AGI Investment Tracks — Energy & Grid
Report date: 2026-05-26
Track 2 of the agi-investment-tracks team.
Analyst lens: AGI arrival is the working premise, not a hedge.
1. Thesis
Once cognition becomes cheap, the binding constraint on AGI-scale compute and physical industry is electrons. The
US is staring at a step-change in load growth — Rystad, EPRI and the EIA's AEO 2026 all converge on
hyperscaler+EV demand adding roughly 250-400 TWh by 2030, on top of an otherwise flat grid. That is the
equivalent of adding California's entire annual consumption inside five years to a system whose interconnection
queue is already 5+ years long and whose transformers ship in 2.5-3 years.
The first-order trade (own existing nuclear baseload — CEG, VST, TLN) is already richly priced. The remaining
edge is in the second-order picks-and-shovels: gas turbines for bridge power, large power
transformers, HV switchgear and cable, transmission EPCs, HALEU enrichment, and uranium. These
companies have multi-year backlogs, real pricing power, and earnings still being upgraded — not just multiple
expansion. The contrarian add-on is geothermal (Ormat) and grid-scale storage (Fluence) where the AI-PPA narrative
just landed in early 2026 and has not yet fully run.
2. The Bottlenecks (quantitative)
2a. Power demand step-up
- US data-center load: ~200 TWh in 2024 → projected 600-1,000 TWh by 2030 (~8-12% of US grid). Rystad pegs combined data-center + EV growth at ~300 TWh by 2030.
- Reported delayed AI capacity in queue: at least 7 GW of US AI data-center capacity has slipped due to power+equipment delays per 2026 tracking reports.
- PJM's 2024-25 capacity auction cleared at $269.92/MW-day — ~9x prior cycle. ERCOT scarcity pricing has firmed similarly. The market is already telling you what's scarce.
2b. Equipment lead times (Wood Mackenzie, Q2 2025 reading)
| Equipment | Current lead time | Price change since 2019 | Demand change since 2019 |
|---|
| Large power transformers (LPT) | 128 weeks (~2.5 yrs) | +77% | +119% |
| Generator step-up units (GSU) | 144 weeks (~2.8 yrs) | +45% | +274% |
| Distribution transformers (large) | ~2+ yrs | up to +95% | +34% |
| HV switchgear | 44 weeks | MV +50%, breakers +47% | heavy |
| Heavy-duty gas turbines (GE Vernova HA-class) | Sold out through 2029, sold-out-through-2030 expected by end of 2026 | New reservations price > existing backlog | ~80 GW backlog |
2c. Nuclear fuel cycle
- Uranium spot: ~$70-75/lb U3O8 range in May 2026 (off the 2024 spike to $107 but well above $40 in 2022). Long-term contract price remains $80+.
- Enrichment (SWU): ~$170-190/SWU in early 2026; the West has lost ~30% of historical SWU capacity after the Russian HALEU ban (Centrus is the only US-owned operator).
- HALEU (advanced reactor fuel): Centrus is the only domestic producer, running at ~900 kg/year — a rounding error vs. the tons needed for SMR deployment.
3. Supply / Demand Gap Model
| Year | US data-center load (TWh) | New nameplate capacity needed (GW, 50% CF) | Plausible new build (GW) | Gap |
|---|
| 2024 | ~200 | baseline | — | — |
| 2027 | ~380 | ~41 GW incremental | ~25 GW (gas peakers + existing nuclear reactivations + a little solar+storage) | ~16 GW short |
| 2030 | ~700-1,000 | ~115-180 GW incremental | ~70 GW (gas, solar, wind, storage); SMRs immaterial <5 GW | ~45-110 GW short |
Estimate. Sources: Rystad, EPRI, EIA AEO 2026, gas-turbine OEM backlog disclosures. The gap is closed by demand
destruction (queue delays, hyperscaler siting overseas) and/or price.
4. Investable Public Companies
"Priced in" = how much of the AGI/data-center thesis is already in the multiple. High = stretched. Medium = partial. Low = still discounted vs the bottleneck it owns.
4a. Existing nuclear baseload (the "obvious trade")
| Ticker | Mkt cap | Valuation | Bottleneck owned | Risks | Priced in? |
|---|
| CEG Constellation |
~$95B (May 2026) |
Fwd P/E ~26x, EV/EBITDA ~13x |
Largest unregulated US nuclear fleet (~32 GW); Three Mile Island restart for Microsoft |
Already up ~3-4x from 2023; 2026 guide disappointed; political risk on co-located load tariffs (FERC TLN/AWS dispute) |
High |
| VST Vistra |
~$53B |
Fwd P/E ~18x, EV/EBITDA ~10x, ~7% FCF yield |
~39 GW fleet incl. 6.4 GW nuclear (Comanche Peak); ERCOT + PJM merchant exposure |
$19.6B net debt; coal exposure; capacity-price normalization risk |
Medium — still cheaper than CEG on every metric |
| TLN Talen |
~$15B |
EV/EBITDA ~12x post-Amazon expansion |
Susquehanna nuclear; 1.9 GW Amazon co-location PPA being restructured after FERC pushback |
Regulatory: FERC has signaled it won't bless behind-the-meter exemptions cleanly |
High |
| NEE NextEra |
~$160B |
Fwd P/E ~22x |
Largest US renewables + Florida regulated utility; reopening Duane Arnold nuclear plant |
Rate-base ROE caps the upside; exposure to IRA repeal noise |
Medium |
4b. Gas turbines & bridge power (sold-out OEMs)
| Ticker | Mkt cap | Valuation | Bottleneck owned | Risks | Priced in? |
|---|
| GEV GE Vernova |
~$165B |
EV/EBITDA mid-30s, Fwd P/E ~50x |
~80 GW gas turbine backlog into 2029; ~18 GW booked in Q4'25 alone; hyperscaler volume agreements in discussion through 2035; HVDC; synchronous condensers |
Multiple already prices in the buildout; cyclical at heart; capacity expansion (to 24 GW/yr by 2028) erodes future scarcity |
High — but pricing on new bookings still moving up |
| 7011.T Mitsubishi Heavy |
~¥6.0T (~$38B) |
Fwd P/E ~22x, much cheaper than GEV |
Doubling gas turbine output; gained data-center share vs Siemens Energy; recently raised guidance citing turbine orders |
Conglomerate (defense, ships, etc.) — turbine signal diluted; JP forex risk |
Medium — the cleanest "GEV at half the multiple" |
| ENR.DE Siemens Energy |
~€85B |
Fwd P/E ~30x |
Grid Technologies (transformers, HVDC, GIS); record backlog; turbines via Siemens Gamesa drag fading |
Has already 10x'd from 2023 bottom; Gamesa wind tail risk |
High |
4c. Transformers, switchgear, HV cable (the deepest shortage)
| Ticker | Mkt cap | Valuation | Bottleneck owned | Risks | Priced in? |
|---|
| ETN Eaton |
~$135B |
Fwd P/E ~30x, EV/EBITDA ~22x |
MV switchgear ($30M Omaha plant; $340M SC transformer plant — 2027 start); Electrical Americas backlog >$13B; "nearly 3,000 planned US data centers" as target |
Premium multiple; switchgear capacity catching up by 2027-28 |
High |
| HUBB Hubbell |
~$22B |
Fwd P/E ~22x |
Utility solutions: meters, grid components, distribution-class transformers |
Less direct AI tailwind narrative; smaller than ETN |
Medium |
| PRY.MI Prysmian |
~€20B |
Fwd P/E ~18x |
Global #1 HV submarine cable for offshore wind interconnects + grid HVDC; orderbook into 2030s; Encore Wire acquisition expanded US grid cable |
Offshore wind project delays in US |
Medium |
| NEX.PA Nexans |
~€6B |
Fwd P/E ~15x |
HV subsea cable (Tyrrhenian Link record); smaller, cheaper Prysmian peer |
More European concentration |
Low-Medium |
| 009540.KS HD Korea Shipbuilding (parent of HD Hyundai Electric) |
~₩30T |
SOTP-cheap |
HD Hyundai Electric is the leading Korean LPT exporter to the US; +30% US production targeted by 2026; Alabama plant |
Conglomerate discount; Korea governance |
Low — direct Hyundai Electric (267260.KS) trades hot; parent is the dilution |
| 298040.KS Hyosung Heavy |
~₩3T |
Fwd P/E ~10x |
Korean LPT/STATCOM/HVDC equipment; major US transformer importer |
KRW; concentration in US data-center demand |
Low |
4d. Transmission EPCs (the builders)
| Ticker | Mkt cap | Valuation | Bottleneck owned | Risks | Priced in? |
|---|
| PWR Quanta Services |
~$60B |
Fwd P/E ~30x, EV/EBITDA ~19x |
$48.5B record backlog (Q1'26), Electric Power $40.1B segment; $2.4T TAM through 2030 per mgmt; doubling adj-EPS by 2030 target |
Skilled-labor constraint is now the binding input; share already 5x off 2022 lows |
Medium-High |
| MYRG MYR Group |
~$3B |
Fwd P/E ~22x |
Pure-play T&D EPC, smaller cousin of PWR |
Less scale; project execution drag in 2024 |
Low-Medium |
| MTZ MasTec |
~$13B |
Fwd P/E ~20x |
Power delivery + clean energy + pipelines; mixed exposure |
Pipeline + telecom drag |
Medium |
| EME EMCOR |
~$25B |
Fwd P/E ~22x |
Mechanical/electrical building services + data-center mission-critical |
Construction labor |
Medium |
4e. SMR / advanced nuclear (mostly priced for hope)
| Ticker | Mkt cap | Valuation | Bottleneck owned | Risks | Priced in? |
|---|
| BWXT BWX Technologies |
~$13B |
Fwd P/E ~30x |
$8.65B backlog; sole US supplier of naval reactors; SMR component manufacturing; medical isotopes |
Defense-budget dependent; revenue growth single-digits |
Medium — the only profitable nuclear-supply name |
| LEU Centrus Energy |
~$5B |
Fwd P/E ~35x |
Only US-owned uranium enricher; sole domestic HALEU producer (Piketon, OH); DOE contract extended; positioned to scale if Russia ban makes Western SWU permanently scarce |
Single-site execution; depends on Congressional appropriations to scale beyond pilot |
Medium |
| OKLO Oklo |
~$1.8B (multiple sources; check live) |
Pre-revenue |
Aurora reactor at INL; Sam Altman-chaired (pre-IPO); LOIs with hyperscalers |
NRC COLA timeline is years; LOIs ≠ binding; dilution risk |
High — story stock |
| SMR NuScale |
~$8B |
Pre-revenue |
NRC-approved SMR design; ENTRA1 industrial partnership; no commercial deployment yet |
UAMPS project cancelled (2023); first commercial unit unlikely before 2030; massive dilution |
High |
| NNE Nano Nuclear |
~$1B |
Pre-revenue |
Microreactor concepts; HALEU fuel transport ambitions |
Earlier-stage than even Oklo; pure narrative |
Very High |
4f. Solar+storage / batteries
| Ticker | Mkt cap | Valuation | Bottleneck owned | Risks | Priced in? |
|---|
| FSLR First Solar |
~$22B |
Fwd P/E ~10x |
Only major US-domiciled thin-film maker; Section 45X IRA credits; backlog through 2027 at locked prices |
IRA repeal risk under current admin posture; China LFP+poly cost competition |
Low — beaten down on tariff/IRA fears |
| NXT Nextracker |
~$9B |
Fwd P/E ~14x |
US solar tracker market leader; >$4B backlog |
Utility-scale solar slowdown if IRA bites |
Low-Medium |
| FLNC Fluence Energy |
~$3B |
Path to profitability 2026 |
$5.6B order backlog (Q2 FY26); two hyperscaler supply agreements announced May 2026; AES + Siemens parentage; battery storage is the cheapest way to firm AI-data-center renewables PPAs |
China cell-cost competition; lumpy bookings |
Low — stock just popped 40% on the hyperscaler news but still well off 2023 highs |
| ENPH Enphase / SEDG SolarEdge |
small/mid |
Distressed |
Residential — not your AGI trade |
N/A for this thesis |
Skip |
4g. Geothermal & other firm low-carbon
| Ticker | Mkt cap | Valuation | Bottleneck owned | Risks | Priced in? |
|---|
| ORA Ormat |
~$5B |
Fwd P/E ~30x, EV/EBITDA ~14x |
Largest pure-play public geothermal IPP; just inked a 150 MW PPA tied to Google via NV Energy (Feb 2026); the only firm-renewable proxy public investors can own (Fervo is private) |
Slow project cycle; small absolute size |
Low-Medium — Google deal not yet a multi-bagger catalyst |
5. Pre-IPO / Private to Watch
- Fervo Energy — enhanced geothermal; Google, Microsoft, Shell are customers/investors. The most credible firm-clean-power scale-up. Likely 2027 IPO candidate.
- Commonwealth Fusion Systems (CFS) — SPARC tokamak target net-Q in 2027; Eni, Google, Bill Gates backed. Multiple-year out from commercial relevance even on best case.
- Helion Energy — Microsoft PPA for 50 MW by 2028. Aggressive timeline; almost certainly slips.
- TAE, Tokamak Energy, Marvel Fusion — long-tail fusion bets.
- Form Energy — iron-air battery for 100-hr storage; Series F in 2024; serves utility load-shifting.
- X-energy — high-temp gas SMR; Amazon backing; Dow industrial partnership.
- Urenco — privately held by UK/Dutch/German govts; ~30% of Western SWU. If governments ever sell, this is THE enrichment trade.
- Westinghouse — owned by Cameco (49%) + Brookfield Renewable (51%). Public route in is via CCJ and BEPC.
6. Commodity / Physical Plays
| Vehicle | Exposure | Why it's a bottleneck | Notes |
|---|
| SPUT Sprott Physical Uranium Trust (U.UN/U.U on TSX) | Physical U3O8 | Direct exposure to spot uranium; trust can absorb supply when it issues units | Trades at small premium/discount to NAV; the cleanest non-equity uranium bet |
| URA Global X Uranium ETF | Basket: CCJ ~22%, Sprott U trust ~10%, NXE, DNN, UEC, KAP, etc. | Liquid diversified basket | Top-heavy in CCJ; some non-pure-play names |
| URNJ Sprott Junior Uranium Miners | Junior miners (NXE, DNN, UEC, EU, UROY) | Higher beta to spot | Volatile; ride spot, not value |
| CCJ Cameco | World's #2 producer + 49% of Westinghouse | Long-cycle production discipline; contract book repricing at $80+ floor | Most "investable" uranium equity; Q1 2026 maintained guidance |
| NXE NexGen Energy | Arrow project (Sask., Canada) — undeveloped, highest-grade undeveloped uranium | ~30 Mlb/yr future production with permitting nearly complete | Development risk; pure leverage to permit + price |
| Henry Hub natgas futures / EQT, AR, CTRA | US natgas producers | Gas turbines burn the marginal molecule for AI peakers; LNG+power competition tightens market | Cyclical; weather-sensitive — buy on weakness, not strength |
| Copper (FCX, COPX ETF) | Electrification metals | Transformers, cables, motors all need copper; see Materials track for full thesis | Cross-references materials track — don't double-allocate |
7. People / Talent Concentration
Worth noting because if you can't hire the people, you can't build the kit — irrespective of capex.
- Nuclear engineers: US graduates ~600/yr; aging workforce. BWXT, CEG, Westinghouse, Bechtel are the absorbers. Talent moat for Cameco/Westinghouse and CEG specifically.
- Transformer-design engineers: Severe shortage in US after 30 years of offshoring. Hitachi Energy and Eaton are paying premiums. Hyundai/Hyosung have intact bench in Korea — strategic edge.
- NRC licensing attorneys / regulatory experts: Pinch point for SMR commercialization. Most senior people are at Pillsbury, Morgan Lewis, and inside DOE — limited supply.
- Linemen: IBEW says ~10,000 lineman shortfall by 2030. Helps PWR/MYRG margins (labor-cost passthrough), hurts utilities.
- Quanta has internally invested in lineman training schools — a structural moat that's easy to miss in the financials.
8. Top Picks Ranked
#1 — VST Vistra (HIGH conviction).
Best risk/reward inside the obvious-trade bucket. Nuclear fleet + Texas gas + 7% FCF yield + investment-grade upgrade in March 2026. ~18x P/E vs CEG's 26x for similar AGI-PPA optionality. The market's preference for CEG is narrative, not numbers.
#2 — 7011.T Mitsubishi Heavy Industries (HIGH conviction).
Same gas-turbine super-cycle as GEV, at half the multiple, with rising guidance from data-center turbine demand. The Japan/conglomerate discount is doing real work here. Cleanest way to own the gas-peaker bottleneck without paying GEV's 50x.
#3 — FLNC Fluence Energy (MEDIUM-HIGH conviction).
Hyperscalers just signaled in May 2026 that grid-scale storage is part of their PPA stack, not optional. Backlog has doubled. Stock has run 40% but is still well off 2023 peaks. Path to profitability now visible. Asymmetric because it's the cheapest way to firm the renewables share of the AGI energy mix.
#4 — LEU Centrus Energy (MEDIUM conviction).
Only US-owned enricher. Only domestic HALEU producer. Permanent loss of Russian SWU + tightened FEOC rules = structural pricing power. Risk is execution and reliance on Congressional appropriations to scale. Position-size accordingly — this is a 2-4% allocation, not 10%.
#5 — PWR Quanta Services (MEDIUM conviction).
Owns the labor-and-execution bottleneck for actually building transmission. $48.5B record backlog; raised 2026 guidance. Expensive but defensible. Add on pullbacks.
Honorable mention: ORA Ormat (only public geothermal, has Google deal); 298040.KS Hyosung Heavy (cheap Korean transformer pure-play if you can buy Korea); SPUT as the physical-uranium core holding. FSLR is a contrarian buy if you believe IRA survives — beaten down enough that downside is bounded.
What I am NOT recommending and why:
- CEG at $308 / 26x — narrative is set, numbers just got cut. The TMI restart for Microsoft is already in the multiple.
- OKLO / NNE / SMR — pre-revenue with multi-year NRC timelines. Lottery tickets, not investments. If you must, size as 0.5%.
- GEV at 50x forward — the bottleneck is real but the price is silly. Wait for a 30% drawdown or buy 7011.T instead.
- ENPH / SEDG — residential solar has no AGI thesis. Skip.
9. What Would Change My Mind
- Hyperscaler capex pullback. If any one of MSFT/META/AMZN/GOOG cuts AI capex guidance by >15%, sell the entire stack down hard before the others react. PJM auctions would re-rate first.
- FERC kills behind-the-meter co-location across the board. Would destroy TLN's premium and dent CEG. Already partly in motion — watch the FERC ELCC ruling expected mid-2026.
- Permitting accelerates. If NEPA reform under the current admin actually halves transmission timelines, the GEV/transformer/PWR scarcity premium compresses.
- China LFP+turbine exports flood global markets. Compresses FLNC, NXT margins. Less impact on GEV (export controls + after-sales installed base).
- Fusion net-energy demonstration. If CFS hits net-Q materially ahead of 2027, terminal value of every fission name resets. Low probability but tail-fat.
- Russia rejoins Western fuel cycle. Would kill LEU's pricing power overnight. Currently almost zero probability under any administration.
- Recursive-self-improvement-driven materials science breakthrough. If AGI ends up designing room-temp superconductors or radically cheaper batteries inside the investment horizon, the entire transmission-bottleneck thesis collapses. This is the AGI-pilled version of "watch what you're betting against."
10. Sources
All accessed 2026-05-26 unless noted. Live quotes/multiples are approximations from the date stamp on each source and should be verified before trading.
Prepared by: energy-grid analyst | agi-investment-tracks team | 2026-05-26
This document is research output, not investment advice. Position sizes and entries are the investor's call.