Thesis: Single-crystal nickel-superalloy turbine-blade caster — picks-and-shovels upstream of GEV in the AI gas-turbine supercycle; ~3-companies-on-Earth metallurgy moat.

The story right now

Howmet is the upstream layer of the gas-turbine bottleneck we've been building thesis on for months. Crusoe's Lochmiller (operator-side) confirmed gas turbine pricing 3x'd ($1M/MW → ~$3M/MW) on 2026-05-02. GEV was the OEM beneficiary. @Gaurab's 2026-04-29 turbine-blade thread points one layer further up: the blade is the actual physical chokepoint. Each blade is a single crystal of nickel superalloy grown in a vacuum furnace at 3 mm/min; a set costs $600K and takes 90 weeks; only three companies on Earth can build them. China spent $42B trying to reverse-engineer a Russian engine and got 30hrs of life vs 4,000hrs Western standard. The metallurgy is a 60-year moat, not a 60-month one.

HWM is one of those three. Its Engine Products segment (the original Alcoa investment-castings business, pre-Arconic separation) makes investment castings + forged jet engine components for both jet engines and industrial gas turbines — the exact product whose order book is sold out to 2030. The aerospace narrative gets all the airtime; the IGT exposure is the asymmetric variant nobody is repricing yet.

The tape agrees something is happening: $256.43, +9.9% past 30 days, RSI 60, golden cross, +63% 1y, +488% 3y, +675% 5y, +35pp 1y alpha vs SPY. Op margins held 22-26% through 2025; FCF margin 24.5% operating-plus-investing cash flow ran 24.5% of Q4 revenue (correction 2026-08-03: OCF $654M + investing −$122M — capex is undisclosed in the cache, so this is not a true FCF margin) — still an extraordinary cash profile for an industrial. Cap structure is clean (no debt reported in the latest snapshot). Net margin 17-20% all four quarters of 2025; revenue accelerating $1.94B → $2.17B Q1→Q4.

The yellow flag: insider trading bearish, 10 sells / 0 buys / -$170M net, with the stock 4% off 52w high. Compounders that have run this hard often see C-suite sell into strength regardless of forward thesis — but it's noise we shouldn't dismiss.

Bull case

• The chokepoint is real. Single-crystal blade casting is a 60-year metallurgy moat. Three suppliers globally. Industrial gas turbine OEMs (GEV, Siemens Energy, Mitsubishi Heavy) cannot meet AI-data-center demand without HWM-class capacity. As gas-turbine prices triple, the upstream blade supplier captures spread expansion regardless of whether GEV holds its 25%+ pricing premium.
• AI-data-center IGT mix is mispriced. Sell-side still models HWM as a commercial-aerospace cycle play (737/787/A350 build rates). The IGT side — historically the smaller stub segment — sees demand step-change as hyperscalers shift from waiting-for-grid to onsite gas. Same product, different end market, different multiple.
• Picks-and-shovels arithmetic. GEV is +169% 1y already; HWM is +63%. If the gas-turbine cycle has another leg, the upstream supplier gets repriced after the OEM does, not before.
• Operator quality. 100% earnings beat rate per leading indicators. Margins held expansion through 2025 even as revenue grew. FCF margin 24.5% gives capital flexibility for capacity expansion or buybacks.
• Recent acquisition tape. Brunner Manufacturing acquired (per recent news) — extending the precision-machining capability that feeds blade-casting and forging.

Bear case

• Aerospace cycle is the dog, IGT is the tail. Engine Products + Fastening Systems + Engineered Structures + Forged Wheels are still mostly aerospace cycle exposure. If 737 MAX rate decelerates, A350 ramp slips, or a recession hits airline capex, the IGT thesis won't carry the whole P&L.
• Insider tape is unambiguous: 10 sells, 0 buys, -$170M net. This is selling into strength near 52-week highs. Insiders see something we don't or are de-risking concentrated stock comp; either way, the asymmetry is wrong.
• Multi-bagger pricing. +488% 3y, +675% 5y, alpha 1y +35pp. Reversion to mean is the default outcome from here, not continuation. RSI 60 isn't euphoric but the multi-year run leaves zero margin for execution slip.
• The "3 companies on Earth" claim is partially folklore. Single-crystal castings exist outside the West (Russia, China, Japan, India have programs) — they're just not yet at par on engine-life. Time/CapEx/IP-theft can compress the moat. Verify the 30hr-vs-4,000hr engine-life number from primary sources before relying on it.
• No revenue acceleration in the leading-indicators composite. Score 0.4 (neutral) — the bull case rests on forward IGT mix shift, not what's already in the numbers.

Catalysts

• Q1 2026 print (mid-2026) — IGT segment commentary; first chance to see whether AI-data-center mix is showing up in management language. Watch backlog by end-market.
• GEV / Siemens Energy / MHI capacity announcements — every multi-billion gas-turbine plant expansion implicitly commits to HWM capacity.
• DPA Title III follow-ons. The 2026-04-20 grid-equipment determination established federal-priority precedent. A turbine-blade addition would re-rate the moat publicly.
• Hyperscaler onsite-gas announcements (Anthropic, OpenAI, xAI, Crusoe, Stargate variants). Each one validates the IGT-mix-shift thesis and pulls forward HWM revenue recognition.

Risks to the thesis

• HWM is largely an aerospace stock with IGT optionality. The IGT optionality is the asymmetry; if it doesn't materialize at expected magnitude, you own a hot industrial in cycle.
• Insider selling is loud. Watch for any insider buying — that would be the single highest-information signal here.
• Single-crystal blade-casting is one product line within Engine Products. Verify what % of HWM revenue actually flows through that line vs. the broader investment-castings + forging book before sizing.

Setup

• Entry zone: $244 - $256 (current $256.43 - 5% to current)
• Stop: $205 (-20%)
• Target: $333 (+30% — modest given 5y +675% baseline; first-look conservative)
• Conviction: medium — moat is real, the asymmetry is real, but multi-bagger pricing + insider selling argue for sizing as a cohort name not a flagship.

Cross-references

• Perspective: 2026-05-03-ai-power-bottleneck — generation-side parent thesis
• Sister deep-dive: the same-day ATI deep dive — alloy-supplier layer above HWM
• Adjacent layer: GEV (gas-turbine assembler, already in ai-power watchlist + key_tickers)
• Source events: @Gaurab's April 29 capture (primary), a April 30 recycler-themed capture (derivative)
• Workspace cluster: same physics-first chokepoint archetype as EHV transformers (NuttyCLD/FrontierMap convergence) and GOES electrical steel (CLF Butler)

Open questions

1. Cross-check China $42B + 30hr-vs-4,000hr engine-life numbers from primary sources (defense aerospace literature, NDIA reports, congressional testimony). Bullet-proofs the moat language if true; downgrades it to "industry talking point" if can't be sourced.
2. What % of HWM Engine Products revenue is IGT vs aero? This is the single most thesis-critical number. Should be in segment disclosures.
3. Who are the 3 companies exactly? Best public guesses: HWM, MTU Aero Engines (DE-listed), Rolls-Royce (UK PLC). Possibly Doncasters (private), PCC (Berkshire-owned, private), or NSK Japan. If only HWM is investable in liquid US listings, that's the cleaner long.
4. HWM's IGT customer concentration. GEV exposure size? Siemens Energy? MHI? Concentration risk is the inverse of pricing power.

Verification — closed 2026-05-06

Cross-validation pass on the four queued questions. Net: thesis trust-hardens; one new disclosure-narrowing risk surfaced.

(a) China reverse-engineering claim — VERDICT: mostly matches (confidence: high on $42B, medium on hours)

$42B confirmed. Observer Research Foundation (ORF Online): "by the end of 2020, total investment in the 'Two Engines' project had doubled to nearly RMB 300 billion (US$42 billion)" across WS-10/WS-15/WS-19/WS-20 in parallel. Source: orfonline.org/expert-speak/aviation-engine-success-propelling-china-s-air-combat-power. The "30hr vs 4,000hr" is a directional narrative compression — actual sourced figures: early WS-10 ~800hr, current WS-10A faulted ~20,000 times in first three years (PLA-logged); WS-15 service life ~3,600hr; F119 4,000–6,800hr. Two confirmed turbine-blade-traced WS-15 explosions (2015, 2018). The order-of-magnitude moat claim is real; the specific "30hr" number is narrative compression not literal precision.

(b) Three-company oligopoly claim — VERDICT: partially overstated, structurally still a duopoly (confidence: high)

The duopoly is 2 companies, not 3: Howmet + Precision Castparts. In Practise expert call: "PCC and Howmet probably own 80% of the [single-crystal] market, maybe even higher." Rolls-Royce operates a captive Rotherham foundry (100k blades/yr capacity, opened 2015) but uses output internally rather than competing for OEM contracts. Doncasters, MTU, and small Spanish foundries fill niche/equiax roles. Net: PCC being inside Berkshire makes HWM the only liquid pure-play, which strengthens the trade rather than weakening it — the @Gaurab "3 companies" framing undersells PCC's scale and overstates the closed-club narrative, but the HWM-as-only-investable-leg conclusion holds.

(c) HWM Engine Products IGT vs aero split — VERDICT: partial / disclosure narrowed (confidence: medium-high)

Engine Products FY2025 = $4.32B (+16% YoY). Adj EBITDA $1.4B, 33.3% margin. End-market growth FY2025: commercial aero +12%, defense aero +21%, gas turbines +25% (fastest line). Disclosure caveat (small new thesis risk): HWM combined IGT + Oil & Gas into "Gas Turbines" starting Q4 2025, eliminating the standalone IGT line going forward. Company-wide Q4 2025 split: ~53% commercial aero / 20% defense aero / 12% gas turbines / 12% commercial transport / 3% other. Engine Products subset not separately disclosed. Action surfaced as followup: pull 2024 10-K filed Feb 2025 to back-fit the last clean IGT$ baseline before the narrowing.

(d) IGT customer concentration — VERDICT: confirmed structurally, not quantified (confidence: medium)

Howmet's Hampton VA foundry serves GE, Siemens, Alstom, Mitsubishi (industry-press confirmed; spinoff.nasa.gov + foundry trade press). Whitehall MI serves GE Aircraft Engines, Pratt & Whitney, Rolls-Royce, Rocketdyne, Honeywell. No single customer >10% disclosed. The GEV/Siemens-Energy/MHI IGT triad is structurally confirmed; quantitative concentration unavailable from public disclosures.

Verification scorecard

3 of 4 questions source clean (a, b, d); 1 partial-with-disclosure-risk (c). Picks-and-shovels framing holds and strengthens — the duopoly is real and HWM is the only liquid leg. Cinematic numbers source as directionally true; the "30hr" specifically is narrative compression. New small risk: Q4 2025 disclosure narrowing means we lose clean IGT$ tracking going forward, so the bull case has to be triangulated via segment growth rates rather than dollar lines. No new tickers added — Doncasters private, PCC inside Berkshire (BRK.B too diversified to count as a thesis vehicle), MTU not pure-play, Rolls-Royce integrated. Coverage of the duopoly is already maximally captured by HWM + ATI in ai-power.json.

Followups surfaced (filed in TASKS.md)

• Pull HWM 2024 10-K end-market revenue table for Engine Products segment — back-fit historical IGT$ baseline before disclosure-narrowing.
• Track HWM "Gas Turbines" line YoY in Q1/Q2 2026 prints — confirm 25% growth rate persists; inflect-down before GEV order-book inflects = leading bear tell.
• Open question: does PCC publish ANY external production figures via Berkshire annual letter or BNSF-adjacent rail-volume data? If yes, that's a HWM-comp benchmark.
• Evaluate Rolls-Royce ADR (RYCEY) as adjunct play — captive Rotherham foundry technically inside the duopoly, but RR is integrated engine-maker not pure-play caster.

Verification sources

• HWM FY2025 release: howmet.com/press-release/2026-02-12 — Engine Products $4.32B +16%; gas turbines +25% YoY
• In Practise expert call: inpractise.com/articles/howmet-and-precision-castparts-single-crystal-hpt-blade-investment-casting-process — PCC+HWM ~80%+ of single-crystal market
• ORF Online: orfonline.org/expert-speak/aviation-engine-success-propelling-china-s-air-combat-power — $42B "Two Engines" cumulative 2010–2020
• Rolls-Royce Rotherham facility: rolls-royce.com — captive, 100k blades/yr (2015 opening)
• Military Watch / Global Defense Corp / globalsecurity.org — WS-15 explosions, life-hour figures
• Foundry industry press (spinoff.nasa.gov and trade refs) — HWM Hampton VA + Whitehall MI customer rosters

10-K Back-Fit Appendix — closed 2026-05-06

Headline finding: HWM has NEVER disclosed standalone IGT$. The 10-K segment-reporting note has reported Engine Products non-aerospace revenue as a single combined "Industrial and Other" line back to at least FY2022. The Q4 2025 disclosure change combined IGT + Oil & Gas in narrative commentary only — the dollar table was already combined the whole time.

Engine Products end-market disaggregation (2024 10-K, Note C — Revenue from Contracts with Customers)

Estimated IGT$ — derived, NOT disclosed

Q4 2024 deck slide-5 footnote provides the only public ratio, at company-wide "Industrial & Other" level: IGT ~45% / General Industrial ~30% / Oil & Gas ~25%. Engine Products skews more IGT-heavy than Fastening/Forged-Wheels segments (Engine Products IGT% inside the $878M FY2024 line is plausibly 50–60%, putting FY2024 IGT$ at ~$440–530M). No public disclosure pins it tighter.

Disclosure timeline

• 2024 10-K (filed Feb 14, 2025): Engine Products disaggregation table shows ONLY Aero-Commercial / Aero-Defense / Industrial & Other. IGT and Oil & Gas appear ONLY in narrative commentary, never as separate dollar lines.
• Q3 2025 earnings call (Oct 30, 2025): combination first announced in narrative.
• Q4 2025 / FY2025 results (Feb 12, 2026): "Gas Turbines market" growth reported as +25% FY YoY, +32% Q4 YoY. No standalone IGT$ or O&G$.

Reframes the existing "Gas Turbines line YoY tracking" followup: best anchor is Engine Products Industrial-and-Other = $878M FY2024 / $798M FY2023 / $677M FY2022 (+10%, +18% YoY); future quarters triangulate via segment growth rates rather than IGT dollar lines. Open: pull FY2025 10-K (drops mid-Feb 2026) to check whether HWM restated FY2024 disaggregation under the new "Gas Turbines" combined scheme.

Sources

• 2024 10-K HTML (CIK 4281, accession 0000004281-25-000011): sec.gov/Archives/edgar/data/4281/000000428125000011/hwm-20241231.htm
• 2024 Q4 earnings deck (Feb 13, 2025) — slide 5 (45/25/30 footnote), slide 7 (Engine Products Q4 mix 55/21/24)
• Q3 2025 earnings call transcript (Oct 30, 2025) — combination announcement
• 2025 Q4 results press release (Feb 12, 2026)

PCC Benchmark Appendix — closed 2026-05-06

BIG FINDING: Berkshire DOES publish PCC line-item revenue + pre-tax earnings annually in MD&A — far more disclosure than typical Berkshire subsidiaries. PCC is broken out specifically alongside BNSF / BHE / GEICO. HWM-comp tracking IS feasible at YoY revenue/margin level; granular tracking (capacity, customer mix) remains opaque.

Reported PCC figures (Berkshire 2024 Annual Report MD&A, filed Feb 2025)

The 2025 letter (Feb 2026) calls PCC "Berkshire's largest industrial manufacturing business" and discusses post-pandemic aerospace recovery narratively. Quantitative pre-tax earnings continue to appear in the MD&A.

What's NOT disclosed

Capex, headcount, capacity by alloy/process, single-crystal-blade-specific revenue, customer concentration (GE/PW/RR/MHI mix), aftermarket vs OEM split, plant-level data. HWM-style segment detail unavailable.

Net answer

HWM-comp tracking IS feasible at YoY revenue/margin level. Track PCC YoY rev growth vs HWM Engine Products YoY — divergence is the signal (e.g., HWM accelerating while PCC flat = HWM taking share, as the late-2025 episode showed per Seeking Alpha/Finterra notes). For the granular layer (single-crystal-blade share, capacity by alloy/process), would need Tegus / AlphaSense / Vertical Research Partners — paid expert-network channel.

Other channels — all dry

• AIA / NDIA / industry trade groups: no PCC-specific disclosure. Industry-wide stats only.
• Customer-side disclosures (GE Aerospace 2024 Investor Day, PW/RR/MHI investor materials): imply but never name PCC. GE Aerospace flagged $100M supplier-investment program for castings/forgings — implied PCC, no name.
• BNSF Railway: no PCC-specific shipment volumes.
• PCC residual SEC filings: went fully dark on EDGAR post-2016 acquisition.
• Third-party expert networks (In Practise): qualitative ("PCC and Howmet probably own 80% of the market, maybe even higher for single crystal"), no revenue split.

Sources

• Berkshire 2024 Annual Report (SEC): sec.gov/Archives/edgar/data/1067983/000119312525054885/d810841dars.pdf — primary source for $10.4B / $1.9B
• Berkshire 2025 Annual Report: berkshirehathaway.com/2025ar/2025ar.pdf (FY2025)
• Benzinga PCC $2B rebound recap (2025-03)
• Berkshire 10-K: sec.gov/Archives/edgar/data/1067983/000095017025025210/brka-20241231.htm
• GE Aerospace 2024 Investor Day deck — implies PCC, doesn't name
• Finterra HWM 2026 deep dive — quantifies the late-2025 PCC stumble episode
• In Practise: HWM/PCC single-crystal investment-casting interview

Followups filed

• Extract literal MD&A PCC sub-paragraph from brka-20241231.htm to inputs/profiles/precision-castparts.md for diff-tracking against the 2025 10-K when it drops Feb 2027.
• Standing PCC-vs-HWM-Engine-Products YoY divergence quarterly tracker.

Sources

• Cash-flow figures in the 2026-08-03 correction re-derived from cash-flow-statements.
• 2026-04-29-single-crystal-nickel-superalloy-turbine-blades-upstream-chokepoint-above-gev-ga — @Gaurab primary
• 2026-05-07 — Massive 8-Q financials + 10-K excerpts + price + leading indicators
• ai-power — cohort comparison
• log — Crusoe / Lochmiller 2026-05-02 capture, Lonis Hamaili horizontal-slice cross-check
• inputs/FOLLOWS.md — @Gaurab Trial follow row (added 2026-05-06)