TPI Composites

Phase A — Understand the business (what it was)

Lens 1 · Company Overview

TPI Composites was the last U.S.-listed independent, pure-play manufacturer of composite wind-turbine blades — a contract manufacturer ("toller") to the wind OEMs. Founded 1968, producing composite blades since 2001; IPO July 2016 at $11.00/share. It produced over 103,000 blades since 2001.

• Model: a "collaborative dedicated supplier" arrangement. OEMs outsource a portion of their blade volume; TPI dedicates manufacturing lines at low-cost-geography plants (Mexico, India, formerly Turkey/China) under multi-year supply agreements with minimum annual volume commitments, liquidated-damages provisions for late delivery, and cost-plus-margin pricing recognized over time (cost-to-cost percentage-of-completion).
• Products: wind blades (~95% of FY2025 net sales) plus field inspection/repair services (~5%).
• Customers: two, and effectively one going forward. In FY2025 GE Vernova = 56.4% and Vestas = 41.9% of net sales — 98.3% combined. Former customers Nordex (Matamoros line closed at contract end June 30, 2024) and Senvion (insolvent 2019) are cautionary footnotes.
• The structural flaw, stated plainly: the toller sits between a concentrated, powerful buyer (2–3 global OEMs) and volatile input costs, with a fixed-price contract and minimal pricing power. The "revenue visibility" of dedicated capacity is the same thing as operating leverage with no demand control — when the OEM slows, the toller eats the idle-line and transition cost. That is exactly what happened.

Lens 2 · Supply Chain

Map: raw material suppliers → TPI plants → OEM → wind-farm owner/operator.

• Upstream inputs (chokepoints): advanced fiberglass fabrics, carbon fiber reinforcements, balsa/foam core, epoxy/resin, adhesives, gel-coat, steel root hardware. Resin and carbon are price-volatile and passed through only with lag under fixed-price contracts — a persistent margin sink. Notably, one customer sources substantially all of its own raw materials, leaving TPI fewer levers on that program.
• TPI plants (at Dec 31, 2025): Juárez ×3 + Matamoros (Mexico), Chennai (India), Newton, Iowa (US, restarted 2H-2025); engineering centers in Kolding (Denmark) and Berlin (Germany); HQ Scottsdale, AZ. Turkey (Izmir) sold Sept 2025; Automotive (TPI, Inc.) divested June 2024.
• Downstream (the buyers who won): GE Vernova and Vestas Wind Systems A/S — global top-3 OEMs. The single most important supply-chain fact of this whole file: the customers ended up buying the plants out of bankruptcy. Vestas took Chennai + Matamoros ($24M); ECP (a PE firm, backed by GE Vernova's supply needs) took the rest ($20M). A supply chain in which your customer can, and does, vertically integrate you at a distressed price is not a supply chain you have power in.
• Trade chokepoint that became fatal: U.S. Customs & Border Protection (CBP) restricted importation of a substantial portion of Mexico-made blade models into the U.S. under the Uyghur Forced Labor Prevention Act (UFLPA) — a rebuttable forced-labor presumption. TPI's largest plants (Mexico, 78% of FY2025 sales) were partially cut off from the largest market. See Lens 10.

Lens 3 · Competitive Advantages (the moat that wasn't)

The bull-era pitch was: proprietary vacuum-infusion process, replicable/scalable plants, "BladeAssure" (AI-vision QA), 100k-blade track record, low-cost geography, and dedicated OEM relationships. In practice none of these were durable moats:

• No pricing power over customers. Two buyers controlling 98% of revenue set the terms. GE Vernova's ability to set off liquidated damages against amounts owed — which it did in late July 2025 — is the definition of negative bargaining power; it triggered the liquidity crisis.
• No switching-cost moat — the opposite. The customer owns the blade IP/design; TPI's process is customer-specific but the asset (the tooled plant) is transferable. When TPI failed, the OEMs simply bought the plants. Switching cost accrued to TPI, not away from it.
• No cost moat. Mexico/India labor advantage was eroded by wage inflation, peso/rupee FX, and — decisively — the UFLPA import ban that neutralized the geography.
• Quality was a negative differentiator. Recurring blade defects/failures and warranty campaigns: additional model-specific warranty accruals of $42.7M (2023), $22.9M (2024), $19.7M (2025) on top of base warranty; accrued warranty reserve $45.8M at Dec 31, 2025. Verdict on moat: a commodity contract manufacturer misclassified as a technology company. The only genuine asset was tooled capacity — worth ~$44M in a forced sale.

Lens 4 · Segments (geographic; USD, continuing ops)

Read: revenue was flat-to-up in FY2025 — this was not a top-line death, it was a unit-economics death. Mexico, 78% of sales, lost $115M at the operating line. India, historically the only profitable segment, flipped to a loss on price/mix. The U.S. (Iowa) burned $64M in restart/pre-petition costs. Every geography was underwater. By-product: Wind blades $869.9M / Field services $48.5M.

---

Phase B — Measure performance (the tape and the print)

Lens 5 · Earnings Result — the final annual print (FY2025 10-K)

The last audited results are a liquidation-grade income statement:

• Net sales $918.5M (+3.2% YoY).
• Total COGS $1,029.1M = 112.0% of sales → gross LOSS of −$110.6M (−12.0% margin), worse than FY2024's −8.1%. TPI lost money on the blades before a dollar of overhead.
• Loss from operations −$179.4M (−19.5%) vs −$134.8M (−15.1%) in FY2024.
• Interest expense, net −$89.0M — the Oaktree debt alone consumed ~10% of revenue.
• Reorganization items, net −$49.6M (post-petition professional fees $54.5M; retention $3.9M; offset $11.6M gain on liabilities subject to compromise).
• Net loss to common ≈ −$340.8M (−37.1% of sales) vs ≈ −$240.7M (−27.0%) in FY2024.
• Balance sheet flags (all red): unrestricted cash $13.9M (from $143.3M a year earlier — ~$184M consolidated cash burn); total stockholders' deficit −$679.5M; working capital deficiency −$484.7M; pre-petition Oaktree secured term loan $476.9M (incl. PIK); liabilities subject to compromise $323.8M; DIP drawn $23.9M of an $82.5M facility.
• Operating cash flow −$126.4M (from +$12.5M in FY2024) — a $138.9M swing.
• The tell hiding in plain sight: TPI factored $746.8M of receivables non-recourse in FY2025 (81% of revenue; up from $474.1M in FY2024), at a $14.2M loss on sale. A company selling nearly all its receivables at a discount is financing itself on the customer's balance sheet — a classic terminal-liquidity signature.
• Auditor / controls: management concluded ICFR and disclosure controls effective at Dec 31, 2025; going-concern substantial doubt is disclosed under ASC 205-40. This was an operating/leverage death, not an accounting fraud.
• Sequential deterioration (Q3-2025 10-Q, period ending Sept 30, 2025): cash $29.5M, stockholders' deficit −$582.8M, working-capital deficiency −$400.7M — each worsened into Q4.

Lens 6 · Earnings Calls (sentiment trend)

No earnings-call transcripts are on the research shelf (transcripts=0), and TPI stopped normal IR cadence as it moved to deregister — so tone is reconstructed from filed MD&A and press language. The arc of management language:

• 2021–2022 (peak): growth/decarbonization tailwind, capacity expansion, AMPC (IRA 45X) as a margin kicker.
• 2023: pivot to "cost discipline," warranty remediation, and the December Oaktree recapitalization sold as a liquidity backstop.
• 2024: "focus entirely on core wind" — code for divesting Automotive, exiting Nordex/Matamoros, shrinking.
• 2025 (terminal): the vocabulary collapses to "restructuring," "debtor-in-possession," "§363 sale," "going concern," "cancellation of equity." The FY2025 10-K's forward-looking section is a 40-item risk litany led by CBP/UFLPA and DIP-default risk. What they stopped saying is instructive: gone are "operating leverage," "margin expansion," and the AMPC tailwind (killed by OBBBA — Lens 10).

Lens 7 · Comps — read-through table (TPI equity = $0; peers are the survivors)

A P/E for TPI is meaningless (equity canceled). The useful comp is who won the consolidation. Multiples are `` where sourced, n/a otherwise (no fabrication).

Read-through: onshore wind is recovering (market ~$64.7B in 2026 → ~$99.2B by 2033, 6.3% CAGR ) — but the value is accruing to the vertically-integrated OEMs, not the independent toller. TPI died into a sector recovery.

Lens 8 · Stock-Price Catalysts — the arc from $81 to zero

• IPO Jul 2016 @ $11.00; all-time high $81.36 on Feb 16, 2021 (peak wind/ESG euphoria, IRA anticipation).
• 2021–2022: customer defections (Senvion insolvency legacy, Nordex/Vestas line closures), first blade-warranty scares, and CFO/COO turnover began the de-rate (a Pomerantz shareholder investigation was floated Jan 2022).
• Dec 2023: the Oaktree recapitalization — nominally a rescue, structurally an "uptier" (Lens 9/13) — signaled equity was now subordinated to a distressed-debt insider.
• Aug 11, 2025: Chapter 11 filing; shares −40% intraday, then a bankruptcy-lottery +90% squeeze on the DIP-approval headline (pure noise — "trading bears little or no relationship to recovery," per the company).
• Aug 19, 2025: Nasdaq delisting → OTC Pink "TPICQ."
• Mar 2026: §363 sale agreements (Vestas/ECP/GE Vernova) confirm equity cancellation. Pattern the market reacted to, in order of force: (1) balance-sheet/liquidity events, (2) customer concentration shocks, (3) policy/subsidy headlines. Never a product story. A pure balance-sheet equity.

---

Phase C — Judge people & books

Lens 9 · Management

• CEO William E. Siwek, 63 — the through-line of the whole story. CFO 2013–2019, President 2019–2020, CEO since May 2020. Career: 18 years at Arthur Andersen (Partner, Audit + Business Consulting) then CFO roles in real estate. A finance operator, not an industrialist — he built the capital structure that later crushed the company and then presided over its liquidation. The accounting-lineage irony (Andersen) is not causal but is thematically hard to miss.
• CFO Ryan Miller (since May 2022); COO-Wind Charles Stroo (since Nov 2023).
• Capital-allocation record: value-destructive. Levered a fixed-margin toller into a cyclical demand downturn; funded startup/transition losses (Iowa, Juárez) with expensive secured debt; carried a debt load whose interest (~$89M/yr) exceeded any realistic mid-cycle EBIT. ROE/ROIC were deeply negative for years; total stockholders' deficit reached −$679.5M.
• Red flag — pre-petition insider retention cash. On June 13, 2025 — ~8 weeks before the Aug 11 filing — the board granted cash Retention Bonuses: Siwek $1,225,459; Miller $518,155; Stroo $487,500, and the executives' equity awards were cancelled/forfeited in exchange. Legal and common in restructurings, but the optics — insiders converting worthless equity to guaranteed cash on the eve of a filing that zeroed public shareholders — are exactly the misalignment a skeptic underwrites. Siwek's total FY2025 comp was $2.20M.
• Board: capable on paper — Jayshree Desai (CFO, Quanta Services), Bavan Holloway (ex-Boeing VP Corporate Audit; audit-committee chair), Tyrone Jordan (auto-supplier COO; also FuelCell Energy board). The board's central failing was allowing the Dec-2023 Oaktree uptier while Oaktree held a board seat (Lens 13).
• Archetype: professional-manager/financier team steering a capital-intensive commodity manufacturer — precisely the profile that optimizes the balance sheet until the balance sheet optimizes the company into Chapter 11.

Lens 10 · Forensic Red Flags

Not fraud — the numbers were honestly disclosed — but the accounting told the death story a year early for anyone reading it:

1. Negative gross margin (−12%) — the core product was structurally unprofitable; no volume fixes a per-unit loss.
2. Receivables factoring to the hilt — $746.8M non-recourse (81% of revenue) at a $14.2M loss; working capital was being sold, not financed.
3. PIK interest on the Oaktree term loan — accruing principal the company had no cash to service; interest expense ($89M) dwarfed operating results.
4. Warranty as a recurring "special" charge — $42.7M/$22.9M/$19.7M model-specific accruals (2023-25) signal chronic product quality liabilities, not one-offs.
5. Cash-flow/earnings divergence — operating cash flow −$126.4M while the reported loss was "only" −$341M because of non-cash PIK and factoring mechanics; the cash picture was worse than GAAP in liquidity terms.

Regulatory findings (required):

• SEC enforcement: None. No Litigation Releases and no AAERs name TPI Composites (2021–2026).
• Bankruptcy adversary proceeding (the material legal event): the Official Committee of Unsecured Creditors sued Oaktree (complaint Oct 1, 2025, S.D. Tex.) over the ~$400M December-2023 "uptier" — alleging Oaktree, as board insider + preferred holder, converted ~$436M of Series A preferred equity into a $393M first-lien senior secured term loan (+$43M to common) without new money, "leapfrogging" unsecured creditors while TPI was insolvent or near-insolvent. Settled for $18.275M (Dec 2025).
• Trade enforcement (a proximate cause of death): CBP UFLPA import restriction on Mexico-made blade models — detained/excluded a substantial portion of TPI's largest-plant output from the U.S. market.
• Item 3 / Note 18 legal: the Senvion GmbH insolvency-estate voidance claim (~$13.3M, filed Jan 2021, re: pre-insolvency payments); a July-2025 court expert found Senvion was solvent at the time, so TPI has meritorious defenses. (Senvion is another failed wind OEM — a recurring counterparty-mortality theme.)
• Securities litigation: only a stale 2022 Pomerantz "investigation" surfaced; no material securities class action found.

---

Phase D — Project & stress-test (the autopsy)

Lens 11 · Forward Projection — recovery waterfall, not EPS

There is no forward EPS — equity is canceled; per SKILL rules for --watchlist no Brier forecast is logged. The only meaningful projection is the §363 recovery waterfall:

• Estimated asset-sale proceeds: Vestas (Chennai $10.0M + Matamoros $14.0M) = $24.0M + ECP Blade Holdings = $20.0M ≈ $44M core (GE Vernova $21M was the backstop if ECP fell through; Türkiye already sold Sept 2025 with liabilities assumed).
• Against: $607M funded debt at petition ($472M Oaktree secured term loan + ~$135M convertible notes); $476.9M Oaktree balance (incl. PIK) at Dec 31, 2025; DIP $23.9M drawn; $323.8M liabilities subject to compromise.
• Waterfall outcome: DIP and admin/professional claims consume the top; the Oaktree secured position takes the assets via credit-bid economics and still suffers a large haircut; unsecured creditors recovered a fraction (the $18.275M uptier settlement is the visible unsecured pool); common equity: $0, canceled. Confirmed.
• The counterfactual base case that never existed: even at mid-cycle utilization, a −12% gross-margin manufacturer paying ~$89M/yr interest cannot reach positive net income; there was no debt-free path to daylight once the Dec-2023 recap locked in the interest burden. The business needed both a demand recovery and a balance-sheet wipe — it got the wipe.

Lens 12 · Bull vs Bear — retrospective

• The bull case (what believers underwrote, and why it failed): "picks-and-shovels of the energy transition," IRA/AMPC subsidy tailwind, dedicated-capacity revenue visibility, 100k-blade moat, low-cost geography. Every pillar broke: subsidies were pulled forward and then cut (OBBBA phased AMPC out to Dec-2027; executive orders froze federal-land permits); "visibility" was operating leverage in disguise; the "moat" was transferable tooling; low-cost geography was neutralized by UFLPA + FX + wage inflation.
• The bear case (which was correct): (1) customer concentration with negative bargaining power (98% two customers who can insource); (2) negative unit economics under fixed-price contracts with pass-through lag; (3) a balance sheet that couldn't survive one downturn after the Oaktree recap. Any one was survivable; together they were terminal.
• Pre-mortem, now a post-mortem: "It's 18 months out and the thesis broke — what happened?" Answer, verbatim from the file: a key customer (GE Vernova) set off liquidated damages in late July 2025, CBP froze Mexico imports, cash hit $14M, the DIP couldn't be repaid, and the OEMs bought the plants for $44M. All of it was visible in the −12% gross margin and 81% receivables factoring a year prior.
• Contrarian view the market still under-appreciates: The independent wind-blade toller is a structurally dead business model. The sector's recovery will be captured by vertically-integrated OEMs (Vestas insourcing LM Wind Power's Poland plant; GE Vernova consolidating LM; both buying TPI's plants). There is unlikely to be a re-listed pure-play blade toller — the category was disintermediated, not merely bankrupted.

Lens 13 · Devil's Advocate (short-seller — the thesis that paid)

A skeptic dismantling the 2021 bull would have won on structure, not timing:

• Where revenue is concentrated: two OEMs = 98%; one of them (GE Vernova) chose to enforce liquidated damages and is the buyer of last resort. When your customer is also your executioner and your acquirer, you have no business.
• Why the moat is weaker than bulls think: the customer owns the design IP; the plant is a transferable, commoditized asset; process "know-how" did not prevent recurring warranty failures.
• The most dangerous "competitor" bulls underestimated — the customers themselves. OEM insourcing is the structural kill: Vestas acquired the Goleniów, Poland blade factory from LM Wind Power (2025); GE Vernova/LM is reinvesting in core in-house EU capacity and divesting outsourced ops. The toller exists only until the OEM decides the economics favor bringing it in-house — and in a downturn they always do (at a distressed price).
• The worst capital-allocation / governance move: the Dec-2023 Oaktree "uptier" — an insider (board seat + preferred stake) converting at-risk equity into first-lien debt on substantially all assets without new money, priming other creditors (later challenged and settled for $18.275M). This is the move that guaranteed common equity would be zero; from that date TPI was a zombie whose assets belonged to Oaktree.
• What breaks the model permanently — and did: a demand air-pocket while carrying ~$600M of debt against a negative-gross-margin product. Plausibility in 2023: high. It happened in 2025.

Lens 14 · Management Questions (the autopsy set — ordered by information value)

Reframed as the questions a board/analyst should have forced, ranked by how much the answer would have changed the outcome:

1. At what utilization does a blade line earn a positive gross margin — and what % of lines were below it in 2023–24? (The −12% margin says: most of them.)
2. Why accept fixed-price contracts with lagged raw-material pass-through when resin/carbon are volatile — and what is the realized pass-through recovery rate?
3. In the Dec-2023 Oaktree recapitalization, what new money entered the business, and how was the priming of unsecured creditors justified to the independent directors?
4. What is the customer-concentration kill-switch plan if GE Vernova or Vestas enforces liquidated-damages setoffs? (There was none.)
5. What is the contingency if CBP/UFLPA restricts Mexico blade imports — i.e., how much of Mexico output can reach the U.S. under a detention?
6. What interest coverage does the business achieve at mid-cycle EBIT, and at what point does PIK accrual become unrepayable?
7. Why factor 81% of receivables non-recourse — what does that say about days-payable pressure and true liquidity?
8. What is the standalone economic case against the OEMs insourcing each dedicated line?
9. What is the cumulative cash cost of blade warranty campaigns since 2019, and is the reserve adequate?
10. What returns did the Iowa restart and Juárez ramp actually earn versus the startup cash burned?
11. What is the plan if AMPC/IRA support is repealed or pulled forward? (OBBBA did exactly this.)
12. How transferable is the plant/tooling to a competitor or the customer — i.e., what is liquidation value vs. going-concern value?
13. Why were pre-petition executive retention bonuses structured as cash while public equity was wiped?
14. What is the realistic re-contracting position for 2026+ given supply agreements expired at end-2025?
15. If the answer to (1) and (6) is "never" and "soon," why is this an equity at all rather than a distressed-debt or private situation? — the question that, answered honestly in 2022, avoids the entire loss.

---