A supply-chain judgment built on 33 A-share and HK-listed PV companies. It answers one question: which segment of the PV value chain is truly supply-locked? The answer is more counterintuitive than instinct — none.
Across the six PV segments, not one has truly formed a “supply-locked” constraint. The few players, heavy capex, and high concentration we see are all “signals,” not “hard conclusions.”
① The six segments span 3–9 entities, a 3× max-to-min ratio with no single segment above half; ② the thin upstream is a scope effect (vertically integrated in-house, assets held by unlisted groups), not supply scarcity; ③ module concentration is an outcome of scale economics — lowest per-GW capex, month-level certification — so it never enters the barrier list.
The PV industry has long leaned on an intuition: few upstream players plus heavy capex means supply is stuck. It sounds right, but no one has tested it against a falsifiable standard. This report does exactly that — it turns “intuition” into a “verifiable conclusion” with a three-layer judgment framework.
The result is a negative finding: none of the six segments reaches the structural-constraint threshold. That is itself a valuable conclusion — it tells you the PV game is not “who blocks supply,” but “who survives the capacity cycle, technology roadmap, and channels.”
The distribution of listed entities — a quantifiable, reproducible structural fact.
A “localization rate” — the report claims no localization rate and does not equate “few entities” with “insufficient supply.”
Conclusions hold only for listed entities; unlisted majors and foundries are outside the sample.
| Segment | Main | STAR | ChiNext | HK |
|---|---|---|---|---|
| Polysilicon | 1 | 1 | 0 | 2 |
| Wafers | 4 | 0 | 0 | 0 |
| Cells | 2 | 0 | 1 | 0 |
| Modules | 4 | 3 | 1 | 0 |
| Inverters | 1 | 4 | 4 | 0 |
| Power plants | 4 | 0 | 0 | 1 |
Wafers sit entirely on the main board (mature, asset-heavy), inverters lean to STAR/ChiNext (growth, R&D-heavy), and 2 of 4 polysilicon firms are HK-listed — a capital-market character difference, not a competitiveness signal.
The PV value chain is relatively balanced — the largest segment (inverters, 9) and the smallest (cells, 3) differ by only 3×, with no single segment above half.
This report does not equate “few players, heavy capex, high concentration” with “supply locked down.” It tightens step by step through a three-layer judgment, and explicitly allows a negative finding — “no structural constraint identified” is itself a valid conclusion.
Few players, heavy assets, high concentration — observable “signals.” Prices are under pressure and losses are absorbed, but no substantive shakeout has been triggered.
Supply is materially locked — new capacity cannot enter and old capacity cannot exit — backed by independently verifiable barrier evidence such as cost-curve position or per-watt cost.
Even alternative technology routes and substitute products pose no real threat. This layer requires import-dependence, local-share, or technology-gap evidence.
The report’s result is “zero constraints” — all six segments stay at layer one. That negative finding is worth more than any “found a chokepoint” claim, because it corrects a widespread industry misjudgment.
Has the structural features of “few players, heavy assets, high concentration,” but they fall short of an entry barrier. Polysilicon, wafers, and cells land here.
High concentration is an outcome of scale economics; the lowest per-GW capex and month-level certification keep the entry barrier low. Modules land here.
No supply scarcity — either entities are dense, or assets sit inside unlisted groups. Inverters and power plants land here.
Six segments, checked one by one: none reaches the structural-constraint threshold. Three upstream segments are “not a barrier,” modules have a “low entry barrier,” and inverters & power plants have “ample supply.”
Inverters are the classic “low-capex, high-R&D” manufacturing segment: the lowest capital intensity (4.2%) with the highest R&D (6.7%). PV’s real competitive axes are the capacity cycle, technology roadmap, and certification channels — not capital or R&D barriers alone.
| Metric | Value | Scope |
|---|---|---|
| Polysilicon localization | 95.3% | capacity · CPIA |
| Wafer localization | 96.9% | output · CPIA |
| Polysilicon capacity CR10 | 79.8% | global · CPIA |
| Module shipment CR5 | 73.9% | 2024 · computed |
| Inverter shipment CR5 | 63.9% | 2025 · S&P Global |
Upstream localization already exceeds 95%, so a “chokepoint” is moot; module and inverter concentration are high but are the result of scale economics and consolidation, not irreplaceable barriers.
Polysilicon, wafers, and cells total just 11 entities (33.3%), but supply tightness is set by capex intensity, expansion cycles, and the capacity cycle — entity count is only a reference.
At 85–105k RMB/tonne and a 15–24-month expansion cycle, it is strongly cyclical with high sunk costs — whoever sits low on the cost curve survives, and the longer capacity idles, the harder it is to restart.
TOPCon holds 83%, and HJT/XBC equipment costs far more. The equipment-investment gap decides how much old capacity is written off on a switch, and who seizes the next technology window.
The lowest per-GW capex (0.8–1.4 bn) and 3–6-month certification make entry easy; what separates winners is manufacturing cost, expansion capability, and vertical-integration efficiency.
Chinese makers already hold a large global share (Sungrow 25.2%, Huawei 24.8%); the next fight is overseas grid certification, channels, and brand — not domestic capacity.
China’s cumulative grid-connected capacity is 1,272 GW, but most assets sit with power groups and unlisted entities — “listed count” is the wrong ruler; one must first unify the numerator across self-owned, rolling-development, and O&M capacity.
Entity identities were verified one by one; for some entities only summary-level disclosure is public, so capacity, shipment, and revenue figures are not word-level verified, and value-chain classification rests on disclosed facts.
PV auxiliaries/equipment, pure distributors, mixed wind-solar operators, and non-PV firms were excluded by explicit rules; unlisted majors and foundries are out of scope, so conclusions cannot be generalized.
The entity distribution reflects a cut-off date (2026-10-01); it shows the current state and makes no trend judgment.
Tongwei is classified under polysilicon and LONGi under modules, subject to revenue-mix review; inverters with storage are classified as inverters, with storage revenue unsplit.
Publicly identifiable entities whose main business belongs to the six PV segments, classified by value-chain position, covering polysilicon through power plants.
Capital formation is measured as fixed-asset cash over revenue, and R&D as R&D expense over revenue, splitting “high barrier” into two measurable axes.
Industry totals are set against the sample to judge whether “few entities” is real scarcity or a scope effect; local shares are public-report estimates.
Industry totals come from IEA PVPS, China’s NEA, CPIA, InfoLink, TrendForce, and MIIT; entity dimensions come from company periodic reports.
Few entities ≠ insufficient supply. The thin upstream stems from vertical integration and asset-holding structure, not a real capacity shortage.
High concentration ≠ an irreplaceable barrier. Module and inverter concentration is a scale-economics outcome, not “newcomers can’t enter.”
Listed-entity count ≠ localization rate. This report measures the distribution of listed entities and claims no localization rate.