Stance: Cautiously Bullish (rising probability of the industry's end-game realization, tangible cash flow improvement; but thematic overvaluation in A-share names has yet to be unwound — industry views must be decoupled from individual-stock views) | Research Window: 24 months | 2026-11-15
One-Line View: The fusion industry chain is transitioning from "scientific engineering" into an "order-driven realization phase" — government funding forms the cash flow floor, the first 10,000-km-class tape long-term order confirms a step-change in procurement volumes, and milestones are densely packed over the next 24 months (SPARC first plasma and Q>1, BEST completion by end-2027); but 100% of current cash flow is still R&D/demonstration reactor capex with zero electricity-generation revenue, the profit pool is locked within the national research institute system and materials leaders' core businesses, and pure-play fusion names listed in A-shares are broadly loss-making — industry momentum is real, individual-stock profits are illusory.
The subject of this study is the controllable nuclear fusion industry chain, focusing on four segments: magnets (REBCO high-temperature superconducting tape and magnet assemblies, NbTi/Nb3Sn low-temperature superconducting wire), vacuum vessels and in-vessel components (first wall, divertor, tritium breeding blankets), specialty materials (tungsten, superconducting tape, low-activation structural steel, large forgings), and EPC/main-machine integration.
One-line industry map: end buyers (ITER's seven parties, Chinese state platforms [Fusion New Energy/BEST, Hefei Institutes of Physical Science/CRAFT, CNNC, China Fusion Energy], UK Fusion Energy, private developers CFS/Helion/Tokamak Energy, etc.) → subsystem integration (magnets, vacuum vessel, power supply, cryogenics, heating) → materials layer (superconducting wire/tape, tungsten, forged steel).
Two caliber red lines run through this report: ① the industry currently has no "electricity-generation operation market" — supplier revenue and orders come 100% from facility construction capex (referred to here as "research cash flow"); ② the boundary for "electricity-generation cash flow" (electricity sales revenue and recurring consumables purchases) arrives at the earliest in the early 2030s. The tokamak/spherical tokamak is the mainstream route (ITER, SPARC, BEST, STEP), coexisting with stellarators (W7-X, Proxima), FRC (Helion, TAE), inertial confinement (NIF/IFE), magnetized target fusion (General Fusion), and fusion-fission hybrid reactors (Xinghuo One) — this report's segment analysis (magnets/vacuum/materials) applies across all routes; route competition mainly affects the reactor-owner side.
| Facility/Sponsor | Funding Status | Amount and Milestones | Nature |
|---|---|---|---|
| ITER (seven parties) | Appropriated + top-up under negotiation | New baseline adds EUR 4.2 billion; research operation 2034 / D-T 2039 | Realized |
| China · BEST (Fusion New Energy) | Funded | ~USD 2.1 billion; completion by end-2027, D-T electricity demonstration around 2030 (Xinhua/Science Net) | Realized |
| China · China Fusion Energy | Funded | ~USD 2.1 billion (CNNC + CNPC + Shanghai state capital), preparing HL-4 | Realized |
| China · Xinghuo One | Raised ≥USD 1 billion | Total investment ~USD 2.76 billion (media caliber), target 2030-31 | Partially realized |
| China · CRAFT | Built | USD 570 million | Realized |
| US · CFS | Cumulative funding ~USD 4 billion | SPARC ~80% complete (2026-07); ARC 400MW PJM interconnection application filed | Realized |
| US · Helion et al. | Ongoing funding | Helion raised USD 465 million in 2026-06; Orion under construction | Realized |
| UK · STEP | Endorsed | >GBP 2.5 billion five-year framework; STEP new phase GBP 1.3 billion through 2029 | Realized |
| Japan | 4 industrial demonstration projects selected 2026-09 | ~USD 400 million (reported caliber, unverified) | Realized |
| China · CFEDR / EU-DEMO / K-DEMO | Planned | Not appropriated | Planned |
Taking 2025 supplier equipment procurement of ~USD 3.8 billion/year as the baseline, cumulative visible orders over 2026-2030: low 14.9 / mid 21.6 / high 29.2 billion USD (scenario probabilities 28%/52%/20%, probability-weighted expectation ~USD 21.2 billion), with annual order flow climbing to USD 4.5-4.8 billion by 2030.
Explicit assumptions disclosed (the bridge's weakness): government procurement conversion rate set at 75% (anchor: ITER in-kind contributions represent about 90% of project value) + 50% ITER top-up realization + 13% private financing-to-procurement conversion (anchor: FIA's 2025 private-side supplier spending of USD 538 million / financing volume). Sensitivity: if procurement conversion falls to 60%, the mid scenario drops to ~USD 18.4 billion. Top-down cross-check: extrapolating F4E's annualized data gives a deviation of +12.6% (within ±15%, pass), but publicly verifiable large orders remain thin (see I4.2); this bridge is an "assumption-driven magnitude judgment" rather than an aggregation of the order book.
Institutional outlook (IAEA WFO 2026 citing EUROfusion's "Fusion at scale" scenarios): deployment rates of 12%/yr (slow) and 23%/yr (fast), fusion at 2%-20% of global electricity by 2100, implying a fusion fleet of ~400-700 GWe × USD 3,600-6,000/kW = cumulative plant capex of USD 1.44-4.0 trillion. But the pace warrants sobriety: per deployment-rate extrapolation, the fleet will be only ~6-16 GWe in 2050 with an annual equipment market of USD 2.7-14.8 billion/year — the true scale-up comes after the 2060s; before 2050, the "generation side" contributes billions, not tens of billions, to suppliers. Equipment value chain split (assumed range, no authoritative first-hand breakdown, unverified): magnets 25-35%, vacuum vessel and in-vessel components 15-25%, power supply and heating 10-20%, tritium cycle + conventional island 15-25%, civil works and installation 10-15%.
The essence of current industry chain demand is 100% R&D/demonstration reactor construction capex; electricity-generation operating demand is zero. Microsoft-Helion (50MW/2028), Google-CFS (200MW), and Eni-CFS (over USD 1 billion equity + offtake) are long-term power pre-sales — prices all undisclosed, not constituting current equipment orders (Eni's equity injection could convert into procurement within 1-2 years; it is current funding but not directly equal to supplier orders). Quantified criteria for the procurement inflection (confirm when ≥2 are met): ① orders for the 2nd/3rd unit of the same design from the same developer; ② 10,000-km-class framework agreements — the first confirmed signal appeared in CFS-Fujikura in 2026-09; ③ consumables (divertor/first wall) repeat purchases >10% of orders; ④ supplier capacity doubling with >3 customers (the REBCO layer is close to meeting this). Inflection window: 2027-2032; before then, the core variable for supplier revenue is government budget timing, not the electricity market.
| Segment | TRL | Key Validation Events |
|---|---|---|
| Nb3Sn/NbTi low-temperature superconducting magnets | 8 | All 18 ITER TF coils on site by 2026-10; Western Superconducting completed all BEST wire deliveries (bond prospectus, 2026-05) |
| REBCO high-field magnets | 7 | CFS TFMC full-performance 20T test in 2021; 8 of 18 SPARC TF magnets installed (2026-10) |
| Vacuum vessel (large forged-welded sections) | 7 | 9/9 ITER sectors on site, 6/9 installed; Heifei Forging delivered BEST's first sector in 2026-04, 8 sectors in batch production |
| First wall/divertor/tungsten | 6 | EAST 100M°C × 1,066 seconds (2026-01); SWIP won ITER first-wall batch production contract (2025-12) |
| Heating and current drive | 8 | ITER gyrotron batch production deliveries (second unit installed 2026-01); 24 EU procurement contracts in execution |
| Power supply systems | 6 | Mature at experimental-facility level (EAST/BEST); no track record at GW-class plant power supply |
| Diagnostics | 7 | Full ITER diagnostic package in manufacture; reactor-grade tritium/neutron environment unvalidated |
| Tritium plant and fuel cycle | 4 | All six ITER tritium plant subsystems reached final design, downstream plant contracts entering from 2026-09; commercial closed loop has never operated |
| Tritium breeding blanket (TBM) | 3 | No full-machine integration on any device; ITER selected 4 systems for testing in 2026-09 |
Interpretation: Magnets and the vacuum vessel (two of the four segments in this report) have crossed the "first-reactor ready" threshold — this is precisely the technical foundation for order cash flow in 2026-2030; while tritium breeding/tritium plant at TRL 3-4 is exactly the physical boundary of "electricity-generation cash flow" — segment heat is extremely uneven.
Current global nameplate capacity is ~8,700 km/yr: Shanghai Superconductor 4,000 km (~46%, global No. 1; Phase III topped out 2026-10 with planned capacity of 20,000 km/yr at full ramp), Liangjing Technology's first line 3,000 km (built 2026-04, ramp unverified), FFJ 1,000 km, SuperOx ~500 km, SuperPower ~240 km (company announcements/official websites). A single commercial reactor consumes 15,000-20,000 km of tape (CFS-Fujikura 10,000-km order, Proxima Alpha ~20,000 km) — current capacity supports only parallel construction of ~0.6 commercial reactors/year; planned 2030 capacity of ~29,000 km/yr ≈ 1.9 reactors/year. Low-temperature wire: Western Superconducting has annual capacity of 2,000+ tonnes, near-monopoly domestically (one of three major global players). Large forgings: Hefei Forging (8 vacuum vessel sectors) and China National Erzhong (8 TF coil cases) have achieved BEST first-reactor-grade batch production, but the "more than one reactor-set per year" replication cadence is untested by commercial orders.
The capacity arms race has begun (Fujikura expanding to secure CFS orders, Shanghai Superconductor Phase III, Proxima planning with Lower Saxony Europe's first fusion-grade tape plant at ~EUR 140 million by 2029) — after concentrated tape capacity release in 2027-28, the industry's main contradiction will shift from "insufficient capacity" to "volume up, price down": a cost-reduction positive for reactor owners, a real margin pressure for tape suppliers (the early structure of a solar-style price war is already visible: prices have broken below RMB 100/meter while expansion continues).
① Tritium supply and self-sufficiency: Global civilian inventory ~25 kg ≈ 0.45 GW·full-power years; CANDU produces ~2 kg/year, while a 1GW D-T reactor requires 55.6 kg/year; TBR>1 has never been validated in any device — "no breeding, no industry" (OSTI/Abdou white paper); ② neutron irradiation materials qualification: IFMIF-DONES will not provide a 14MeV high-flux source until the early 2030s; until then, low-activation steel/tungsten lifetime data can only be extrapolated — gating licensing and insurance; ③ REBCO capacity and the price-decline slope; ④ helium: the 2026-03 Qatar disruption doubled spot prices; cryogenic systems/leak detection/tritium extraction all depend on it; ⑤ batch replication of large forgings; ⑥ absence of standards and licensing frameworks.
Completed: NIF scientific ignition Q>1 (2022-12) and the 8.6MJ record (2025-04); JET D-T finale 69MJ; CFS TFMC 20T; all ITER magnets + 9/9 vacuum vessel sectors on site (2026-10, core components of main-machine assembly complete); EAST 100M°C × 1,066 seconds (2026-01); BEST four main-machine module sections positioned + first TF magnet test completed (2026-10). The next 24 months: see the I7 catalyst calendar.
SPARC Q>1 (2027-28 window) is the industry's single largest catalyst: SPARC is ~80% complete overall (company caliber, 2026-07), with 8 of 18 TF magnets installed; but first plasma has slipped twice (2025→2026→2027; both Nuclear News and IAEA WFO 2026 cite 2027), with a realistic prior that Q>1 could slip to 2028-29. BEST is the China-side anchor: upon completion by end-2027 it will proceed directly to D-T burning experiments (target fusion power 20-200MW), with electricity demonstration around 2030 — completion is not the endpoint; the subsequent burning/electricity demonstrations will continue generating blanket, tritium system, and power supply orders. ITER's significance lies in the supply chain, not the schedule: component delivery ≠ progress (new baseline first plasma 2034, D-T 2039), but its ongoing annual assessments at "USD 2-3 billion scale" are the stable floor of the global supply chain.
| Order | Buyer → Supplier | Amount | Status |
|---|---|---|---|
| ITER first-wall batch production (China's largest procurement package) | ITER China → Southwestern Institute of Physics | Undisclosed | Signed (2025-12) |
| ITER magnet power supply Phase II turnkey | ITER China → ASIPP consortium | Undisclosed | Signed (2025-11) |
| BEST vacuum vessel, 8 sectors | Fusion New Energy → Hefei Forging | RMB 209 million | Delivered |
| BEST TF magnet / tritium breeding / cryogenics three major packages | Fusion New Energy → Hefei Institutes/CAEP-system consortium | RMB 888.5M / 1,140M / 728M | Signed (broker weekly caliber, pending original award verification) |
| STEP construction partners Phase I | UK Fusion Energy → ILIOS consortium | GBP 197 million | Signed (2026-03) |
| STEP magnet partnership / digital twin | UKFE → Tokamak Energy / Dassault | GBP 70M / 30M | Signed (2026-04) |
| REBCO tape long-term agreement (largest single order ever) | CFS → Fujikura | >10,000 km; amount undisclosed (~USD 140-420 million at current prices) | Signed (2026-09-30) |
| Tritium fuel cycle, four major systems | ASIPP → tendering | Combined budget RMB 1,345.4 million | In tender (published 2025-11) |
| European fusion-grade tape plant | Proxima → Lower Saxony, Germany (MoU) | ~EUR 140 million | Intent |
The public order book totals ~USD 960 million to 1.24 billion, equivalent to only 25-33% of annual order flow (~USD 3.8 billion) — the bulk is hidden in undisclosed-amount ITER/CRAFT/private procurement. This means the market is effectively pricing an "imagined order book," the structural root of thematic volatility; it also means improved order visibility (original award announcements, company disclosures) is itself a catalyst.
① Materials/component suppliers (where 90%+ of industry cash flow currently lands: 2025 private-side supplier spending of USD 538 million + Chinese government procurement packages); ② subsystem assembly (TF magnets/vacuum vessel/cryogenics: institute-system consortia + a few listed/leading players); ③ facility owners/operators (pure capex, zero revenue, sustained by fiscal funding and private capital); ④ long-term power PPA layer (cash flow = 0; nominal boundary 2028 [Helion-Microsoft, likely delayed], base case early 2030s [ARC Fall Line commissioning]); ⑤ tritium fuel and consumables repeat purchase layer (≈0; divertor/first wall replaced every 2-5 years, forming only after 2035 — this is the true recurring component of "electricity-generation cash flow").
Three-tier mechanisms: custom R&D components = public tender + budget cap + comprehensive evaluation (research institutes lead the assembly, listed companies subcontract processing segments, one-time pricing); tape long-term agreements = commercial negotiation of multi-year frameworks (CFS-Fujikura price undisclosed); PPAs = bilateral negotiation (Microsoft/Google/Eni prices all undisclosed, anchored to nuclear LCOE: Lazard 2025-06 basis USD 142-222/MWh). Value split of a single 400MW commercial reactor (midpoint USD 2.4 billion): magnets ~USD 720M, vacuum vessel + in-vessel components ~USD 480M, power supplies and heating ~USD 360M, civil construction EPC ~USD 290M, others (tritium cycle/diagnostics etc.) ~USD 550M; of which tape: 15,000-20,000 km × USD 14-42/m = USD 210-840M (~68% of magnet cost) — tape is the single material with the highest value density per reactor.
REBCO tape price base-case path (pricing framework = incentive price floor [USD 10-50/kA·m] + capacity ramp dilution + order pacing): 2026Q4-2027 RMB 100-300/m → 2028 RMB 80-200/m → 2029-30 RMB 50-150/m (midpoint 200→140→100); industry annual order flow: USD 3.9-4.1bn → 4.3bn → 4.5-4.8bn. Consistent with the base-case scenario (cumulative USD 21.6bn).
Reactor integrators/operators (national buyers ≤5, all state-owned platforms: Fusion New Energy, Hefei Institutes of Physical Science, ITER·Fusion Center, CNNC group, Southwestern Institute of Physics; overseas: ITER seven parties, UKFE, private developers) → subsystem suppliers (magnets, vacuum vessel, power supplies, cryogenics, in-vessel components) → materials layer (Western Superconducting near-monopoly in LTS wire domestically, Shanghai Superconductor #1 globally in HTS tape, tungsten duopoly of AT&M/Plansee, large forgings via China National Erzhong/Hefei Forge/Oriental Electric). Competition revolves around "certification track record + relationship binding" rather than price wars, but buyer concentration overwhelmingly favors buyers — national owners use public tendering, with the institute system and major-shareholder affiliates prioritized in contract awards.
The current profit pool is not in A-share fusion concepts. BEST mega-packages are retained in-house by the institute system (TF magnets RMB 888.5M → Hefei Institutes consortium, tritium breeding RMB 1.14bn → CAEP Materials Institute, cryogenics RMB 728M → Hefei Institutes pre-selected winner; all amounts per broker weekly reports, to be verified); listed companies mainly take processing segments like wire and forgings. Pure A-share fusion names are generally loss-making: Guoguang Electric 2025 revenue RMB 276M (-48%), loss RMB 99M, with continued losses of RMB 93M in the first three quarters of 2026 plus RMB 79M in impairments; Hefei Forge lost RMB 249M in 2025; AccoTest losses in consecutive years. Profitable players earn everything from other core businesses: Western Superconducting relies on titanium alloys (first three quarters of 2026 net profit -59.6%, core business declining), Yongding on submarine cables/harnesses (net profit +126%), AT&M on tungsten/molybdenum (revenue +40%). Migration direction: as BEST is delivered (2027) and Xinghuo/CFEDR break ground, profits migrate to leaders with delivery track records — roughly in order: magnet wire (Western Superconducting) > divertor in-vessel components (Guoguang Electric, if orders land) > forgings (Hefei Forge) > power supplies (Yingjie/AccoTest); but "in-house institute circulation + bulk-procurement price pressure" is a dual ceiling on listed companies' revenue elasticity.
A living specimen of crushing buyer bargaining power is Guoguang Electric: due to ITER design changes and the prime contractor not placing orders, 2025 revenue halved and turned to losses — a single concentrated customer can directly absorb the entire industry upturn. Scarcity segments enjoy temporary strength: high-field REBCO tape has only 4-6 mass producers globally (CFS forced to sign a 10,000-km long-term contract to lock capacity), large vacuum vessel forgings, divertor tungsten-copper components (premium during the certification-moat period; once multi-source certification completes, tendering erases the premium).
| Company | Segment | Fusion Orders & Delivery | Core Business & Profitability (as of 2026-11) | One-Line Take |
|---|---|---|---|---|
| Western Superconducting 688122 | LTS wire (near-monopoly) | ITER conductor packages supplied for years; BEST wire fully delivered (first-hand from prospectus); ITER-grade Nb3Sn package ~RMB 55M | 2025 revenue RMB 5.226bn / net profit RMB 839M; 2026 first three quarters net profit -59.6% (titanium drag); market cap RMB 27.5bn (2% 1-yr percentile), PB 3.92 (4% 10-yr percentile) | The most certain beneficiary, but fusion elasticity diluted by core-business cycle |
| 联创光电 600363 | HTS magnets (off-balance-sheet equity stake) | Xinghuo One magnet supplier is Lianchuang Superconductor, ~40%-owned (2023 revenue only RMB 75M) | 2025 revenue RMB 3.34bn / net profit RMB 480M | High thematic content but profits off-balance-sheet, lowest tier of reporting delivery |
| Guoguang Electric 688776 | Divertor / first wall / tritium-handling pumps & valves | ITER/EAST historical supply confirmed; BEST divertor order ~RMB 820M is second-hand, to be verified | 2025 revenue RMB 276M (-48%) / loss RMB 99M; market cap RMB 4.69bn (4% 10-yr percentile), PS ~17x | The rarest positioning and the most painful delivery, a living specimen of downstream concentration risk |
| Hefei Forge 603011 | Vacuum vessel forgings | BEST vacuum vessel RMB 209M delivered (~10% of 2025 revenue); chairman concurrently chairs Fusion New Energy | 2025 revenue RMB 2.12bn / loss RMB 249M | A shadow-shareholder beneficiary bound to the owner, dragged down by core-business losses |
| Yongding 600105 | REBCO tape (Eastern Superconductor) | Tape revenue not separately disclosed; no announced BEST tape order | 2026 first three quarters net profit RMB 743M (+126%, submarine cable-driven); 2025 share price +310.7% | Typical concept-delivery mismatch: improvement comes from core business |
| AT&M 000969 | Tungsten materials / divertor materials | Multiple batches of tungsten-copper parts to WEST and ITER; fusion revenue share undisclosed | 2026 first three quarters revenue +40%, net profit RMB 372M | Core-business boom carries it; fusion is a free option |
| Yingjie Electric 300820 | Fusion power supplies | Cumulative historical orders tens of millions; company itself says "no significant volume" (2026-10) | 2025 revenue RMB 1.5bn / net profit RMB 223M | Authentic but elasticity not yet triggered; valuation rests on semiconductor power supply business |
| AccoTest 688719 | Power supply integration | No orders retrieved | 2025 revenue RMB 903M / loss RMB 45M | Pure valuation-driven thematic beta |
| Lanshi Heavy Industry 603169 | Large vessels / ex-core structures | No fusion order announcements found | Not retrieved | Watchlist, not a delivery pool |
| (unlisted) Shanghai Superconductor | REBCO tape | Global #1 in nominal capacity (~46%) | Yield ~90%, cost in global #1 percentile | The most valuable unlisted asset in the chain |
Proxy basis (no standalone SW fusion index): SW Other Power Equipment II PE 41.24x / 37% percentile since July 2014, PB 2.50x / 47% percentile (actual calculation as of 2026-11-13) — the proxy index sits at a neutral-to-low historical level. At the single-stock level: after the 2025 thematic bull run (Yongding +310.7%, Hefei Forge +231.8%, Guoguang +96.4%, Kaiyuan Securities as of 2025-12-19), 2026 saw deep drawdowns (Guoguang Electric market cap down 53% from ~RMB 10bn; Western Superconducting at 2% 1-yr market-cap percentile) — the bubble of valuation outrunning revenue has been substantially squeezed but not fully cleared: loss-making Guoguang Electric still trades at PS ~17x. The current setup is "upcycle × mid-low valuation," fundamentally different from 2025's "upcycle × extreme valuation"; but this does not change the return-structure judgment that "profits sit in core businesses, fusion is an option."
Return quality takeaway: this business is currently "hard to make money in" overall — real cash sits in materials leaders' core businesses (Western Superconducting superconducting products, AT&M tungsten/molybdenum) and inside the institute system; fusion revenue itself is "loss-making or negligible" for nearly all A-share names. Improvement in return quality follows the order of order batching (wire → forgings → in-vessel components → power supplies), and must pass the gate of "bulk-procurement price pressure."
| Jurisdiction | Framework | Status |
|---|---|---|
| US | NRC: fusion regulated under Part 30 byproduct material (2023 ruling codified); 2026-02-26 proposed rule + draft licensing guidance; also proposed "standardized design approval + production certificate" for batch manufacturing | Proposed rule published; final rule timing undecided, with agreement-state jurisdiction dispute risk |
| UK | EA+HSE environmental/occupational safety framework (not licensed as fission reactors); EN-8 national policy statement in development; insurance with nuclear exclusions + decommissioning funding mechanism | Framework in place; EN-8 not finalized |
| China | MEE 2025-04 notice: fusion devices licensed as radiation safety facilities for nuclear technology use, three-tier classification, consolidated review pilot | In force (BEST pathway proven) |
| Japan | NRA treats tritium environmental release containment as top issue; regime not finalized | Under discussion |
| EU | First EU fusion strategy being finalized; ENSREG fusion working group established 2026-03; missing legal basis is the biggest gap | In progress |
| Russia/Germany/South Korea | Russian amendment effective early 2027; German radiation protection law amendment expected end-2026; Korea to propose framework direction end-2026 | In progress |
Judgment: global licensing converging toward "lighter than fission" is the main trend (US, China, UK — the three largest markets — have all offered executable pathways), but uncontrollable first-of-a-kind licensing timelines remain the biggest administrative risk to 2030s grid connection — especially the US final rule and state-level execution capacity, and the EU's legal basis.
US: FY2026 FES budget request USD 744.8M (US ITER share USD 200M); the Milestone-Based Fusion Development Program received a multi-year Congressional authorization of USD 415M, but actual annual appropriations fall far short (first tranche USD 46M + USD 107M for FIRE in 2025) — federal cash support for private pilot reactors still relies mainly on "financing endorsement + procurement commitments"; DOE 2026-06 roadmap targets deploying early plants mid-2030s. UK: over GBP 2.5bn over 5 years (STEP GBP 1.3bn through 2029). EU: Euratom EUR 549.4M + member-state matching ~EUR 450M. China: BEST/CRAFT are national mega-projects; the "15th Five-Year Plan" recommendations list "hydrogen and fusion energy" among future industries for forward deployment; China Fusion Energy Co. (CNNC + CNPC + Shanghai SOE capital) is the national industrialization platform. Japan: 2026-09 selected 4 industry-led demonstration projects. Net direction: supportive; two uncertainties: US FY2027 appropriations fight, and whether the EU's additional EUR 4.2bn for ITER lands.
China's 2025-04 export controls on 7 categories of medium/heavy rare earths including yttrium/dysprosium/terbium (yttrium is the main REBCO raw material), expanded in October with extraterritorial provisions, and in November a one-year suspension of some measures against the US — the suspension expiring around November 2026 is a key window for tape costs and the pace of overseas capacity expansion; directional impact: accelerated overseas tape localization (Fujikura expansion, German plant construction), short-term cost and lead-time pressure upward. Tritium is concentrated in Canada's CANDU system (UK-Canada tritium-cycle JV); helium concentrated in Qatar/US (spot doubled 2026-03).
Zero-carbon baseload attributes put fusion on the UK's clean-energy priority list and the US strategic technology list; waste is mainly low-activation, short-lived materials (NRC proposed rule specifically solicits input on waste classification); the UK insurance/decommissioning framework opens the door to institutional capital. ESG funding flows are broadly friendly, providing structural support for private financing.
| Scenario | Probability | Narrative | Implications for Orders |
|---|---|---|---|
| Bear | 28% | SPARC Q>1 slips to 2028-29; BEST delayed to 2028; private financing cools (2023 prior); ITER top-up stalled by EU budget; CFEDR not initiated | 2026-2030 cumulative orders USD 14.9-16.0bn; second A-share thematic washout (high-PS loss-makers pull back again) |
| Base | 52% | Milestones broadly on time (SPARC first plasma 2027, Q>1 2027-28; BEST complete end-2027); ITER top-up lands year by year; China tendering normalizes; STEP platform awards continuously | Cumulative orders USD 21.6bn (range 20.0-23.0); annual order flow USD 3.9→4.8bn; tape ASP down to RMB 50-150/m |
| Bull | 20% | Q>1 achieved triggers ARC Units 2-3 / Xinghuo / CFEDR initiation in 2027 + STEP main FID and construction in same period | Cumulative orders USD 27.2-31.2bn; tape/forging arms race, ASP resilience beats expectations |
| Time Window | Event | Directional Impact | Related Debate |
|---|---|---|---|
| 2026Q4 | China rare earth export control one-year suspension expires (extend/tighten/lift) | Tape costs and overseas expansion | D2 |
| End-2026 | German radiation protection amendment, Korean regulatory framework direction, NRC final rule (FIA expectation, undecided) | Licensing pathways | D1 |
| ~Feb 2027 | US FY2027 budget request (Milestone Program / ITER funding showdown) | Government cash flows | D2 |
| 2027H1 | EU fusion strategy release, UK EN-8, NRC final rule window | Licensing pathways | D1 |
| 2027 | SPARC first plasma (company guidance) | Master switch, stage one | D1 |
| 2027-2028 | SPARC DT Q>1 experimental window | The industry's single biggest catalyst | D1 |
| End-2027 | BEST completion (IAEA basis) | China-side anchor | D1 |
| 2027-2028 | CFS-Fujikura first tape deliveries; ARC Virginia main construction starts | Procurement inflection confirmed | D2 |
| 2027-2028 | ITER tritium plant first component installation; LIBRTI facility milestones | Tritium chain | D3 |
| 2028 | Helion-Microsoft PPA nominal delivery window (likely deferred) | Power-generation cash flow expectations | D1 |
| ~2030 | BEST DT power demonstration | Generation-side milestone | D1/D3 |
Bullish on the cycle ≠ worth buying: the lower-right quadrant positions — Guoguang Electric (rare positioning but single-customer risk already realized as halved revenue + widening losses, PS ~17x), Hefei Forge (orders genuinely delivered but core business bleeding, orders only ~10% of revenue), Yongding (fusion tape revenue undisclosed, results from submarine cables), 联创光电 (fusion assets held as off-balance-sheet stake), AccoTest (consecutive losses, pure valuation-driven), plus the "tape expansion wave" segment itself (highest cycle but volume up price down after 2027-28 capacity release) — share a common trait: industry boom is real, but the names' reported delivery is fake, a classic value-trap structure. In the upper-right quadrant, Western Superconducting/AT&M/Yingjie Electric also earn current profits from core businesses (titanium/tungsten-molybdenum/semiconductor power supplies); fusion revenue is a free option — the "core allocation" valuation basis should be core-business value + option elasticity, not discounting the fusion narrative into current profits.
Beneficiaries: ① Western Superconducting — near-exclusive domestic supplier of low-temperature superconducting wire; has actually supplied both ITER and BEST (BEST wire delivery is sourced from a primary prospectus); its titanium alloy main business provides an earnings floor, valuation sits at a 10-year low percentile, and it offers the most positive leverage when fusion volumes ramp up; ② AT&M (Antai Technology) — one of the duopoly players in tungsten materials/divertor materials; its main business is in its own upcycle, with fusion as a low-risk option; ③ Hefeng Forging & Intelligent (high-risk tier) — has delivered the BEST vacuum chamber and is deeply tied to the project owner; if Xinghuo/CFEDR break ground, its order continuity will be the strongest, but it needs validation that its main business stops the bleeding; ④ Yingjie Electric (watch tier) — a legitimate player in the power supply segment with profitable main business; wait for a "volume ramp" signal; ⑤ Unlisted: Shanghai Superconductor (world's largest tape production capacity, best cost percentile) and Fujikura (CFS multi-thousand-kilometer long-term contract) are the two highest-quality non-A-share assets in the value chain.
Stocks to Avoid: ① LIANCHUANG Optoelectronics — its core fusion assets are held as an outside equity stake, the lowest earnings realization on the listed company's statements; ② Aiko Saibo (ACPower) — consecutive losses plus no disclosed fusion orders, purely thematic beta; ③ Guoguang Electric — scarce positioning, but downstream concentration risk has already materialized (revenue halved, expanding losses, impairments), while PS ratio remains ~17x — overextended, not yet cleared (if the BEST divertor order is confirmed via the original announcement, it becomes the highest-elasticity reversal candidate, an event-driven bet rather than a core allocation); ④ Weakly-related concept stocks (cryogenics/capacitor/valve names riding the theme).
| Scenario | Biggest Beneficiaries | Biggest Losers |
|---|---|---|
| Bear (28%) | Western Superconducting, AT&M (main business floor + low valuation percentile, relatively defensive) | Guoguang Electric, Aiko Saibo (no main business buffer + high PS, second-round clearing) |
| Base (52%) | Western Superconducting (BEST final payment + wire for new facilities), Hefeng Forging & Intelligent (delivery-driven revenue recognition) | LIANCHUANG Optoelectronics (off-balance-sheet assets continue to fail to materialize) |
| Bull (20%) | Guoguang Electric (highest elasticity from divertor ramp), Western Superconducting (wire volume and price); unlisted Shanghai Superconductor/Fujikura benefit most | AT&M has relatively the lowest elasticity (underperforms but is unharmed) |
① Milestone slippage (a SPARC Q>1 delay would reprice financing and order pacing across the entire industry); ② Tritium self-sufficiency and materials qualification (the physics bottleneck means the power-generation cash flow inflection point may come later than the early 2030s); ③ Financing cycles (private capital already accounts for nearly half of funding flows; cooling in the primary market will transmit upstream to orders); ④ Closed-loop procurement by state research institutes and centralized purchasing price pressure (listed companies' revenue elasticity systematically lags the sector narrative); ⑤ Tape overcapacity (concentrated release in 2027-28; rising volume and falling prices erode materials segment gross margins); ⑥ Policy and geopolitics (US appropriation intensity, additional EU funding for ITER, the direction of rare-earth export controls upon expiry, NRC final-rule delays).
This is industry-level research. For stocks of interest (e.g., Western Superconducting, Guoguang Electric, Hefeng Forging & Intelligent), a separate single-stock research file can be established to obtain target prices and position recommendations.
Key Sources: IAEA World Fusion Outlook 2026; FIA 2026 Global Fusion Industry Report; F4E Global Fusion Energy Trends Observatory (2026-07); CFS press release (2026-09-30); NRC Federal Register proposed rule (2026-02-26); Ministry of Ecology and Environment notice (2025-04-02); UK Fusion Strategy 2026; Nature Energy (Tang et al., 2026); OSTI tritium supply chain white paper; Kaiyuan Securities Controlled Nuclear Fusion Industry Weekly (2025-12-21, BEST procurement package amounts are secondhand figures pending verification); Shanghai Superconductor website; Western Superconducting bond prospectus (2026-05).