brain/
conceptstock-market

Nuclear as the only viable AI data center baseload

Notes

Nuclear as the only viable AI data center baseload

One-line summary: AI data centers require 24/7 firm power that solar and wind cannot provide without storage that doesn't exist at scale — nuclear is the only current technology that meets the requirement, and Microsoft's Three Mile Island PPA (reportedly >2x spot rate) established that hyperscalers will pay a significant premium.

The insight

AI training runs and inference workloads cannot tolerate intermittent power. A data center running on renewable energy requires either firm backup (gas peaker plants) or grid-scale storage (not yet viable). Nuclear provides what solar/wind cannot: baseload, carbon-free, 24/7. The combination of ESG-driven decarbonization targets and reliability requirements creates structural demand for nuclear that didn't exist at scale five years ago.

The chain

AI DC load growth → PJM capacity prices clear 10× ($329/MW-day, data centers 63% of increase) → dispatchable nuclear earns structural premium in capacity markets and hyperscaler PPAs → CCJ/UEC (uranium producers) and CEG (nuclear operator) benefit from derived demand and pricing. Canonical: pjm-capacity-prices-to-nuclear-premium.

Canonical: pjm-capacity-prices-to-nuclear-premium.

Evidence

  • From 2026-05-01-all-in-podcast-openai-misses-targets-codex-vs-claude-elon-vs: David Friedberg (All In): "Microsoft's Three Mile Island PPA was more than 2x the prevailing spot rate" — establishing that hyperscalers will pay a substantial reliability premium for nuclear power.
  • From 2026-05-11-autoresearch-macro-energy-critical-minerals-may-2026: Nuclear is identified as the only viable baseload for AI data centers in the current technology landscape. Solar/wind paired with storage doesn't meet the uptime requirements at reasonable cost.
  • From 2026-05-11-autoresearch-macro-semis-ai-infrastructure-may-2026: CSP CapEx $830B in 2026 (+79% YoY) is the upstream demand signal for all AI infrastructure inputs, including power.
  • From 2026-05-11-autoresearch-macro-energy-critical-minerals-may-2026: Microsoft has a nuclear PPA with Constellation Energy (CEG). Other hyperscalers (AWS, Google) are pursuing similar long-term power agreements to meet data center baseload requirements and decarbonization targets simultaneously.
  • From 2026-05-15-autoresearch-semis-ai-infra-ma-partnerships-may-12-15: NANO Nuclear Energy + Supermicro MOU (May 6, 2026): microreactor + AI server integration agreement — an early signal that hyperscale AI server vendors are actively partnering with microreactor companies to co-locate power and compute. Supermicro is a major server supplier (>10% AI server market share); its public commitment to microreactor co-location strengthens the SMR-for-data-center thesis. Terrestrial Energy + Riot Platforms (May 6): nuclear-powered data center deal; Riot Platforms (bitcoin mining → AI transition) signing with Terrestrial Energy to co-locate a molten-salt reactor with a data center, further validating demand for direct nuclear co-location beyond traditional PPA structures.
  • From 2026-05-18-autoresearch-macro-bucket-scan-may-18-2026: Meta signed agreements for up to 7.8 GW of nuclear capacity (Q1 2026) — the largest single-entity nuclear PPA bundle yet. Includes Oklo + Meta 1.2GW nuclear campus (Pike County, Ohio; Meta prepaying and funding project advancement). Microsoft has arranged >800 MW exclusively for data centers, including Three Mile Island Unit 1 under a 20-year PPA. The dominant corporate energy procurement model has shifted from "passive renewable PPAs" to "active, strategic underwriting of new firm power generation." — highly confirmatory for this thesis. US data center electricity demand: 176 TWh → 580 TWh estimated by 2028.
  • From 2026-05-18-autoresearch-macro-bucket-scan-may-18-2026: Uranium pricing: spot spiked to $100/lb in January 2026 (first time in 2 years), corrected to ~$89/lb; term prices at 14-year high of $90/lb; no new mine supply before 2030 — structural tightness. CCJ is the primary listed uranium exposure; CEG is the operator.
  • From 2026-05-20-autoresearch-nuclear-uranium-meta-ppa-ccj-ceg-smr-may-2026: Uranium spot updated: $85.25/lb as of May 18, 2026 (-1.90% past month, +19.6% YoY) — pulled back ~4% from the $89/lb wiki figure; downtrend near-term but YoY bull trend intact. See ccj.
  • From 2026-05-20-autoresearch-nuclear-uranium-meta-ppa-ccj-ceg-smr-may-2026: Meta Vistra portion starts LATE 2026 — near-term catalyst. Vistra delivers 2,176 MW from existing Perry/Davis-Besse plants to Meta with delivery beginning late 2026, plus 433 MW incremental from uprates by 2034. This is the first major hyperscaler nuclear PPA at scale with a confirmed near-term start date; makes Meta/Vistra a concrete 2026 catalyst rather than a 2030+ thesis. The 6.6 GW total (Vistra 2.6GW + TerraPower 2.8GW + Oklo 1.2GW) from January 2026 announcement is additive to prior ~1.2GW Constellation deal (~7.8GW total for Meta).
  • From 2026-05-20-autoresearch-nuclear-uranium-meta-ppa-ccj-ceg-smr-may-2026: CCJ Q1 2026: revenue $845M, adjusted EBITDA $509M, EPS $0.34 (beat $0.31 consensus by 9.7%); uranium sales volume 7.8M lbs (+13% YoY); prices up, costs down. India deal (March 2, 2026): 22M lbs U3O8 over 9 years, total $2.6B at market-related pricing, deliveries 2027–2035. CCJ stock ~$104.89 (+105% YoY). See ccj.
  • From 2026-05-20-autoresearch-nuclear-uranium-meta-ppa-ccj-ceg-smr-may-2026: AWS-Talen 1.92GW Susquehanna PPA transitioned to front-of-meter (FTM) delivery in spring 2026 (after scheduled refueling outage). $18B deal through 2042; full 1.92GW by 2032. This is now operational — the first large-scale grid-connected hyperscaler nuclear PPA running.
  • From 2026-05-20-autoresearch-nuclear-uranium-meta-ppa-ccj-ceg-smr-may-2026: CEG Q1 2026 earnings miss (May 11, 2026): Revenue $11.12B (+64% YoY — Calpine acquisition). Adjusted EPS $2.74 (beat $2.59 consensus). BUT 2026 adjusted EPS guidance midpoint $11.50 vs $11.60 consensus → single-day -8% sell-off; stock -11.6% since earnings, -13% YTD. Calpine adds 23 GW of gas/renewable capacity + retail platform; concern is share dilution and integration risk rather than nuclear thesis deterioration. Six analysts revised EPS expectations downward.
  • From 2026-05-26-autoresearch-nuclear-ai-datacenter-ppa-smr-update: Cumulative hyperscaler nuclear commitment: 9.8 GW across 13 projects. NEW: Google signed with NextEra Energy to restart Iowa's 615 MW Duane Arnold Energy Center (shut since 2020, storm damage) — power delivery to Google AI data centers by 2029. Second brownfield restart deal (after Microsoft TMI Unit 1), confirming hyperscalers prefer guaranteed brownfield delivery (2027–2029) over unproven SMRs (2028–2030+). NextEra (NEE) benefits from restart contract — not currently in suggested-tickers. See contradictions/tensions.
  • From 2026-05-26-autoresearch-nuclear-ai-datacenter-ppa-smr-update: Cameco CEO Tim Gitzel (May 5, 2026): "The AI, the data centers, the hyperscalers. It's demand increase like we haven't seen." Data centers are "bringing dead reactors to life." Global context: 78 GW nuclear under construction globally; 38 nations at the 2026 Paris summit pledged to triple nuclear capacity by 2050 — sovereign fuel commitments locking in multi-decade uranium demand.
  • From 2026-05-26-autoresearch-nuclear-ai-datacenter-ppa-smr-update: HALEU fuel scarcity is the binding constraint for most SMR timelines. TerraPower explicitly cited fuel scarcity as reason their 2028 target may slip to 2030+. SMR status: TerraPower NRC permit expected H1 2026; Kairos Power DOE HALEU TRISO fuel contract signed (Hermes 1 startup fuel secured); X-energy NRC safety evaluation complete November 2026; Oklo backed by Altman/Meta.
  • From 2026-05-28-autoresearch-energy-critical-minerals-may-28: Uranium spot late May 2026: ~$86.5/lb (near 2-month high); Q1 2026 long-term uranium price: $91.50/lb avg — approaching highest since 2012 in constant dollars; 2026 peak so far $101.41/lb. CCJ Q1 2026: EPS $0.34 (~30% beat vs. $0.26 consensus); uranium revenue $712M (+15% YoY); gross profit +28%; full-year guidance maintained (not raised).
  • From 2026-05-28-autoresearch-energy-critical-minerals-may-28: $80B Westinghouse reactor deployment partnership — Cameco, Brookfield, Westinghouse, and the US Government (announced October 2025, reaffirmed Q1 2026 filings). US Government arranges financing and permits for new Westinghouse reactors in the US; aggregate investment value at least $80B. Categorically different from a single utility PPA — a federal government commitment to a domestic reactor buildout pipeline. CCJ shares jumped 5% on the strength of this framing.
  • From 2026-05-28-autoresearch-energy-critical-minerals-may-28: CEG planned 1 GW of nuclear uprates over the next decade (announced April 21, 2026), including 135 MW at Braidwood and Byron Clean Energy Centers (Illinois). Secured Meta 20-year agreement to keep Illinois reactor operating. Prioritizing long-term hyperscaler contracts for AI datacenter power.
  • From 2026-05-29-autoresearch-nuclear-ai-datacenter-ppa-smr-may-29: Kairos Power Hermes 2 groundbreaking (late April/early May 2026) — first Gen IV commercial reactor to receive an NRC construction permit. 50 MW fluoride salt-cooled high-temperature reactor (FHR); Google PPA via TVA grid for Tennessee/Alabama AI data centers; commercial power target ~2030. This NRC construction permit is the most significant SMR regulatory milestone since Oklo's DOE NSDA approval (March 2026) and de-risks the technology path for subsequent commercial SMR orders.
  • From 2026-05-29-autoresearch-nuclear-ai-datacenter-ppa-smr-may-29: X-Energy IPO: Nasdaq: XE, April 24, 2026 — $1.02B raised at $23/share, above $16–$19 target range, in an upsized offering. Xe-100 HTGR (80 MW, high-temperature gas-cooled reactor) design. The above-range raise signals institutional demand for SMR pure-plays, validating the commercial IPO path for this sector.
  • From 2026-05-29-autoresearch-nuclear-ai-datacenter-ppa-smr-may-29: Oklo 14 GW pipeline: customer pipeline exceeded 14 GW with commitments exceeding $10B. Key deals: Meta pre-payment on 1.2 GW Ohio project; Switch master agreement up to 12 GW; 500 MW with Equinix. Oklo + Vertiv collaboration on nuclear data center power/cooling integration announced — picks-and-shovels crossover: Vertiv is the lead thermal management vendor for AI data centers ([datacenter construction concept]). First Aurora powerhouse broke ground at INL September 2025; first power ~2030.
  • From 2026-05-29-autoresearch-nuclear-ai-datacenter-ppa-smr-may-29: Uranium LT price: $90/lb in Q1 2026 — highest since 2008 in constant dollars. Spot at ~$85–86/lb late May 2026 (easing slightly from Q1; 2026 peak $101.41/lb). Utilities prepaying for long-term contracts "because they've concluded spot availability cannot be relied upon at future buildout volumes" — structural repricing, not cyclical. UEC (Uranium Energy Corp): analyst consensus price target ~$18 (vs ~$13.65 current, ~+32% upside) per tickernerd.com snippet.
  • From 2026-05-30-autoresearch-energy-critical-minerals-uranium-helium-copper-nuclear: PJM capacity prices 10x in one auction cycle: $28.92/MW-day (2024/25) → $329.17/MW-day (2026/27) — an order-of-magnitude increase in just two years. Data centers responsible for 63% of the 2025/2026 auction price increase, translating to $9.3B in cost recovery from PJM customers. US average retail electricity ~19¢/kWh (+27% vs. 2019). Goldman Sachs (Feb 2026): electricity prices rising at double the rate of inflation. This PJM price spike is the direct forcing function behind nuclear PPA urgency — hyperscalers are not just seeking carbon-free power, they're escaping escalating grid capacity charges.
  • From 2026-05-30-autoresearch-energy-critical-minerals-uranium-helium-copper-nuclear: UEC Burke Hollow ISR production commenced (Texas) — world's newest ISR mine, first new US ISR operation in over a decade. Uranium spot $85.20/lb as of May 28 (2-month low; -2.07% past month; +18.91% YoY). Long-term contract price: $90/lb — highest since 2008. CCJ $112.70 (+2.07% May 29); CCJ president: "the forward demand that has yet to come to the market has never been bigger." US government committed to 65M lbs for 10 new reactors; Sprott Physical Uranium Trust resumed active purchasing after 6-month pause in Q1 2026.
  • pippa-malmgren in 2026-05-28-podcast-macro-voices-macrovoices-534-dr-pippa-malmgren-superpower-war (TRISO fuel and portable SMRs): "there's a company...one in California called Valor Atomics. Valor is making a box that's literally half the size of a car...they're putting them in these small boxes that then you can literally move anywhere you want...in Austin...another one called Aloe...The new technology evolves around what they call triso fuel, which are little tiny pellets the size of a poppy seed which have super hard shells that are almost impossible to break...it takes literally half a handful of triso pellets to fuel...5,000 homes...indefinitely." — Valor Atomics (California) and Aloe (Austin TX) are new entrants alongside the NuScale/Oklo/Kairos landscape; TRISO fuel's "poppy seed pellet" form factor and meltdown-proof chemistry are a recurring industry-distinguishing feature. Malmgren reports an Austin-area company built a working SMR prototype "with less than 300 people in less than 360 days."
  • pippa-malmgren in 2026-05-28-podcast-macro-voices-macrovoices-534-dr-pippa-malmgren-superpower-war (Genesis Mission — AI + classified national labs): "the announcement by the White House that all our national labs, so that's Los Alamos and Lawrence Livermore and the Argonne and all these amazing labs...they've all been totally classified. None of the data has ever been let out the door...We're going to lift the lid off all of them. We're going to run AI over all the data and we're going to connect the dots within the labs and across the labs and we're betting there's a whole bunch of world shaking technologies that are going to come out of that and massively accelerate the solutions, like new forms of energy." — if Genesis Mission proceeds, it could accelerate nuclear/fusion breakthroughs faster than consensus models assume, potentially pulling forward the SMR timeline.
  • From 2026-05-28-autoresearch-us-industrial-policy-tariffs-may-28: Big Beautiful Bill creates structural nuclear vs. solar/wind policy asymmetry. Per enacted law (signed July 4, 2025):
    • Nuclear, geothermal, hydro, batteries: Full 100% clean electricity credits if project begins construction by end of 2033; phase-out 2033–2036
    • Solar/wind: Projects not breaking ground by ~July 4, 2026 are on phase-down — 60% credit (2026), 20% (2027), 0% by 2028
    • Net effect: Nuclear has a 7+ year credit advantage over solar/wind. For AI datacenter PPAs, nuclear is the only carbon-free baseload that is both technically viable (24/7) AND policy-advantaged. The policy has taken solar/wind off the table as long-term hyperscaler power sources. See first-solar-ira-domestic-content-advantage for the First Solar beneficiary angle within the solar residual.

Names and exposures

TickerExposureConviction
CCJCameco — largest Western uranium miner; uranium supply for reactor fuelMedium-high — uranium fuel demand is a derivative of reactor restarts and new builds
CEGConstellation Energy — largest US nuclear operator; already signed Microsoft PPAMedium (downgraded from High) — PJM delay to 2031 is a new material risk; Calpine dilution; -13% YTD
BWXTBWX Technologies — nuclear components and small modular reactor (SMR) servicesMedium — longer-dated optionality on SMR buildout; less near-term catalyst
UECUranium Energy Corp — analyst consensus target ~$18 (vs ~$13.65, ~+32% upside); US-domiciled uranium producerLow-medium — pure uranium price leverage; no reactor/fuel infrastructure moat

Why nuclear and not natural gas

  • Natural gas provides baseload but does not satisfy hyperscaler ESG commitments (Scope 2 emissions).
  • Nuclear provides baseload and qualifies as carbon-free, meeting both reliability and emissions targets simultaneously.
  • SMRs (small modular reactors) are a 2030+ theme; the near-term catalyst is existing large reactor restarts and PPAs (Three Mile Island, Constellation's fleet).

Contradictions / tensions

  • ⚠ CONTRADICTION (see calibrate): CEG/TMI PJM interconnection delay to 2031. Prior wiki entry and prevailing thesis framing assumed Crane Clean Energy Center (Three Mile Island Unit 1 restart, 835 MW) online target 2027. New: Constellation Energy Q1 2026 commentary flagged PJM interconnection process delay — potential slip to 2031. If the interconnection is delayed, CEG's $1.6B restart investment and Microsoft's 20-year PPA both face material disruption. CEG stock already -13% YTD as of May 2026 on guidance miss + Calpine dilution; PJM delay adds a second re-rating risk. This is a direct CONTRADICTION of prior belief; log via /calibrate. From 2026-05-29-autoresearch-nuclear-ai-datacenter-ppa-smr-may-29.
  • CEG's valuation already reflects the nuclear PPA premium — forward multiples are not cheap. CEG -11.6% post Q1 2026 earnings on guidance miss ($11.50 vs $11.60 consensus EPS) and Calpine dilution concerns; the stock is re-rating from "nuclear pure-play premium" to "gas acquirer with nuclear assets" given the $21.8B Calpine deal. From 2026-05-20-autoresearch-nuclear-uranium-meta-ppa-ccj-ceg-smr-may-2026.
  • CCJ is exposed to uranium price cycles; spot at $85.25/lb (May 18) has pulled back from $89/lb recent reference — YoY still +19.6%. The India $2.6B deal locks in multi-year volume at market-related prices, reducing spot exposure for 2027–2035. From 2026-05-20-autoresearch-nuclear-uranium-meta-ppa-ccj-ceg-smr-may-2026.
  • Regulatory risk: nuclear permitting and licensing in the US is slow even for restarts; any grid-interconnect delay slips the timeline.
  • SMR timelines remain pre-commercial: NuScale is the only SMR with full NRC design certification (May 2025); Oklo/Kairos/TerraPower/X-energy all pre-commercial. No SMR online in US before 2028–2030. Meta's Vistra 2026 deliveries are from existing large reactors, not SMRs. From 2026-05-20-autoresearch-nuclear-uranium-meta-ppa-ccj-ceg-smr-may-2026.
  • NextEra (NEE) is getting nuclear restart contracts but is not in suggested-tickers. Duane Arnold (Google/NEE, 615 MW, 2029) is the second brownfield restart after TMI. If the brownfield-restart pattern multiplies, NEE is a larger, more liquid nuclear play than CEG with a diversified utility book. From 2026-05-26-autoresearch-nuclear-ai-datacenter-ppa-smr-update.
  • NextEra/Dominion merger ($67B all-stock, May 18, 2026) creates a dominant AI DC power infrastructure utility (Dominion mid-Atlantic/Southeast grid + NextEra clean-energy expertise). If FERC-approved, this combined entity is the preferred hyperscaler PPA counterparty in the highest-density AI DC geography. See nextera-dominion-merger-ai-power-consolidation for hypothesis. From 2026-05-29-autoresearch-nuclear-ai-datacenter-ppa-smr-may-29.
  • If battery storage improves dramatically (solid-state, grid-scale iron-air), the nuclear moat on baseload narrows.

What would weaken this thesis

  • Grid-scale storage breakthrough that makes solar + storage cost-competitive with nuclear 24/7 reliability
  • AI training workloads migrate to inference-at-edge models that don't require large centralized baseload
  • Regulatory reversal on nuclear (unlikely in current US political environment but possible)

Valuation snapshot

Last refreshed 2026-05-30 (source: twelvedata, 2026-05-29 close). Fwd P/E / mkt cap from paid tier — not available on free tier. PJM capacity auction: $28.92→$329.17/MW-day (10×); data centers 63% of increase. Uranium spot $85.20/lb; Sprott Physical Uranium Trust resumed purchases. UEC Burke Hollow ISR commenced production May 2026. Per 2026-05-30-autoresearch-energy-critical-minerals-uranium-helium-copper-nuclear.

TickerPrice52w rangeMkt capFwd P/EYTDWhat's priced in (one line)
CCJ$112.70$57–$135———-3.6% from $116.93 entry; uranium demand cycle partially priced; 14-yr term price high + Meta 7.8GW PPA catalyst = additional upside not priced
CEG$287.75$243–$413———-30% from 52w high; TMI PPA + PJM $329/MW-day priced; new hyperscaler PPAs and additional nuclear restart optionality remain unpriced
BWXT$195.88$123–$242———-5.3% from $206.83 entry; defense nuclear services priced; SMR optionality underpriced vs 2030+ catalyst path
UEC$13.77$5.63–$20.34———Burke Hollow production start (world's newest ISR, May 2026) partially priced; lowest-cost US uranium producer thesis intact

Forward-looking outcomes (12-month)

Bull case — Meta 7.8 GW nuclear PPA triggers industry-wide hyperscaler procurement wave; AWS, Google announce comparable arrangements; uranium spot returns above $95/lb: CCJ uranium volume contracts accelerate; CEG announces 2–3 additional hyperscaler PPAs; BWXT wins advanced microreactor design contracts. Implied price: CCJ +40–50%; CEG +25–30%; BWXT +20–25%. Cited: 2026-05-18-autoresearch-macro-bucket-scan-may-18-2026, 2026-05-01-all-in-podcast-openai-misses-targets-codex-vs-claude-elon-vs.

Base case — Microsoft TMI and Meta PPAs hold; AWS/Google sign 1–2 more nuclear PPAs in 2026; uranium spot stays $85–95; US DC electricity demand grows toward 580 TWh by 2028 as projected: CEG compounds on existing PPA book; CCJ volume grows on long-term contracts; BWXT steady on defense work. Implied price: CCJ +15–20%; CEG +10–20%; BWXT +5–15%. Cited: 2026-05-18-autoresearch-macro-bucket-scan-may-18-2026, 2026-05-11-autoresearch-macro-energy-critical-minerals-may-2026.

Bear case — Grid-scale battery storage proves sufficient for hyperscaler reliability at lower cost; uranium spot softens below $80; no further hyperscaler PPA announcements beyond current signings: CEG multiples compress from premium baseload to utility comps; CCJ tracks uranium spot lower. Implied price: CCJ -15–20%; CEG -15–20%; BWXT -10%. Cited: nuclear-baseload-for-ai-data-centers.md (contradictions section).

Currently undervalued vs base case? CEG: No — wiki notes valuation already reflects PPA premium; further hyperscaler PPAs = unpriced upside. CCJ: Marginal — term price 14-yr high signals market anticipation; further upside requires sustained >$95/lb. BWXT: Yes on SMR optionality if 12-month catalyst materializes.

Catalyst path:

  • Meta 7.8 GW nuclear PPA concrete volumes and timeline (Q2/Q3 2026): flows to CCJ fuel demand
  • Additional hyperscaler nuclear PPA announcements (AWS, Google): confirms broader-than-Microsoft adoption
  • US NRC advanced reactor licensing milestones for CEG/BWXT products

Related

Referenced by
Mechanisms
AI capex sprint → power-supply gap → grid-component + materials bottleneck → nuclear/copper/transformer beneficiary cascadeAI DC load growth → PJM capacity auction 10× → nuclear baseload structural premium → CEG/CCJ/UEC benefitTerrestrial power-flat → chip-output exponential → orbital DC arbitrage
Concepts
Copper supercycle: AI data centers and energy buildoutUS critical mineral independence
Questions
Hypothesis: Big Beautiful Bill IRA Changes → First Solar as Domestic Solar WinnerHypothesis: NEE/Dominion merger Virginia SCC block risk → NEE priced above standalone clean-energy valueIs the $830B CSP CapEx cycle a one-year spike or a sustained multi-year structural level?
Entities
Cameco (CCJ)Cheniere EnergyConstellation Energy (CEG)David FriedbergElon MuskGE VernovaPippa MalmgrenUranium Energy Corp
Sources
Autoresearch: Macro bucket scan — May 18 2026