibrated forecast
60% (45% confidence)
My independent estimate of 67% relied on a liberal reading of 'start the process' that likely over-credits the ANPI program, which is fundamentally a base-power microreactor initiative with data centers as one possible tenant use case, not a dedicated nuclear-powered data center program. I accept Challenges 1, 2, and 4: the ANPI/Pele reference class suggests DoD nuclear timelines slip (Pele announced 2019, still not built), and my three sub-questions measure milestones of the same pathway rather than independent routes, so weighted-averaging inflates the estimate. I partially reject Challenge 5 since the evidence-driven perspective captures real momentum (Radiant's 2028 target, INL testing summer 2026, Trump EOs bypassing some NRC review on military bases). Settling at 60% reflects the genuine positive momentum while respecting resolution strictness; the modest premium over the shallow 55% market is justified by specific April 2026 ANPI site-selection evidence that low-volume traders may not have priced.