Bull case

• CRS is named by ATI as the legacy-nickel-superalloy competitor. ATI's FY2025 10-K (filed 2026-02-20) explicitly identifies Carpenter Technology as the primary HPMC-segment competitor for "legacy nickel-based alloys and superalloys and specialty steel alloys." That alloy class is the load-bearing input for single-crystal IGT-blade casting at HWM. The named-competitor relationship establishes CRS's position in the actual supply chain that drives the AI-data-center IGT thesis — not the inferred position the 2026-05-06 sequence assigned to ATI.

• Margin expansion at high rates of change. Op margin 19.0% → 23.0% in 4 quarters (4-quarter sequence ending Mar 2026). FCF margin and net income margin moving similarly. For a specialty-metals business with a fixed asset base, +400bp of operating margin in a year reflects mix shift into higher-spec products (nickel superalloys, titanium powders, additive-manufacturing feedstocks) — exactly the direction the AI-IGT / commercial-aerospace / defense + medical demand bundle points.

• Specialty Alloys Operations is the levered segment. CRS's SAO segment is the alloy-supplier business; PEP is downstream powder + parts. SAO is the segment that captures alloy-layer pricing power. As nickel-superalloy ingot/billet/forgings into engine OEMs grow (commercial jet recovery + IGT acceleration + defense), SAO's margin should outpace PEP. Need segment-level Q3 disclosure to confirm but financial-summary trajectory is consistent.

• Surfaced by blind research-trace, then independently confirmed. The 2026-05-06 blind trace spike found CRS at 2/3 convergence ("nickel-superalloy powders for additive manufacturing of turbine parts") — the trace surfaced CRS before any human-narrative-down framing. ATI's 10-K then independently confirmed CRS's position. Two independent paths converging on the same name without coordination is the substantive-class convergence signal pattern (compare to the four-skip-class-recyclers-on-Aschenbrenner pattern, where convergence inverts to skip).

• Defense + commercial aerospace + medical + energy diversification works. CRS serves aerospace, defense, medical, transportation, energy, industrial, and consumer markets. The defense/aerospace/medical mix is structurally growing; medical specifically (orthopedic implants, surgical instruments) uses CRS specialty steels in volumes that the broader specialty-metals competitors don't always match. This is quality of revenue diversification, not a "diluted exposure" weakness.

• Materials-cycle alignment with HWM. If HWM Gas Turbines +39% YoY is the leading indicator of an IGT pull-through wave, the alloy-layer-supplier-of-record should rise in parallel. CRS's +11.6% revenue + 35.3% op income YoY is consistent with that wave but more diversified — CRS gets the IGT lift through alloy mix without being singular-exposed to it.

• CRS named in ATI's competitor list AND in independent research-trace = the opposite of the ATI-stack misread. The thesis here is not based on inferred vertical relationship; it's based on an explicit disclosure plus an independent algorithmic surface.

• No-debt balance sheet, fragmented float ownership. CRS doesn't carry the cap-structure complexity of larger A&D names. Smaller cap (vs HWM $97B / vs ATI $21B mcap) at $21.3B — CRS is comparable to ATI in size, ~1/5 the size of HWM.

Bear case

• Multi-year-run pricing. +110% trailing 12 months is multi-bagger geometry. Mean reversion is the default outcome from this kind of run. Entry at $429 is -6.5% off 52w high — not a deep entry zone.

• No direct customer-concentration disclosure pulled yet. The CRS 10-K customer-concentration paragraph hasn't been read directly (TODO — defer to user). If CRS has 10%+ customer concentration the way ATI explicitly doesn't, the supplier-risk profile differs. Worth pulling.

• Specialty-metals cycle exposure. Like ATI, CRS is structurally a cyclical specialty-metals company. Op margin 23% is good but compresses in a downturn. The thesis depends on the engine-cycle staying strong.

• PCC (private inside Berkshire) is the larger competitor. Berkshire's PCC subsidiary competes with CRS across nickel-superalloys + titanium + precision forgings + investment castings. PCC's scale + Berkshire balance-sheet support is a real competitive moat. CRS captures share at the margin, not in the core.

• Aubert & Duval is the European equivalent of CRS for some forgings. Cross-border substitution exists on the high-spec product class. European engine OEMs (RR, Safran) historically source forgings from European suppliers preferentially — limits CRS's growth in that specific lane.

• The "alloy supplier to which caster" question stays partially open. ATI's 10-K names CRS as a competitor in nickel-superalloys but doesn't disclose CRS's specific customer mix vs HWM vs PCC vs internal-melt-at-OEMs. CRS could be selling alloy primarily to engine OEMs (GE/PW/RR/Safran) who specify both alloy and caster separately — the same OEM-mediated structure ATI has. If so, CRS's "alloy layer above HWM" framing has the same caveats (not direct supplier-to-HWM, but supplier-to-OEM-who-specifies-HWM-as-caster). Read CRS 10-K Item 1 to confirm. ~10 min.

• PEP segment is lower-margin and more cyclical. Performance Engineered Products is the downstream powder + parts business — competitive with broader specialty-metals players. SAO is the alloy story; PEP dilutes the segment-level alloy-layer pure-play narrative.

• Insider-tape and 13F overhang not yet checked. Standard discipline missing from this first-pass deep-dive. Worth a follow-up read.

Catalysts

• CRS Q4 FY2026 print (~Aug 2026) — the near-term catalyst. If revenue and margin continue accelerating (Q3 print was +11.6% rev / +400bp op margin / +35% op income YoY), the alloy-layer story compounds. Crucially, if SAO segment margin expansion outpaces PEP, that confirms the alloy-mix-shift thesis vs the broader specialty-metals cycle.
• CRS FY2026 10-K filing (~Aug 2026) — will disclose customer concentration; will say whether HWM, GE Aerospace, RTX, Rolls-Royce, Safran are top customers. The customer-concentration disclosure is the cleanest test of the "alloy layer to engine OEMs" framing.
• HWM Q2 2026 10-Q (~early Aug 2026) — the standing HWM tracker (TASKS.md:422). If HWM Gas Turbines continues at +30%+ YoY rates, CRS's alloy mix is reinforced; if it inflects down, CRS's lift is at risk too.
• AI-data-center IGT pricing news — Crusoe, Anthropic, OpenAI, Stargate announcements that name specific GW commitments + power equipment + turbine OEMs. Each datapoint reinforces the IGT-data-center thesis.
• Defense FY27 appropriations — defense end-market is structurally growing for specialty-alloy suppliers; appropriations cycle is a calendar-trigger.
• HWM segment disclosure changes — if HWM ever breaks out alloy-vendor mix (unlikely; competitive-harm redactions are standard), that would directly answer whether CRS supplies HWM materially.

Risks to the thesis

• The "CRS is the alloy-layer-above-HWM-for-IGT" thesis is inferred-by-elimination (ATI's not it; ATI's 10-K names CRS as the nickel-superalloy competitor; CRS surfaced independently via blind trace). It's not yet confirmed by direct customer-disclosure on the CRS side. If CRS's own 10-K shows CRS sells primarily to medical/energy/non-aero markets rather than engine OEMs, the framing weakens.
• Multi-year-run reversion risk is real at +110% 1y.
• PCC (private) is the larger competitor; CRS captures share at the margin.
• Same vertical-reality-check caveat from the 2026-05-10 meta note applies in reverse: don't assume CRS supplies HWM directly; assume OEM-mediated until disclosed otherwise.

Setup

• Entry zone: $410 - $430 (current $429 / -6.5% off 52w high). RSI not currently extreme (need to verify in ai-power summary after CRS is added). Above $430 the entry geometry compresses; below $410 is fresh setup.
• Stop: $345 (-20% from $430 mid-range entry).
• Target: $525 (+22% — modest first-look at multi-year-run geometry; explicitly does NOT extrapolate the +110% 1y forward).
• Conviction: medium. Same tier as HWM and original ATI grading. The thesis rests on (a) ATI 10-K naming CRS as the nickel-superalloy competitor, (b) blind research-trace independent surface, and (c) the operational financial trajectory confirming a strengthening alloy cycle. The customer-concentration disclosure is the missing direct test; pull when the FY2026 10-K drops.
• Cohort framing: CRS sits at the alloy layer alongside ATI (specialty-alloy diversified) and below HWM (caster + assembler + vertical integration). The cohort sizing should reflect alloy-layer exposure as a sub-position, not a flagship.

Cross-references

• Perspective: 2026-05-03-ai-power-bottleneck — CRS added as key_ticker 2026-05-11 (structural-correction event)
• Sister deep-dives:
  • the May 6 HWM deep dive — caster downstream; vertically integrated
  • the May 6 ATI deep dive (revised 2026-05-10) — diversified specialty-alloy; commercial-jet driver; reframed away from IGT-stack
• Adjacent: GEV (turbine OEM assembler), HWM (single-crystal blade caster + own titanium melt), PCC (Berkshire-owned vertical-integrated competitor)
• Source events:
  • 2026-04-29-single-crystal-nickel-superalloy-turbine-blades-upstream-chokepoint-above-gev-ga — @Gaurab primary (the original surfacing event for the layer above casters)
  • 2026-05-10-ati-10k-hwm-competitor-not-customer-three-layer-stack-was-wrong — the structural-correction event that elevated CRS to key_ticker
  • a May 6 blind research-trace smoke test that surfaced CRS at 2/3 convergence
• Methodology references:
  • a May 10 methodology note — the lesson from ATI/HWM that promoted CRS to the right layer
• Watchlist: ai-power (added 2026-05-11)

Open questions

1. CRS customer-concentration disclosure. Pull FY2025 10-K (June year-end, filed ~Aug 2025) and read Item 1 / Risk Factors / financial-statement-notes customer-concentration paragraph. Are HWM, GE Aerospace, RTX, Rolls-Royce, Safran named as 10%+ customers? Or does CRS have ATI-like zero-concentration? ~10 min.
2. SAO vs PEP segment breakdown. Specialty Alloys Operations is the alloy-layer business; Performance Engineered Products is downstream powder + parts. What % of total revenue is SAO? What's the SAO segment margin trajectory specifically? Quarterly disclosures should answer; pull most recent 10-Q segment table.
3. Aerospace & defense mix. ATI is 68% A&D. HWM is 70% aerospace. What % of CRS revenue is aerospace + defense specifically? Critical for sizing the IGT-thesis weight.
4. Insider tape + 13F overhang. Standard discipline check.
5. Same vertical-reality-check question that fixed ATI: is CRS actually a direct supplier to HWM's casting operations, or is the relationship OEM-mediated (CRS sells alloy to GE/RR/PW/Safran who specify both alloy and caster)? Read CRS 10-K business-description for customer/end-market detail.

2026-05-26 update — FY2025 10-K customer-concentration read (resolves Open Questions 1, 3, 5)

Direct test of the thesis via CRS's FY2025 10-K (CIK 0000017843; full read in the May 26 customer-concentration investigation). Mirrors the 2026-05-10 ATI correction.

• (a) No 10%+ customer concentration — OEM-mediated structure, same as ATI. CRS discloses zero >10% customers across FY2025/FY2024/FY2023 and does not name HWM / GE Aerospace / RTX / Pratt & Whitney / Rolls-Royce / Safran as customers-of-record anywhere in the 10-K. The bear-case "OEM-mediated, not direct-supplier-to-HWM" framing (Risk note above, Open Question #5) is confirmed — assume CRS sells alloy to the engine OEMs who specify both alloy and caster, not to HWM's casting line directly.
• (b) New evidence ATI does not have — ~40% of net sales under firm-price LTAs. 40% of CRS net sales ($1.15B annualized) sit under firm-price long-term agreements with annual purchasing commitments and consumption schedules. That's structural multi-year supplier evidence at the grade × volume × duration level — a step more concrete than ATI, whose 10-K carries no equivalent LTA-share disclosure. Supplier-of-record evidence at the contract level, even if not at the customer-name level.
• (c) Re-read framing — diversified A&D pull-through, not direct HWM supply. CRS captures the IGT / jet / defense pull-through through a diversified A&D mix, not a named HWM relationship: A&D revenue +15% YoY ($1,538.8M → $1,768.6M), SAO operating margin +630bp YoY (16.7% → 23.0%), A&D mix-share 51% → 56% → 62% over three years.

Net: the thesis does not break — it sharpens. The structure is OEM-mediated (like ATI), but the ~40% LTA share + the accelerating, A&D-weighted mix give CRS more concrete supplier-of-record evidence than ATI had. Routing unchanged (stays in ai-power); conviction unchanged at medium — operational evidence still supports the bull case, but the absence of a direct HWM customer-of-record name keeps it from hardening to high.

Sources

• 2026-05-26-crs-fy2025-10k-customer-concentration-read — CRS FY2025 10-K (CIK 0000017843): zero >10% customer concentration FY25/24/23; HWM/GE/RTX/PW/RR/Safran not named; 40% net sales under firm-price LTAs ($1.15B annualized); A&D +15% YoY ($1,538.8M→$1,768.6M); SAO op margin 16.7%→23.0%; A&D mix 51%→56%→62%.
• ATI FY2025 10-K (filed 2026-02-20, SEC EDGAR accession 0001628280-26-010140): names CRS as HPMC-segment competitor in "legacy nickel-based alloys and superalloys and specialty steel alloys"
• info — $21.3B mcap, $429.41 close, +110% 12m, -6.5% from 52w high
• income-statements — 8-quarter trajectory; FY26Q3 +11.6% rev / +35.3% op income / +400bp op margin YoY
• 2026-05-06-derive-skill-smoke-test-gas-turbine — blind trace surface at 2/3 convergence
• log — 2026-05-10 structural-correction entry

Methodology note

This deep-dive is the first ticker added to ai-power.json via derived-then-verified path rather than narrative-down path. The 2026-05-06 blind research-trace independently surfaced CRS at 2/3 convergence before any human had named it; the 2026-05-10 ATI 10-K disclosure independently confirmed CRS's position by elimination. The path here is the inverse of the typical narrative-down flow (capture → frame → deep-dive). It's the path the Track B / B4 cold-problem-practice loop was built to validate: did our own-generated reasoning land on the same name our sources eventually land on? CRS is now an open hypothesis we can track for the next 30/90/180 days against the @Gaurab / @ChurchillWw / @vikramskr substantive-class roster and see whether one of them surfaces CRS independently in a public post. If yes → we matched their lead time. If no → either we surfaced something they missed, or we surfaced something the market didn't reward (worth a retrospective either way).