Daqo New Energy

Phase A — Understand the business

Lens 1 · Company Overview

Daqo New Energy makes one thing: high-purity polysilicon — the refined silicon feedstock that downstream players pull into ingots, slice into wafers, and build into solar cells and modules. It is a pure-play commodity-chemical producer at the very top of the solar value chain, with one reportable segment (Polysilicon) and substantially all revenue and assets in China. It commenced commercial polysilicon production in July 2008.

Corporate structure (important, and NOT a VIE): DQ is Daqo New Energy Corp., a Cayman Islands holding company (NYSE: DQ; 1 ADS = 5 ordinary shares) with no substantive operations of its own. It owns its business through direct equity — principally a ~72.8% stake (as of March 31, 2026) in Xinjiang Daqo New Energy Co., Ltd., which is itself separately listed on the Shanghai Stock Exchange STAR Market (raised ~RMB6.07B in its July 2021 IPO). This is a crucial distinction from the Alibaba/PDD-style VIE: there is no contractual-control fiction here, the foreign holder owns real equity. The cost is a ~27% minority-interest leakage — roughly a quarter of every dollar of Xinjiang Daqo's profit (or loss) belongs to the A-share public, not to DQ's ADR holders.

Products / customers / contracts. Output is solar-grade polysilicon (plus a small 1,000 MT semiconductor-grade line in trial production since May 2024). Customers are China's large wafer/module makers. Contracts are framework agreements that bind volume but price at "prevailing market prices when specific sales orders are placed" — i.e. volume commitments, spot-linked pricing, which is exactly why revenue is so violent. There is no take-or-pay price protection. Customer concentration is high and rising: top-3 customers were 64.4% of revenue (2023), 53.8% (2024), 63.5% (2025); the single largest customer was 38.9% in 2025 (up from ~21.5%).

The newest twist: in June 2026 DQ announced a pivot to manufacture electrical infrastructure for AI data centers — a "back to the future" power-equipment gamble that diversifies away from pure polysilicon. Early-stage; not yet a P&L line. Treat as optionality, not thesis.

Lens 2 · Supply Chain

Upstream → Daqo → end customer, named:

• Upstream inputs: metallurgical-grade silicon (MGS), liquid chlorine, hydrogen, nitrogen, calcium oxide — and, dominantly, electricity (polysilicon via the Siemens process is electro-intensive). DQ's structural edge is sited here: its Xinjiang (Shihezi) and Inner Mongolia (Baotou) plants sit on cheap captive coal-fired power, "much lower than most areas in China". The chokepoint is power cost and TCS (trichlorosilane) handling safety.
• Daqo (the converter): ~305,000 MT nameplate capacity (reached Q3 2024 with Phase 5B); ~350,000 MT per third-party tallies. Production was deliberately throttled: 197,831 MT (2023) → 205,068 MT (2024) → 123,652 MT (2025) → annualizing ~150k+ in 2026 guidance.
• Downstream buyers: China wafer/ingot makers (the framework-agreement customers), who feed cell and module makers (LONGi, Trina, JinkoSolar, JA Solar tier). DQ does not integrate forward — it exited wafers (Chongqing, discontinued 2018) and modules (Nanjing Daqo, sold 2012) to stay a pure upstream play.

Chokepoints / single-source dependencies: (1) power — the entire cost moat is captive Xinjiang/Inner-Mongolia electricity; (2) geography — substantially all assets in two PRC provinces, no geographic diversification; (3) the China demand sink — virtually all sales are domestic-China, so DQ is fully levered to one country's installation cycle (China added ~317 GW of the world's ~580 GW in 2025).

The supply-chain scar: Xinjiang Daqo is on the U.S. UFLPA Entity List and the BIS Entity List (added July 2021) over alleged forced-labor/XPCC links — so Daqo polysilicon (and any module containing it) is banned from import into the United States. DQ sells ~all output into China, so direct US revenue exposure is negligible; the damage is reputational and indirect (its downstream customers are foreclosed from the US module market) and it permanently caps any "sell to the West" optionality.

Lens 3 · Competitive Advantages (moats)

The honest answer: in a commodity glut, the only moat is the cost curve, and even that is not unique.

• Cost position (the real moat). FY2025 all-in production cost was $6.61/kg, up from $6.44 (2024) and $6.78 (2023) — the 2025 rise was not operational slippage but depreciation spread over idled volume (negative operating leverage from running at low utilization). Daqo is consistently cited as one of the lowest-cost producers globally on the back of cheap captive power. But peers match or beat it: Tongwei is building closed-loop TCS recovery to supply at ~$5.50/kg (≈20% under Western producers), and GCL offers FBR granular silicon at ~$6/kg using ~40% less energy. So Daqo is top-quartile on cost, not the low-cost king.
• Brand / quality: "one of the best-quality polysilicon makers in China" — n-type-capable, which matters as the industry shifts to TOPCon/n-type. Modest premium, not a durable moat.
• Switching costs / network effects: essentially none — polysilicon is a fungible commodity sold on spec and price. Framework agreements lock volume, not loyalty.
• Bargaining power: weak on both sides at the trough. Spot-linked pricing means customers hold the whip on price; the rising single-customer concentration (38.9%) further tilts power toward buyers. Upstream, MGS and power are commoditized/regulated.
• The non-obvious moat: the balance sheet + state backing. Zero debt, ~$2.0B liquid, and participation in the state-blessed "anti-involution" capacity cartel give Daqo staying power its weaker rivals lack. In a war of attrition, survivorship is the moat — but that is a balance-sheet attribute, not a business one.

Lens 4 · Segments

One segment, one geography — by management's own designation. The CODM (the CEO) reviews consolidated results and the company reports a single operating and reportable segment: Polysilicon. "Substantially all of the Group's revenues are derived in the PRC… and no geographical information is presented". So there is no product- or geography-level breakout to trend — the consolidated P&L is the segment.

The one meaningful "segment" split is DQ-attributable vs. minority interest, because Xinjiang Daqo's separate A-share float carves ~27% out of every result:

. The takeaway: ADR holders eat ~79% of the loss but also would get ~73% of any recovery — model the ADR off the attributable line, not the headline.

---

Phase B — Measure performance

Lens 5 · Earnings Result

The three-year collapse (FY2023 → FY2025), all ``:

Operating drivers (FY2025 vs FY2024):

• ASP: $11.48/kg (2023) → $5.66/kg (2024) → $5.25/kg (2025). The annual ASP decline decelerated: −66.7% (2023) → −50.7% (2024) → only −7.3% (2025), "mitigated by China's anti-involution initiatives since H2 2025". With cost at $6.61/kg, DQ sold below cash cost through 2025.
• Volume: external sales 200,002 → 181,362 → 126,707 MT (−30.1% in 2025); production cut harder, to 123,652 MT, to avoid dumping into the glut.
• Margin: stuck at −20.7% gross both years. The 2025 loss "improved" only because revenue and the loss both shrank ~35% in lockstep — not because economics turned.
• Below the line: the 2024 result carried a $175.6M long-lived-asset impairment (older Xinjiang lines); FY2025 booked nil impairment because "polysilicon selling prices rebounded substantially in 2025," restoring recoverability — a quiet but real positive tell. Income-tax benefit of $21.0M (2025) and a 15% preferential rate cushion the loss.

The Q1 2026 print (NEWER than the 20-F, ``): Revenue $26.7M (−78.5% YoY); gross loss $139.4M (−521.5% margin); net loss $88.4M (−$1.31/ADS). Sales 4,482 MT (−88.3% QoQ) — management deliberately stopped selling at trough prices; production was 43,402 MT. Q2 2026 guide: production 35,000–40,000 MT; FY2026 reaffirmed 140,000–170,000 MT. This is the operational nadir: revenue is now a rounding error because the company is hoarding inventory rather than realize losses.

Balance-sheet flags (the whole bull case): at Dec 31, 2025: cash + restricted cash $980.3M + fixed-term deposits $1,035.6M ≈ $2.0B liquid, zero bank borrowings, working-capital surplus $2,110.3M. BUT the clock: aggregate liquidity fell from $2.27B (Q4 2025) → $2.00B (Q1 2026) "primarily due to operational losses". That is ~$270M of cushion gone in one quarter — a ~$1B/yr burn at the trough. The fortress is real but it is draining on a timer.

Market reaction: the stock sits at $12.84 (Jun 26–29, 2026), the very bottom of its 2025 range ($12.74–$35.69) and near the 52-week low of $12.52, market cap ~$869M. The tape has priced DQ at ~0.4x its own net cash — the market is saying the cash will be burned or trapped.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts on the research-layer shelf (transcripts/ empty), so this is ``. The trajectory of management tone across the cycle:

• FY2023 calls: confidence, capacity-expansion narrative (Phase 5B ramp to 305k MT), cost-leadership boasts.
• FY2024 calls: shift to "discipline" and "weathering the downturn"; first impairment; first talk of idling older lines.
• Q4 2025 / Q1 2026 calls: "deep strain." Management explicitly frames the environment as severe overcapacity — monthly industry supply ~93,000 MT, industry utilization ~39% — and quantifies the price route: n-type poly RMB48 ($7.02/kg) end-2025 → RMB35 ($5.12/kg) end-Q1 2026. The recurring new vocabulary is "anti-involution," "production discipline," "self-discipline," and the capacity-buyout fund. What they stopped saying: growth, expansion, new phases. The pivot to AI data-center power equipment is the first "new story" injected to change the subject. Net read: management has moved from operator to survivor + policy-advocate, which is the correct posture for a trough but tells you the operating lever is gone — the catalyst is now external (the cartel).

Lens 7 · Comps

[capacity + cost: web: Bernreuter / Solar Power World / pv-magazine / market reports, 2025–2026].

What IS sourced and decision-useful: (1) the industry is an oligopoly in structure but a free-for-all in conduct — top-5 ≈ 64% of global capacity, China ≈ 93.5% of output, yet they competed prices below cash cost. (2) The only comp that matters right now is DQ-vs-its-own-balance-sheet: at ~$869M market cap against ~$2.0B net liquid assets and ~$4.41B DQ-attributable book, DQ trades at ~0.4x net cash and ~0.2x book. That is a deep-distress / sub-liquidation multiple — the relevant question is not "what P/E vs. Tongwei" but "is the book real and is the cash reachable?" (see Lens 10/13).

Lens 8 · Stock-Price Catalysts

DQ is a high-beta proxy for polysilicon spot price + China-ADR sentiment, not for its own operations. The >5% move pattern, ``:

• 2021–2022: parabolic up on the polysilicon shortage (ASP peaked; FY2023 ASP was still $11.48/kg after a −66.7% fall) — Daqo printed $429.5M attributable profit in 2023.
• 2021 (Jul): BIS Entity List addition — sharp sentiment hit; permanent US-market overhang.
• 2022–2025: structural de-rate as the glut built; ASP fell ~95% peak-to-2025.
• Aug 2025: polysilicon prices +~70% in a single month on the anti-involution cartel announcement — the sharpest up-catalyst of the cycle.
• 2025 range: $12.74–$35.69 — a ~2.8x peak-to-trough within one year, entirely driven by cartel hope vs. glut reality.
• Apr–Jun 2026: new lows ($12.52–$12.84) on the catastrophic Q1 print.
• The live macro wildcard: the 20-F itself quotes U.S. Treasury Secretary Scott Bessent declining to rule out forced delistings of Chinese issuers as a trade-war lever — a binary tail that can move DQ regardless of polysilicon.

What the market actually reacts to: (1) polysilicon spot price / cartel headlines (the dominant driver), (2) China-ADR delisting/trade-war risk, (3) far less to Daqo's own earnings, which are now a foregone conclusion (loss). This is a policy-and-commodity instrument.

---

Phase C — Judge people & books

Lens 9 · Management

Founder-family controlled, high insider ownership, recent succession. +.

1. Track record: the Xu family built Daqo from a 1980s industrial group into a top-3 global polysilicon producer with a genuine low-cost asset base; they timed the 2021 Xinjiang-Daqo STAR IPO superbly (raised RMB6.07B at the top) and rode the shortage to a $429.5M profit year (2023). That is real capital-cycle skill on the up-leg.
2. Tenure & skin in the game: CEO + Chairman Xiang Xu (mid-50s) — director since 2007, took CEO in Aug 2023, president of Daqo Group since 2006, directly owns ~11.41% of DQ. Heavy, aligned insider stake. Founder Guangfu Xu (80s, Xiang's father, Daqo Group chairman since 1984) stepped back from the DQ board in Aug 2023. Xu Xiaoyu (Xiang's sister, ex-J.P. Morgan IB) is Deputy CEO + Board Secretary + head of IR (joined May 2023, fast-tracked). This is a family business with a Western-finance-literate next generation — a positive for capital-markets navigation, a flag for governance independence.
3. Capital allocation: mixed-to-good through the cycle. Up-leg: disciplined expansion (Phases 4B/5A/5B) funded substantially by the STAR IPO and customer advances, not leverage — hence zero debt at the trough. Down-leg: capex spigot slammed shut — investing outflow $1,196M (2023) → $1,481M (2024) → $134M (2025); "we currently do not expect to incur material capital expenditures in the near future". They are hoarding cash correctly. The blemish: two $100M buyback authorizations (Jul 2024 and Aug 2025) — and zero shares actually repurchased despite the stock at ~0.2x book. For a board that owns 11%+ and claims the stock is cheap, authorizing-but-not-buying is a capital-allocation tell that cuts against the deep-value thesis (Lens 13).
4. Red flags: related-party web is dense (Daqo Group has 30+ affiliated entities under Guangfu Xu); amounts due to related parties of $5.1M at year-end (small). The fast-track nepotistic promotion of two family members is a governance watch-item. No SEC enforcement (Lens 10).
5. Archetype: founder/family operators, not professional managers — implies long horizon, willingness to idle and wait out the cycle (good for survivorship), but also concentrated control and limited minority-holder leverage.

Lens 10 · Forensic Red Flags

Acting as a forensic analyst against financials.csv-equivalent data in the 20-F:

• Revenue recognition: clean and simple — point-in-time commodity sales, spot-linked. Low manipulation surface. The aggressiveness is the opposite of inflation: they are deferring revenue by not selling (Q1 2026 sales 4,482 MT) and building inventory — watch for a future inventory write-down if prices don't recover, the mirror image of the 2024 $175.6M long-lived impairment.
• Cash flow vs. earnings: operating cash flow is the bright spot and broadly tracks the loss — OCF $1,616M (2023) → −$435M (2024) → +$49.7M (2025). FY2025 generated positive operating cash despite a $170.5M attributable loss (depreciation add-back + working-capital release), which is reassuring — earnings quality is better than the GAAP loss suggests at the cash line. No earnings-vs-cash divergence to flag.
• Receivables / inventory: the allowance for expected credit loss is growing ($18.1M 2024 → ~$19.3M 2025) on "long-aged receivables" — a small but persistent collectability flag in a buyer's market.
• SBC: non-cash share-based comp is large and flatters the cash story — $121.0M (2023), $72.4M (2024), $55.8M (2025) run through SG&A. Falling, but still ~$56M against a company with $26.7M of quarterly revenue — dilutive and worth haircutting.
• Goodwill/intangibles: minimal; the 2024 impairment was PP&E (older lines), already taken. The recoverability reversal logic in 2025 (no impairment because "prices rebounded") is judgmental — if 2026 prices stay at RMB35, expect impairment pressure to return.
• The buried asset-quality question: $3,232.8M of PRC-subsidiary capital, statutory reserves and IPO proceeds are legally PROHIBITED from distribution as dividends/loans up to the Cayman parent. Parent-only cash was just $311.2M (Dec 2025) / $559.4M (Mar 2026). So the "$2.0B fortress" is largely trapped inside China — book is real, but a large slice is not reachable by the ADR holder without clearing PRC capital-account controls (SAFE approval).

Regulatory findings (required sub-section):

• SEC Litigation Releases / AAERs: None. regulatory/regulatory-findings.md (SEC EDGAR EFTS, LR + AAER, period 2021–2026) returned 0 findings.
• Non-SEC enforcement (web search): the material item is trade/forced-labor, not securities — Xinjiang Daqo is on the U.S. BIS Entity List (Jul 2021) and the UFLPA Entity List, banning its polysilicon from US import over alleged XPCC/forced-labor links. Daqo denies forced labor; the listing nonetheless stands and is a permanent overhang. No FTC/DOJ/FDA/CFPB monetary penalties found.
• Item 3 / Legal Proceedings (10-K equivalent — 20-F Note 13(c)): Xinjiang Daqo contract dispute with Xinjiang Xian'an New Materials (silicon-core processing). Claim oscillated RMB1,958.5M → RMB742.7M → RMB744.9M (Jan 2026); courts have repeatedly rejected the consequential-damages claims, awarding only ~RMB3.3M; case is under retrial; company expects "no adverse effect on daily production and operations". Immaterial to solvency.
• Auditor / HFCAA: auditor is Deloitte Touche Tohmatsu (Shanghai), PCAOB-registered and inspected; DQ was a Commission-Identified Issuer only for FY2021 and has not been since FY2023 (PCAOB regained China access Dec 2022). Tail risk = re-determination if PRC obstructs PCAOB access.
• Verdict: No material accounting/securities regulatory findings — verified via SEC EDGAR EFTS (LR, AAER), web search, and 20-F Note 13 as of 2026-06-30. The real regulatory burden is the UFLPA/BIS trade listing and the HFCAA/delisting tail, both disclosure-confirmed.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection

EPS modeling here is low-confidence by nature — DQ's near-term EPS is dominated by one un-forecastable variable (polysilicon ASP), and the company is currently choosing not to sell. Per --watchlist rules, no forecast.ts forecast is logged. Three scenarios for DQ-attributable EPS per ADS (1 ADS = 5 ordinary; ~67.7M ADS), `` with arithmetic:

Anchors: FY2025 attributable loss $(170.5)M ≈ $(2.52)/ADS. Q1 2026 was $(1.31)/ADS. FY2026 volume guided 140–170k MT. Cash cost ~$6.6/kg; current ASP ~$5.1/kg.

• Bear (glut persists, ASP ~$5/kg all-year): sells ~150k MT below cash cost; gross loss ~$(200)M+, attributable net loss roughly $(2.50)–$(3.50)/ADS for FY2026. Cash burns another ~$0.8–1.0B → liquidity ~$1.0–1.2B exiting 2026. The cushion has ~2 years at this rate.
• Base (cartel half-works, ASP recovers to ~$7/kg = cash breakeven-plus by H2 2026): roughly breakeven-to-small-loss, ~$(0.50)–$(1.50)/ADS FY2026, with FY2027 turning to a modest profit as utilization normalizes. This is the consensus-ish path implied by the $26.77 average analyst PT (a >2x from spot).
• Bull (cartel cuts ~1M MT, ASP normalizes to ~$9–11/kg = mid-cycle): at $9/kg vs $6.6 cost on ~200k MT, gross profit ~$480M, attributable EPS could swing to ~$+3–5/ADS within 2 fiscal years. This is the "eps snaps back to 2023-like" case and is what a return to $30+ would discount.

The honest framing: DQ's FY2026–27 EPS is a leveraged option on the polysilicon price, which is a leveraged option on cartel execution. Do not anchor on a point estimate; size to the scenario, not the EPS. Base call (not logged): FY2026 attributable EPS negative, FY2027 the first plausible profit year if the cartel holds.

Lens 12 · Bull vs Bear

Bull case. DQ is a survivor with a balance-sheet put. Zero debt, $2.0B liquid, top-3 global scale, top-quartile cost, n-type-capable — it will outlast the weaker half of Chinese capacity that is bleeding faster. The anti-involution cartel is real and state-blessed: a RMB50B ($7B) fund, a registered JV (Beijing Guanghe Qiancheng Technology, Dec 2025, RMB3B initial capital), and an explicit plan to buy and shut ~1M MT (~1/3 of capacity); poly already +~70% in Aug 2025 on the announcement. The 15th Five-Year Plan formally names anti-involution a national priority, with a proposed mandatory energy-consumption standard that would force-retire obsolete capacity — and Daqo's efficient Xinjiang/Inner-Mongolia lines are winners of that cull. Trading at ~0.4x net cash and ~0.2x book, with an embedded AI-data-center-power optionality and a $100M buyback authorized, the asymmetry on a cycle turn is large (2025 already showed a 2.8x range). Contrarian view the market refuses to see: the cartel doesn't need to fully succeed — it just needs to put a floor under ASP at cash cost; at $7/kg DQ stops bleeding and the 0.4x-net-cash multiple is absurd.

Bear case (2–3 permanent-impairment risks). (1) The cartel fails — "any plan to shut capacity may struggle to secure funding and could face pushback from local governments"; China has announced supply discipline before and defected. If ASP stays at RMB35 (~$5/kg), DQ burns ~$1B/yr and the cushion is a 2–3 year fuse, not a fortress. (2) The cash is not the ADR holder's — $3.23B is PRC-distribution-restricted; parent-only cash is ~$0.56B; a HFCAA re-determination or Bessent-style forced delisting could strand the ADR above a wall of trapped RMB (the 0.2x book is partly justified by un-reachability). (3) Structural demand air-pocket — China's 2026 PV install guide is a wide 500–667 GW band that includes a possible decline from 580 GW, and the PV VAT export rebate is cancelled from Apr 1, 2026 — a fresh demand/margin headwind. Pre-mortem (18 months out, thesis broke): the cartel stalled on local-government resistance, a soft 2026 China install year kept utilization ~40%, DQ's liquidity fell through $1.0B, a new inventory write-down hit, and a trade-war delisting headline took the ADR to single digits — the "cheap vs. cash" thesis dies because the cash was both burning and trapped. Are multiples too high? No — they are low, but appropriately so given a negative-EPS, cash-burning, governance-and-jurisdiction-impaired asset. Cheapness is not the catalyst.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the deep-value bull:

• "It's below net cash" — but whose cash, and is it still cash next year? $3,232.8M is legally trapped in PRC subsidiaries; the parent has ~$0.56B. And the consolidated cash fell $270M in a single quarter. A discount to burning, ring-fenced cash is not a margin of safety — it is the market correctly pricing optionality decay.
• Revenue concentration is worsening into a buyer's market: one customer is 38.9% of revenue. At the trough, that customer dictates price.
• The moat is weaker than bulls think: Tongwei (~$5.50/kg) and GCL (FBR ~$6/kg, −40% energy) can underprice Daqo; Daqo is top-quartile, not low-cost king. In a commodity, second place on the cost curve still loses money in a glut.
• Most dangerous competitor bulls underrate: Tongwei. At ~910k MT it is ~3x Daqo's capacity and the natural cartel leader/winner — any consolidation likely concentrates share toward Tongwei, not Daqo.
• Worst management tells: $100M buybacks authorized twice, zero executed while insiders claim the stock is cheap and own 11% — they are not buying their own "bargain"; large ongoing $55.8M SBC dilutes the ADR holder; dense related-party group + family-promotion governance.
• What must hold for today's $12.84: that the cash is real and reachable and the company survives to a cycle turn and the ADR isn't delisted. Break any one and you're below $10.
• Growth disappoints 20–30% (here: ASP stays at trough 6 more months): liquidity through $1.0B, inventory write-down, possible covenant-free but confidence-shattering capital raise at the A-share sub — ADR to high single digits.
• Single scenario that permanently impairs: a forced HFCAA/trade-war delisting while ASP is sub-cost — strands ADR holders above trapped RMB with no US venue. Plausibility: low-but-non-trivial and explicitly flagged by the company quoting Bessent. That is the real short thesis — not the business, the wrapper.

Lens 14 · Management Questions (ordered by information value)

1. At RMB35/kg, what is your monthly cash burn, and at what liquidity level (e.g. $1.0B? $0.5B?) do you change strategy — raise capital, sell the A-share stake down, or restructure?
2. Of the $2.0B liquidity, how much can actually be upstreamed to the Cayman parent within 12 months given the $3.23B PRC distribution restriction and SAFE controls — and what is the realistic cost (tax, time) of doing so?
3. On the anti-involution capacity fund: what is Daqo's committed capital, what is the binding mechanism that prevents members defecting, and what specific tonnage has actually been bought and shut to date?
4. You authorized $100M of buybacks twice and bought nothing — if the stock is below net cash and you own 11%, why not buy? What price would trigger repurchases?
5. What is your base-case polysilicon ASP for H2 2026 and 2027, and at what ASP does Daqo return to operating cash breakeven?
6. How real is the AI-data-center electrical-infrastructure business — projected revenue, capex, margin, and timeline — and is it a genuine pivot or a narrative bridge?
7. What is your plan if the PCAOB loses China inspection access again or the US moves to delist Chinese issuers — is a Hong Kong primary listing prepared?
8. Given the VAT export-rebate cancellation (Apr 1, 2026) and the wide 500–667 GW 2026 install band, what is your demand assumption and downside plan?
9. How do you defend against Tongwei (~$5.50/kg) and GCL (FBR ~$6/kg) undercutting you — what is your credible path back to cost leadership?
10. What inventory are you carrying after the Q1 sales halt, and at what ASP would you take an inventory write-down?
11. What is the strategic logic of keeping the dual listing (NYSE ADR + Shanghai STAR sub) given the ~27% minority leakage — would you ever simplify it?
12. What capital-return policy applies once the cycle turns — special dividend, buyback, or reinvestment?
13. How exposed is your cost base to PRC power-price reform / the proposed per-unit energy-consumption standard — winner or loser?
14. What is the realistic 2026 industry utilization you're planning for, and what share of the ~1M MT capacity cut do you expect to be your competitors' vs. yours?
15. What governance steps protect minority ADR holders given the family control and related-party group — independent-director majority, related-party transaction policy?

---