The $2.7 Billion Receipt
How the DOE Wrote the AI Energy Put — And Why the Market Priced It Backwards
On January 5, 2026, the US Department of Energy issued $2.7 billion in fixed-price task orders for HALEU enrichment — the fuel form required by advanced reactors and small modular reactors. Four counterparties: American Centrifuge Operating ($900M + $170M options), Orano Federal Services ($900M), General Matter ($900M), and Global Laser Enrichment ($28.5M). That is the AI energy policy put, written down and dated.
Of the four awardees, only Centrus (LEU, ~$4B) trades freely on a US exchange. ASP Isotopes (ASPI, ~$640M) holds the only definitive HALEU supply contract with TerraPower — Centrus holds a July 2023 MOU. The equity market is pricing visibility, not contracts. Sections 3–8 walk the same inversion across grid, compute substrates, cooling, and minerals.
Why This Matters Now
- The receipt is dated and dollar-denominated. $2.7B in fixed-price task orders, four counterparties, a single day. The DOE bought redundancy across the HALEU supply chain — not throughput from any one supplier.
- NDAA §854 binds January 1, 2027. Eight months from publication, Chinese magnets are out of US defense procurement by statute. MP Materials, Lynas, Energy Fuels, and downstream magnet integrators face a hard cliff — not a guideline.
- The grid layer hasn't received its task order yet. Wood Mackenzie reports 128-week LPT lead times and 143-week GSU lead times against a 30% US power-transformer supply deficit projected for 2025. Hammond Power, MYR Group, and Preformed Line Products sit on this constraint.
- X-energy IPO'd at $23 on April 23, 2026 and closed +27% on day one. The market reads the put once it is dated. Five small-caps in this piece sit in front of the same recognition arc.
Section 1 — The ReceiptFree
On January 5, 2026, the US Department of Energy issued $2.7 billion in fixed-price task orders for HALEU enrichment — High-Assay Low-Enriched Uranium (uranium enriched to 5–20% U-235, the fuel form required by advanced reactors and small modular reactors, or SMRs). The recipients were: American Centrifuge Operating (a Centrus subsidiary at the Piketon, Ohio enrichment facility) at $900 million base plus up to $170 million in options, a $1.07 billion ceiling; Orano Federal Services at $900 million; General Matter at $900 million; and Global Laser Enrichment — the Silex/Cameco joint venture — at $28.5 million.[1]
That is the AI energy policy put, written down and dated.
The connection runs like this. Microsoft, Amazon, Google, and Meta have announced roughly 15 GW of nuclear-backed datacenter power purchase agreements (PPAs) since September 2024 — Microsoft-Constellation (Three Mile Island restart, 835 MW), Amazon-Talen (Susquehanna, ~2 GW), Amazon-X-energy (5 GW SMR target), Meta’s 6.6 GW package, Google-Kairos (50 MW initial framework). Those agreements aren’t coal or gas — they require advanced reactors and SMRs to actually get built. Advanced reactors run on HALEU. There is no HALEU without a functioning domestic enrichment supply chain. The DOE just funded that supply chain’s redundancy four ways, in fixed-price task orders, in a single day.
January 5 wasn’t a one-off. The Russian uranium ban (May 2024) cut off the cheapest enrichment alternative by statute. The HALEU Availability Program funded the bridge. Defense Production Act (DPA) Title III invocations have been applied across critical materials. NDAA §854 — binding January 1, 2027 — strips Chinese magnets from defense procurement. The DOE task orders are the line item with the cleanest paper trail: a date, a dollar figure, four counterparties.
Each policy layer reduces the optionality of inaction. The fuel supply chain is being capitalized. The question this raises is narrow and specific: if the put is real and dollar-denominated, why is the equity market pricing it asymmetrically?
General Matter received $900M without operating a commercial facility. That tells you the DOE prioritized counterparty diversification over proven throughput. The put backstops the supply chain, not any single supplier.
Section 2 — One Name Everyone Owns, Three They Don’t, One They’ve InvertedPro
Of the four DOE awardees, only one trades freely on a US exchange.
Centrus (via American Centrifuge Operating, ticker LEU on NASDAQ, approximately $4 billion market capitalization) is the name the market owns.
It is in every nuclear energy ETF. It is the default answer to “how do I own HALEU?” It received the same base award as Orano and General Matter. At $4B, LEU trades at roughly 6x the next-cleanest public HALEU exposure — and the comparison set is thin because Orano is state-adjacent French, URENCO is private, Westinghouse is private. LEU is the liquid name, and it’s priced like the only name.
The other three DOE awardees are not available in the same form. Orano Federal Services is the US subsidiary of Orano SA, a French state-adjacent entity — accessible only indirectly and with significant dilution through any public French energy complex. General Matter is a pre-commercial enrichment startup; it received $900 million in fixed-price task orders before operating a commercial facility. The DOE wasn’t buying throughput. It was buying redundancy. Global Laser Enrichment (GLE) offers the most accessible second path: it is a joint venture between Silex Systems and Cameco. Cameco (CCJ, approximately $50.5 billion market capitalization) is liquid and widely owned, but GLE represents a fraction of Cameco’s business, and GLE received $28.5 million — roughly 1% of the total task order pool. The HALEU exposure through CCJ is real but thin.
ASP Isotopes (ASPI, NASDAQ, approximately $640M market capitalization) signed a definitive 10-year HALEU supply agreement with TerraPower — via its HALEU subsidiary Quantum Leap Energy — on May 19, 2025, covering deliveries from 2028 through 2037 sourced from its Pelindaba facility in South Africa.[2] The agreement covers the first core of TerraPower’s Natrium reactor at the Kemmerer, Wyoming site. It is definitive, not a memorandum of understanding (MOU). It is subject to permitting by South Africa’s National Nuclear Regulator (NNR) — that conditionality is material and should be priced as risk — but it is a binding commercial agreement with scheduled delivery windows.
Centrus’s relationship with TerraPower is documented as a July 2023 MOU.[3] No definitive HALEU supply contract between Centrus and TerraPower has been identified through April 2026.
There’s a fair counter-argument: Centrus has running cascades at Piketon; ASPI is still building. The execution gap is real. The contract gap runs the other direction — and the equity market is paying for execution it can see while ignoring contracts it hasn’t read.
$4B versus $640M is not a contract-weighted distribution. It’s a visibility-weighted one. NNR permitting is a real haircut. It does not close a 6x gap.
Here is the math a hostile reader should work through. If LEU at $4B trades at fair value — reflecting commercial certainty, operating enrichment cascades, and full DOE backstop — and ASPI at $640M trades at a 6x discount, then the gap implies a market-assigned joint probability for the ASPI contract-completing path. Call it the combined probability of (NNR permits granted) × (TerraPower Natrium reaches commercial ops by 2031). At 50% NNR approval × 60% Natrium execution, combined probability is 30%. Even at 30%, ASPI’s expected contract value per dollar invested looks nothing like 1/6th of LEU’s. To justify the 6x gap, the market needs to be implying a combined probability below 15–20%. That is the bet. The NNR approval for a facility with an established operating history at Pelindaba, for a technology licensed from SILEX Systems’ aerodynamic separation process, is not a coin flip. TerraPower has a 2031 first-power target, Bill Gates committed capital, and a Wyoming site with state regulatory support. Sophisticated readers can disagree with that probability estimate — and the disagreement is the trade.
This is one inversion. Sections 3–7 walk the rest: turbines, transformers, magnets, switchgear, the grid interconnect itself. Each one has a Centrus and an ASPI. The market owns the Centrus.
Section 3 — The Demand-Side ReceiptPro
The DOE put has a counterpart. While the supply side was being written into task orders and HALEU enrichment contracts, Microsoft, Amazon, Google, and Meta were writing the demand side into PPAs. The deals are dated, the megawatts are contracted, and they define exactly which utilities and reactor developers re-rate — and when.
The deal book is public. Microsoft and Constellation Energy agreed in September 2024 to restart Three Mile Island’s Unit 1 as the Crane Clean Energy Center: 835 MW, targeting 2027, the largest nuclear PPA in US history by reported MW at signing.[9] Constellation ($114B) re-rated immediately. Amazon and Talen Energy closed a 1,920 MW deal from the Susquehanna nuclear station, with FERC blessing arriving in 2025;[10] Talen ($16.7B) moved with it. Google signed a framework with Kairos Power and the Tennessee Valley Authority in August 2025 — Hermes 2, 50 MW initial, 500 MW target by 2035, the first US Generation IV reactor PPA.[11] Kairos is private. Google also announced a deal with NextEra in October 2025 to restart the Duane Arnold plant in Iowa at 615 MW, targeting around 2029;[12] NextEra ($186B) is large-cap and mostly priced for other reasons.
Then there’s the Meta transaction. On January 9, 2026, Meta announced a 6.6 GW package: 2,609 MW from Vistra across Perry, Davis-Besse, and Beaver Valley; 1,121 MW from Constellation’s Clinton plant starting 2027; 1.2 GW from Oklo’s Aurora reactor at Pike County, Ohio; and a prepayment for TerraPower’s first two Natrium units.[8] That’s four counterparties, two of which are publicly traded (Vistra, Constellation), one of which is a listed pre-revenue developer (Oklo, ~$11.6B), and one of which is private.
The asymmetry is the same one Section 2 described in the fuel layer. The market owns Constellation and Vistra. They’re priced. The market thinly owns Oklo — still pre-revenue, no reactor operating yet, but the Meta contract is as close to a demand-side policy put as a pre-revenue developer gets.
X-energy presents the clearest case study in how quickly the put gets priced once it’s visible. X-energy IPO’d on April 23, 2026 at $23 per share (upsized offering), opened at $30.11 on April 24, and settled near $29.20 — a +31% open / +27% day-one close.[6] The market understood the thesis on day one: X-energy’s TRISO-X subsidiary holds a Special Nuclear Material License (SNM-7007) from the NRC for the TX-1 HALEU fuel fabrication facility, granted February 13, 2026,[7] and the Amazon partnership targets up to 5 GW of SMR capacity by 2039. The put has now been priced into the equity — XE popped on day one because the market read the Amazon binding agreement and the TRISO-X license together, understood the demand-side certainty, and bid it up before most institutional allocations cleared. The Cascade Advanced Energy Facility is not yet under construction. Post-IPO, XE is scaffolding, not asymmetric edge.
The hyperscalers are not making bets on clean energy. They are placing dated, MW-denominated, counterparty-specific orders that force a fuel cycle, a reactor build, a turbine procurement, and a substation. The market is pricing the operators and ignoring the dependency tree those deals create.
Section 4 — The Grid Is Where the Put Hasn’t Landed YetPro
Every hyperscaler PPA in Section 3 requires physical interconnection. Power purchase agreement to delivered electrons requires a substation build, a large power transformer (LPT), a generator step-up unit (GSU), transmission and distribution (T&D) infrastructure, and an engineering, procurement, and construction (EPC) contractor to install it. The policy put has reached the fuel cycle. It is reaching the reactor developers. It has not yet reached this layer.
Wood Mackenzie’s Q2 2025 transformer market analysis reported 128-week lead times for LPTs and 143-week lead times for GSUs, with a projected 30% US power-transformer supply deficit for 2025.[4] A hyperscaler signing a PPA today does not get steel in the ground for more than two and a half years. The Infrastructure Investment and Jobs Act (IIJA) transmission provisions and Inflation Reduction Act (IRA) grid spending help at the margin. There is no $2.7B fixed-price task order for transformers dated and signed. Not yet.
The market doesn’t own Hammond Power Solutions (HPS-A.TO, ~$2.3B USD, listed on the Toronto Stock Exchange). Hammond describes itself as the largest manufacturer of dry-type transformers in North America. In its Q3 2025 earnings release dated October 23, 2025, the company disclosed that “orders received after the close of the quarter were valued at 53% of the closing backlog.”[5] That’s an order-velocity figure on a $2.3B TSX-listed company’s earnings call — datacenter-driven acceleration showing up in the print. The market also thinly owns MYR Group (MYRG, ~$4.95B), the pure-play T&D EPC contractor that reported $973M in Q1 2026 revenue, up 17% year-over-year, and Preformed Line Products (PLPC, ~$1.36B), a small-cap T&D hardware manufacturer.
The market does own the obvious large-cap names. GE Vernova (GEV, ~$230B) reported $2.4B in datacenter-related equipment orders in Q1 2026 alone — exceeding all of fiscal year 2025 — and completed its Prolec transformer acquisition in February 2026, adding a reported $5B backlog. Eaton (ETN, ~$164B), Quanta Services (PWR, ~$94B), and Hubbell (HUBB, ~$29B) round out the large-cap T&D basket. All of them are priced for general industrial demand at minimum. GEV is meaningfully priced for datacenter demand. The Buy America preference matters here, but it’s worth noting GEV operates inside an oligopoly — HD Hyundai Electric runs an LPT plant in Montgomery, Alabama, and Hitachi Energy and Siemens Energy maintain US LPT manufacturing — so the policy preference accrues across multiple producers, not just the one the market has bid up.
The inversion here is sharper than in the fuel layer. The fuel layer at least has a dated DOE receipt. The grid layer has Wood Mackenzie lead-time data, a specific backlog-velocity disclosure from Hammond, and a structural Buy America preference baked into IIJA. What it doesn’t have is the task order. That’s the remaining asymmetry — assets priced for yesterday’s industrial demand, demand-side PPAs that require physical grid buildout, and the policy put not yet written for the bottleneck between them.
Section 5 moves up the stack to compute substrates and cooling — where the same inversion pattern has, in most names, already closed.
Section 5 — Compute and Cooling Are Mostly Priced, Except in the SubstratesPro
The compute layer is not where the asymmetry lives anymore. NVDA at roughly $5 trillion, ASML at $517 billion, TSMC, AVGO, AMD — these trade as AI policy-put names already. The market has read the CHIPS Act, read the executive orders, and priced in the government’s revealed preference for not losing this race. Buying them at current valuations is buying the put after it has been written, recognized, and bid up by every long-only fund that needed AI exposure before year-end. Returns will be respectable. They will not be asymmetric.
The asymmetry is in two directions from the compute core: upstream into substrates, and downstream into cooling and minerals.
Upstream — substrates
AXT Inc. (AXTI, ~$4.88B market cap) manufactures indium phosphide (InP) substrates — the wafer-level foundation for the optical interconnect chips inside AI accelerators. AXTI is the sole US-domiciled merchant supplier of InP substrates. Coherent runs an internal InP fab, but captive production doesn’t count as a merchant alternative for procurement-restricted programs — defense CMMC (Cybersecurity Maturity Model Certification) contracts, CHIPS Act-funded fabs, and FedRAMP-cleared hyperscaler workloads all require US-domiciled sourcing. Sumitomo (JP), JX Metals (JP), and Freiberger (DE) are real global competitors. None of them are US-domiciled. That buyer-constrained substitutability — not just global market share, but the structural inability of procurement-restricted buyers to use foreign suppliers — is what drove AXTI from roughly $150M market cap in 2023 to over $1B in 2024, and to $4.88B today. Small-cap pure-play, procurement constraint, demand inflecting, multiple un-rerated — that’s the template. But AXTI’s $4.88B today is not the entry point. The trade now is finding the next one in adjacent layers, which is what Sections 6 and 7 are for.
One step downstream — memory IP
Rambus (RMBS, ~$17B). Memory interface IP royalties scale with HBM4 (High Bandwidth Memory 4th generation) volume. Every advanced AI accelerator bill-of-materials includes HBM; CHIPS Act allocations name it explicitly. RMBS is less asymmetric than AXTI was at $150M, but the put applies and the royalty stream is durable.
Cooling
Vertiv (VRT, ~$119B) is the liquid cooling pure-play and is mostly priced. Hyperscalers have been disclosing datacenter cooling capex explicitly enough that VRT’s re-rate through 2025 was orderly and well-telegraphed. Super Micro, Modine, and HPE are adjacent; none are purer than VRT on the cooling thesis.
This layer is structurally different from the others: the demand pull is thermodynamics, not policy. GPU racks at AI density dissipate too much heat for air cooling. That is not a regulatory judgment call. The policy put matters less here because the demand exists independent of government action — which also means the valuation reflects it more completely. VRT at $119B is scaffolding. Real exposure, low remaining edge.
Section 6 — Minerals: The Deepest Inversion, the Clearest ReceiptsPro
The mineral layer is where the policy put has been most explicitly receipted and least continuously priced. The pattern is consistent across names: re-rate sharply around discrete catalysts — DPA Title III awards, Chinese export restriction announcements, NDAA implementation dates — then partially mean-revert as if the structural change were temporary. It is not temporary. The re-rates are recognitions; the mean-reversions are buying opportunities.
Large-cap mineral exposure — Freeport-McMoRan (FCX), Southern Copper (SCCO), Albemarle (ALB) — trades on commodity cycles, not policy receipts. The edge is elsewhere.
MP Materials (MP, ~$6.1B)
NDAA §854 imposes a binding January 1, 2027 hard cliff on Chinese-sourced magnets in US defense procurement. MP is the Western Hemisphere’s most advanced integrated rare earth operator by reported throughput, per company filings, running from mine at Mountain Pass through separated oxide to magnet precursor. Its Fort Worth magnet facility is the first US-built commercial rare earth magnet plant in a generation. The catalyst is date-certain: January 1, 2027 is not a regulatory judgment call, it is a statutory deadline. Defense primes cannot source Chinese magnets after that date. MP and its allied-jurisdiction complement Lynas (LYC, ASX-listed) are the two names with operating infrastructure positioned to serve that demand. The market has partially priced NDAA §854; it has not fully priced the procurement shift that executes against that cliff starting in eight months.
Energy Fuels (UUUU, ~$4.95B)
This is the dual-put name in the table. White Mesa Mill in Utah is the only operating conventional uranium mill in the US, per Energy Fuels disclosures and Energy Information Administration (EIA) fuel cycle reporting — that is the nuclear fuel put from Layer 2. The same White Mesa facility achieved first US primary dysprosium qualification in December 2025 and produced the first US primary terbium in March 2026.[13] Dysprosium and terbium are the heavy rare earth elements (HREEs) that increase the heat tolerance of neodymium-iron-boron (NdFeB) magnets for EV motors and wind turbine generators. China controls an estimated 90%+ of global HREE processing. One facility, one operator, two policy puts — uranium on the fuel cycle, HREEs on NDAA §854 and DPA Title III. The market has not priced them simultaneously into a single small-cap equity. UUUU trades as if it’s either a uranium stock or a minerals stock depending on which catalyst is most recent.
5N Plus (VNP, TSX, ~$2.15B USD)
Compound semiconductor precursors: tellurium, bismuth, gallium. VNP refines the upstream materials that flow into wafer production for AXTI-type InP substrates, as well as cadmium telluride (CdTe) solar cells and other compound semiconductor applications. DPA Title III candidacy is the catalyst — a formal Title III determination would lock in US government offtake and potentially fund capacity expansion, replicating the HALEU task order structure one layer upstream from the wafer fabs. The CHIPS Act fab qualification pull is the second path: as InP and GaAs fabs qualify under CHIPS, their upstream precursor demand becomes US-domiciled preference-eligible.
The mineral equity inversion is structural, not cyclical. Chinese export restrictions on gallium (August 2023), germanium (August 2023), and graphite (October 2023) were not temporary bargaining chips. They were supply chain policy instruments. NDAA §854 is not an advisory — it is a procurement prohibition with a hard statutory date. DPA Title III invocations are capital commitments, not statements of intent. The re-rates around each of these events were real. The mean-reversions are the market treating policy receipts as if they were price markers rather than structural realignments.
Section 7 — One Factor, Twenty-Five Names, Four Counter-ScenariosPro
These 25 names share one factor: federal procurement and industrial policy treating AI infrastructure as something the US cannot afford to lose. That is not a thesis — it is observable government behavior, receipted across the federal procurement, defense, and energy regulatory apparatus: DOE (HALEU task orders), Congress (NDAA §854, IRA, CHIPS Act), Commerce (CHIPS Act allocations), Defense (DPA Title III), and the NRC (advanced reactor licensing). The investment question is where the market has least priced what the government has most committed to.
The answer, by layer: Layers 1 and 2 (compute and power-fuel) are mostly priced except in the small-cap substrate and fuel-cycle pure-plays. Layers 3 and 4 (grid and cooling) are partially priced; edge in smaller hardware and EPC names. Layer 5 (minerals) is episodically priced; the largest edge, the clearest policy-receipt replication.
Counter-scenarios
These are analyst judgment ranges, not model outputs.
Algorithmic efficiency breakthrough (15–20% probability) — distillation advances, sparse mixture-of-experts architectures, or sub-2nm process gains bend the AI compute demand curve sharply downward. The 15–20% range reflects the gap between current model cost-per-token trajectory (down roughly 3x annually over the past two years) and the sustained demand growth required to absorb the announced 15 GW PPA pipeline. If efficiency gains compound faster than demand growth, the fuel-grid-cooling stack remains structurally overprovisioned and the put loses most of its demand-pull. The policy backstop survives as written; the economic justification for building against it weakens materially.
AI capex cycle breaks (10–15% probability) — hyperscalers cut datacenter spend in response to disappointing model returns or investor pressure. The 10–15% range reflects that current hyperscaler capex commitments are contractually embedded in PPAs with 10–20 year terms and break fees, making a sharp pivot operationally costly even if sentiment shifts. The government commitment persists regardless, but the private demand that amplifies it collapses and the timeline extends past most equity holding periods.
Policy reversal (roughly 10% probability) — a future administration unwinds DPA invocations, removes IRA nuclear production tax credits, or defunds DOE loan programs. The 10% range reflects the bipartisan framing of AI competitiveness, the contractual duration of HALEU task orders (10 years, with treaty-level friction to unwind), and the political cost of reversing domestic energy investment in swing-state manufacturing districts. Not zero, but not a base case.
Recession or capital cost shock (10–15% probability) — sustained high real rates push 5–7 year SMR and nuclear restart projects below their economic hurdle rates even with policy support. The 10–15% range reflects that SMR economics are most sensitive to project finance cost of capital, not operating cost — a 200bp sustained rate increase over current levels would extend payback periods enough to defer final investment decisions on projects not yet under construction.
Aggregate: roughly 45–55% probability that the thesis encounters meaningful headwinds within 24 months. The put is real. It is not a guarantee.
Section 8 — Ship List25 Names
These 25 names share the AI policy put as their dominant factor. The five names highlighted in the first five rows are the asymmetric exposures — small-cap or recently listed, contract-backed, demand-inflecting, not yet fully re-rated. The rest are scaffolding: real beneficiaries, less edge. Catalysts are the events that close the gap between contract and market recognition.
| # | Ticker | Company | Mkt Cap | Layer | Catalyst | Edge |
|---|---|---|---|---|---|---|
| 1 | ASPI | ASP Isotopes | ~$640M | Fuel | NNR permit, TerraPower 2031 first power | High |
| 2 | HPS-A.TO | Hammond Power Solutions | ~$2.3B | Grid | Datacenter LPT order velocity (53% post-Q3) | High |
| 3 | MYRG | MYR Group | ~$4.95B | Grid | T&D EPC pure-play, 17% YoY revenue growth | High |
| 4 | PLPC | Preformed Line Products | ~$1.36B | Grid | T&D hardware, small-cap pure-play | High |
| 5 | VNP | 5N Plus (TSX) | ~$2.15B | Substrate | DPA Title III determination, CHIPS fab pull | High |
| 6 | LEU | Centrus Energy | ~$4B | Fuel | DOE task order recognition (priced) | Scaffolding |
| 7 | CCJ | Cameco | ~$50.5B | Fuel | Uranium cycle, GLE thin exposure | Scaffolding |
| 8 | UUUU | Energy Fuels | ~$4.95B | Fuel/Minerals | Dual-put: White Mesa uranium + Dy/Tb | Scaffolding-to-edge |
| 9 | XE | X-energy | ~$12B est | Fuel/Reactor | TRISO-X license, Amazon 5 GW SMR | Scaffolding (post-IPO) |
| 10 | OKLO | Oklo | ~$11.6B | Reactor | Meta Aurora 1.2 GW Pike County | Scaffolding |
| 11 | CEG | Constellation | ~$114B | Utility | TMI restart, Meta Clinton | Scaffolding |
| 12 | VST | Vistra | ~$60B | Utility | Meta 2,609 MW Perry/Davis-Besse/Beaver Valley | Scaffolding |
| 13 | TLN | Talen Energy | ~$16.7B | Utility | Amazon Susquehanna 1,920 MW | Scaffolding |
| 14 | NEE | NextEra | ~$186B | Utility | Google Duane Arnold 615 MW | Scaffolding |
| 15 | GEV | GE Vernova | ~$230B | Grid | $2.4B Q1 2026 datacenter orders, Prolec | Scaffolding |
| 16 | ETN | Eaton | ~$164B | Grid | Datacenter T&D | Scaffolding |
| 17 | PWR | Quanta Services | ~$94B | Grid EPC | Transmission build | Scaffolding |
| 18 | HUBB | Hubbell | ~$29B | Grid | T&D hardware | Scaffolding |
| 19 | AXTI | AXT Inc | ~$4.88B | Substrate | InP — already re-rated 30x | Scaffolding (template) |
| 20 | RMBS | Rambus | ~$17B | Memory IP | HBM4 royalty volume | Scaffolding |
| 21 | VRT | Vertiv | ~$119B | Cooling | Liquid cooling — priced | Scaffolding |
| 22 | NVDA | Nvidia | ~$5T | Compute | AI accelerator demand — priced | Scaffolding |
| 23 | MP | MP Materials | ~$6.1B | Minerals | NDAA §854 cliff Jan 1, 2027 | Scaffolding-to-edge |
| 24 | LYC.AX | Lynas | ASX | Minerals | NDAA §854 allied-jurisdiction | Scaffolding-to-edge |
| 25 | FCX | Freeport-McMoRan | ~$96B | Minerals | Copper cycle, not policy | Scaffolding |
The receipt was January 5. The window — between the put being written and being recognized — closes when the rest of the equity universe figures out which of these 25 are sitting on contracts the market hasn’t read yet.
Disclosures
Not investment advice. This piece is research and commentary. It is not a recommendation to buy, sell, or hold any security. Markets move; specific names mentioned can lose substantial value regardless of the structural thesis described. Readers should conduct their own due diligence and consult licensed advisors before making any investment decision. Past performance is not indicative of future results, and the policy environment described is subject to change without notice.
Position disclosure. ForcedAlpha and its principals may hold positions in one or more of the 25 names referenced in this piece. Specific positions, sizes, and entry points are not disclosed. ForcedAlpha does not receive compensation from any company referenced and has no banking, advisory, or paid promotional relationship with any of them.
Methodology link. The analytical framework underlying this piece — the supply chain knowledge graph, the procurement-constraint severity rubric, and the asymmetric-exposure scoring — is documented at /methodology/. Definitions of terms used here (HALEU, PPA, DPA Title III, NDAA §854, buyer-constrained substitutability) appear in that reference page.
Sources
- US Department of Energy, “DOE Awards $2.7 Billion in Fixed-Price Task Orders to Restore American Uranium Enrichment,” press release, January 5, 2026. energy.gov/articles/us-department-energy-awards-27-billion-restore-american-uranium-enrichment
- ASP Isotopes Inc., “ASP Isotopes Inc. Enters into Definitive Agreements with TerraPower including Loan Agreement for Construction of a HALEU Production Facility and Supply Agreements for HALEU,” press release, May 19, 2025. globenewswire.com (ASPI definitive TerraPower agreements, May 19 2025)
- Centrus Energy Corp., “TerraPower and Centrus Expand Efforts to Commercialize Domestic HALEU Production,” press release, July 17, 2023. centrusenergy.com/news/terrapower-and-centrus-expand-efforts-to-commercialize-domestic-haleu-production
- Wood Mackenzie, “Mind the Gap: US Power Transformer Lead Times” (Q2 2025 transformer market analysis), October 15, 2025; “Untangling the US Transformer Supply Chain Crisis” (press release), August 14, 2025. woodmac.com (US power transformer supply deficits 2025)
- Hammond Power Solutions Inc., Q3 2025 Earnings Release and Management Discussion, October 23, 2025. globenewswire.com (Hammond Power Q3 2025 results)
- X-Energy Inc., “X-energy Announces Pricing of Upsized Initial Public Offering,” press release, April 23, 2026; first trade April 24, 2026 on Nasdaq Global Select Market under ticker XE. Open and close prices per public market data.
- US Nuclear Regulatory Commission, Special Nuclear Material License SNM-7007 issued to TX-1 fuel fabrication facility (X-energy/TRISO-X subsidiary), February 13, 2026.
- Meta Platforms Inc., “Meta Announces Long-Term Nuclear Energy Agreements to Support American AI Leadership,” press release, January 9, 2026. about.fb.com/news/2026/01/meta-nuclear-energy-projects-power-american-ai-leadership
- Constellation Energy Corp., “Constellation Energy and Microsoft Sign 20-Year Power Purchase Agreement to Restart Three Mile Island,” press release, September 20, 2024.
- Talen Energy Corp., “Talen Energy Expands Nuclear Energy Relationship with Amazon,” press release, June 11, 2025. ir.talenenergy.com (Talen-Amazon nuclear expansion, June 2025)
- Google LLC, “Google’s First Advanced Nuclear Reactor Project with Kairos Power and Tennessee Valley Authority,” blog post, August 19, 2025. blog.google (Google-Kairos-TVA, Aug 2025)
- NextEra Energy Inc., “NextEra Energy and Google Announce Strategic Collaboration to Restart Duane Arnold Energy Center,” press release, October 27, 2025.
- Energy Fuels Inc., “Energy Fuels Achieves First U.S. Heavy Rare Earth Oxide Qualified for Use in Permanent Magnets,” press release, December 19, 2025; “Energy Fuels Begins First U.S. Primary Terbium Production at White Mesa Mill,” March 2026.