GlobalFoundries Inc. (NASDAQ: GFS) — A Mature-Node Price-Taker Wearing an AI-Datacenter Multiple

An independent, fundamental, evidence-driven research note. The analysis below carries no investment recommendation and no price target; the single exception is the clearly-labeled “Claude’s Take” block, which is the author’s own subjective view. General information only — not investment advice.

⚡ Claude’s Take

This block is the author’s own subjective opinion. It is general information, not investment advice. The analysis in the sections below takes no position and sets no price target.

Verdict: HOLD / AVOID-at-this-price; accumulate-on-weakness in the low-to-mid $40s; not a short. Low-to-medium conviction. Tag: “A price-taker priced like a price-maker.”

GlobalFoundries is a genuinely improving business strapped to a genuinely demanding price. The improvement is real: the revenue mix is rotating away from price-concession smartphone wafers toward record automotive (+17% in FY25) and a silicon-photonics/optical-connectivity datacenter franchise that already grew +29% and is on a credible path to a ~$1B run-rate by end-2028; the balance sheet holds net cash; free cash flow turned positive as capex was cut 79% off the 2022 peak; the build-out is heavily funded by the U.S. CHIPS Act ($1.5B direct + a $16B U.S. plan) and customer prepayments rather than GFS’s own cash; and it owns the one thing no competitor can buy — DoD Trusted-Foundry status and the only scaled, pure-play, non-China/non-Taiwan foundry footprint on earth. That geopolitical moat is the most durable, least-appreciated part of the story. The problem is the price, not the business. At ~$85.83 (≈$47.7B market cap, ~$46–47B EV) GFS trades at ~22× EV/EBITDA and ~54× GAAP earnings on revenue that is still below its 2022 peak and on a return on invested capital of ~6% — below its own cost of capital. The stock’s own valuation-history percentiles tell the whole story: P/S in the 99th, P/B 92nd, P/E 87th, composite 93rd — this is the richest the stock has ever been on its ~4.5-year public record. The factor tape confirms it: a beta-1.8, negative-alpha (−0.25) momentum name that has tripled off its April-2025 low as part of the AI-semis trade, not on company-specific quality, with a −55% three-year max drawdown on record to prove what happens when the semis factor turns. You are being asked to pay a TSMC-style multiple for a business with a quarter of TSMC’s gross margin and a fifth of its ROIC, on top of a relentless Mubadala selldown overhang. That is a poor entry, even for a good story.

What would flip me bullish: sustained gross margin above ~32–34% and ROIC moving durably above ~10% (WACC) as the photonics/auto/IP mix ramps — i.e., hard proof the pivot earns its capital — ideally with the Mubadala overhang cleared. What would flip me bearish: a renewed China mature-node price war or a utilization/auto-inventory rollover that resets EBITDA toward the low-$1.5B’s while the multiple is still 20×+ — the classic high-beta round-trip. The accumulate zone (low-to-mid $40s, ~10–12× EV/EBITDA, ~2.5–3× EV/sales) is where the reshoring optionality and government-funded build come free rather than fully paid for. Today they are fully paid for.

📈 Stock Price Action — Five-Year Event Map

GlobalFoundries IPO’d on the Nasdaq on 28-Oct-2021 at $47.00, so its public history runs ~4.5 years, not five — there is no earlier tape. Over that span the stock traced a deep V: a post-IPO climb to a 2021–22 high near $70, a grinding ~3-year de-rate to an all-time-low close of $30.37 (8-Apr-2025) through the auto/industrial downcycle and repeated Mubadala secondaries, and then a violent ~3× re-rating to an all-time-high close of $89.96 (26-May-2026) on the AI/reshoring narrative. As of 18-Jun-2026 the stock is ~$85.83, roughly −4.6% off the all-time high, inside a 52-week range of ~$31.51 → $92.55. The defining feature: price sits near record highs while revenue remains below its 2022 peak — a multiple re-rating, not an earnings-driven move (own valuation-history percentiles: P/S 99th, P/B 92nd, P/E 87th, composite 93rd).

Cycle narrative. (1) GFS debuted at $47 and ran toward ~$70 within weeks as the global chip shortage made foundry capacity maximally scarce — the high-water mark of the cycle. (2) As the shortage unwound and rates spiked, the stock gave back gains to ~$54 even though revenue actually peaked at ~$8.1B in 2022 — multiple compression, not an earnings miss. (3) A ~$65 rebound on early AI optimism that did not hold. (4) The long grind: the auto/industrial/comms inventory correction flattened revenue ($6.75–6.79B in 2024–25, below the 2022 peak) and recurring Mubadala secondaries added share supply, dragging GFS to its all-time-low close of $30.37 on 8-Apr-2025 (amid broad tariff/macro risk-off). (5) The $16B U.S. plan (4-Jun-2025) seeded the reshoring/AI thesis, but with the stock near $37 the initial reaction was muted. (6) The Q4-2025 print (11-Feb-2026) — a beat plus silicon-photonics and design-win momentum — jumped the stock from ~$42 to ~$49 and crystallized GFS as a recognized AI-semis name. (7) The Q1-2026 print (5-May-2026), with non-IFRS EPS of $0.40 beating $0.34 and a Q2 guide-up (~$1.76B), carried the stock to its all-time-high close of $89.96 on 26-May-2026. (8) It has since eased ~5% to ~$86 on broad semis chop — sector noise, not company news. (Price moves are FACT, from public price history; attributed drivers are INTERPRETATION.)

1. Executive Summary

GlobalFoundries is the world’s #4–5 dedicated semiconductor foundry and the only scaled, pure-play foundry with a manufacturing base spanning the United States, Europe and Asia. It deliberately abandoned the leading edge (no sub-12nm, no EUV) to specialize in differentiated “essential” mature-node platforms — FD-SOI (22FDX/12FDX), RF-SOI, FinFET 14/12LP, BCD power, SiGe, eMRAM/NVM, GaN and silicon photonics — sold into smart mobile, automotive, IoT, communications/datacenter and aerospace/defense end markets, with U.S. DoD Trusted Foundry status. It is a controlled company: Mubadala (Abu Dhabi) owned ~81% as of the FY25 20-F and is steadily selling down.

The central tension is the gap between an improving business and a demanding valuation. The business is improving: FY25 revenue was roughly flat at $6.79B (wafer volume +10%, offset by ASP erosion), but the mix rotated toward the best end markets — record automotive ($1.41B, +17%) and communications/datacenter ($745M, +29%) — and the silicon-photonics/optical-connectivity franchise is a credible AI-era growth vector, targeted at a ~$1B run-rate by end-2028. Free cash flow turned positive (~$1.0B FY25) as capex was cut 79% from the 2022 peak; the balance sheet holds net cash; and the heavy build-out is co-funded by the CHIPS Act and customer prepayments. The valuation is demanding: at ~$85.83 the stock trades at ~22× EV/EBITDA, ~54× GAAP earnings and the 93rd percentile of its own valuation history — on revenue still below its 2022 peak and a ~6% ROIC that sits below the cost of capital.

The quality verdict is mixed and honest: GFS has a real but narrow and shallow switching-cost moat (single-sourced specialty platforms that are costly to requalify, especially in auto/RF/defense) plus a genuinely durable geopolitical moat (Trusted Foundry; non-China/non-Taiwan footprint) — but the financial fingerprint of that moat is a 25% gross margin and a 6% ROIC, i.e., customer captivity without scale economics, a structural contrast with TSMC’s 60% gross margin and 30%+ returns. Management built capacity at the 2021–22 cycle top, impaired $935M of it in 2024, and is now allocating more defensively and pivoting up the value chain via IP/connectivity M&A (MIPS, Advanced Micro Foundry, InfiniLink, Synopsys ARC) — coherent in direction, but with undisclosed deal prices and rising goodwill. The price action and factor data frame the stock as a high-beta (β≈1.8), negative-alpha momentum name that has tripled off its April-2025 low as part of the AI-semis trade. The market is underwriting a structural margin and revenue step-up — and a re-rating that holds — that the financials do not yet demonstrate.

2. Business Overview

What GlobalFoundries does. GFS manufactures integrated circuits for ~200 customers under a pure-play foundry model: customers (fabless designers and IDMs) own the chip design; GFS owns the fabs and the differentiated process technology and is paid to manufacture the wafers (plus mask, NRE and packaging “non-wafer” services). Crucially, GFS does not compete at the leading edge. After years of trying — and after famously abandoning its 7nm program in 2018 — it repositioned around feature-rich mature and specialty nodes from 600nm down to 12nm, where differentiation comes from what the process can do (ultra-low power, RF performance, high-voltage power handling, non-volatile memory, photonics) rather than from raw transistor density. Its platform portfolio — FD-SOI (22FDX/12FDX), RF-SOI, FinFET 14LPP/12LP, SiGe, BCD/high-voltage BCD, RF and power GaN, eMRAM, and silicon photonics — is the differentiation engine. (FY25 20-F, “Business,” filed 2026-02-27.)

Manufacturing footprint. Fabs in Malta, NY (Fab 8, the flagship 300mm fab); Essex Junction, VT (Burlington, the legacy 200mm fab acquired from IBM); Dresden, Germany (Fab 1, 300mm, the European hub); and Singapore (Woodlands/Tampines, 200mm + 300mm). This US+EU+Asia spread, plus the addition of the Advanced Micro Foundry silicon-photonics fab in Singapore (Nov-2025), is the physical basis of the “geographically diverse, non-China/non-Taiwan” pitch. ~13,000 employees; HQ Malta, NY; Cayman-incorporated; reports in IFRS as a foreign private issuer (20-F/6-K, no 10-K/10-Q/DEF 14A).

Revenue by end market (FY25 20-F). The flat FY25 top line masks a sharp, favorable mix rotation:

The two structurally best end markets — automotive (a record, +17% on content and share gains) and communications/datacenter (+29%, five consecutive double-digit-growth quarters on SATCOM/optical networking) — are now 32% of revenue, up from 26% in FY24. The drag is Smart Mobile (−12%), which management attributes to GFS-initiated one-time price concessions to a small number of dual-sourced mobile customers plus the roll-off of LTA underutilization payments. (FY25 20-F; Q4-2025 call, 2026-02-11.)

Volume, ASP and the LTA model. GFS shipped ~2.3M 300mm-equivalent wafers in FY25, +10% y/y, at ~85% utilization — yet revenue rose only +1%, meaning blended ASP fell (the mobile concessions plus the disappearance of take-or-pay LTA payments that had inflated FY23–24 reported pricing). During the 2021–22 shortage GFS signed multi-year Long-Term Agreements (LTAs) under which customers reserved capacity, made prepayments, and owed underutilization (take-or-pay) payments if they under-shipped. As of Q1-2026, CFO commentary confirms those underutilization payments are “largely in the rearview mirror” — a quality-of-revenue cleanup, since they cushioned 2023–24 and are now gone. Management says the next capex wave will again be co-funded by customer prepayments and government grants, so the prepayment model is being re-armed for the onshoring build rather than retired. (Q1-2026 call, 2026-05-05.)

Customer concentration and design wins. Top-10 customers were ~63% of wafer volume in FY25 (down from 65% FY24 and 72% FY23) — high but falling, consistent with “the broadest customer base in our history.” GFS booked 500+ design wins in 2025, of which 95%+ were sole-source to GFS — the single most important moat lever (see §4). (Q4-2025 call.) Historically AMD (GFS’s former parent) was a wafer-supply-agreement anchor; that relationship has normalized and the book is far more diversified today.

Recurring vs. cyclical. Foundry revenue is fundamentally cyclical — both volume and ASP swing with end-market inventory cycles, as the 2022→2024 decline showed. The “recurring-like” element is the multi-year, single-sourced design-win install base on hard-to-requalify specialty platforms (especially automotive and defense, where requalification can take years) plus the ~11% non-wafer (NRE/mask/packaging) revenue. Verdict: a differentiated specialty foundry with an improving end-market mix, a real single-source design-win base, but inherently cyclical wafer economics and an ASP profile under pressure where it is not the sole supplier.

3. Industry Dynamics

Two industries under one label. “Foundry” is not one industry; it is two. The leading edge (≤7nm) is a structurally excellent, near-monopoly business: TSMC holds >90% share, EUV and escalating capex create near-insurmountable barriers, and gross margins run ~60%. The mature/specialty segment where GFS lives is structurally far weaker — more suppliers, lower barriers in generic nodes, and a capital cycle distorted by Chinese state subsidy. GFS competes in neither’s best part: above the commodity 28nm-and-larger logic fray it walked away from, but below the leading edge. Its arena is differentiated specialty mature — feature-rich CMOS, FD-SOI, RF-SOI, BCD power, SiGe, photonics, GaN.

Market structure and the competitive set. GFS’s blended foundry share is ~3.9% (TrendForce, via the TSM report). The competitive set (FY25 20-F competition section):

• TSMC — dominant overall (~70% blended share) and the largest specialty/mature player; it is the gorilla in every niche GFS touches and is selectively raising specialty prices.
• UMC — the closest pure-play mature peer (~4.4% share, slow growth).
• SMIC / Hua Hong / Nexchip (China) — the key structural threat: state-subsidized mature-node capacity. China’s share of top-10 mature capacity is set to exceed ~25% by end-2025.
• Samsung Foundry, Intel Foundry, in-house IDMs (e.g., Texas Instruments) — IDM/foundry hybrids.
• Tower Semiconductor (independent after the Intel acquisition collapsed), Vanguard (VIS), PSMC, X-FAB — smaller specialty foundries.

The capital cycle (Marathon lens) — broken, then inflecting. The 2023–25 mature-node downturn was a textbook bust: auto/industrial/IoT over-ordered into the 2021–22 shortage, then destocked; mature-node utilization fell to ~70% in 2024; ex-leaders’ mature-node operating profit fell ~23% that year; and China kept adding subsidized capacity — state capitalism breaking the normal market-clearing mechanism, exactly the “capital cycle breakdown” Marathon warns about. The 2026 supply side is inflecting more favorably than the bear case assumes: SMIC, Hua Hong, Nexchip and TSMC-affiliate VIS are reportedly raising mature-node prices ~10% (BCD, 12-inch) into AI pull-forward demand, and TSMC is scaling back legacy capacity (Fabs 2/5 to close by end-2027). Tightening and price hikes by the very players that were the overcapacity threat is a meaningful near-term positive — but it is partly AI-demand-driven and could reverse; the structural overhang has been absorbed, not eliminated. (TrendForce/DigiTimes/EE Times, Dec-2025–May-2026.)

The geopolitical tailwind. Tariffs, export controls and “non-China, non-Taiwan” sourcing mandates are driving reshoring. GFS’s US (Malta NY, Essex Junction VT) + EU (Dresden) + Singapore footprint and DoD Trusted Foundry accreditation make it a genuine structural beneficiary, with CHIPS Act, EU Chips Act and Singapore incentives subsidizing the build. This is the one place where the industry’s structural challenges work for GFS rather than against it.

Verdict: a structurally mediocre-to-poor industry in which GFS occupies the better-than-average seats. Generic mature foundry is an oversupplied, China-distorted, capital-cycle-broken sub-industry; GFS partly escapes via differentiated specialty niches and the onshoring tailwind, but it remains a price-taker wherever its product commoditizes, and its 25% gross margin (vs. TSMC’s ~60%) is the financial fingerprint of an inferior industry position.

4. Competitive Position

Name the moat. GFS’s differentiation maps to Greenwald’s customer-captivity / switching-cost advantage (a demand-side advantage), not economies of scale (TSMC owns that) and not a proprietary cost advantage (its 25% gross margin disproves a cost edge). Three pillars:

(a) Single-sourced, qualified specialty platforms = real switching costs. Re-qualifying a chip design at a different foundry is a multi-quarter, multi-million-dollar engineering effort carrying yield and schedule risk — acute in automotive (AEC-Q100 requalification spans years), defense (radiation-hardened, secure manufacturing) and RF/analog (process-specific and notoriously hard to port). With 95%+ of 2025’s 500+ design wins sole-sourced and top-10 concentration falling, this captivity is real and shows up as design-win persistence and pricing defensibility on single-sourced platforms. This is the genuine, identifiable moat.

(b) Trusted Foundry + US/EU footprint = a geopolitical moat. GFS is the only scaled US-based pure-play foundry besides Intel, with DoD Trusted Foundry status and a non-China/non-Taiwan footprint. In a deglobalizing sourcing regime this is a regulatory/access advantage that competitors cannot replicate at any rational cost — you cannot buy “located in the US with DoD accreditation.” Early evidence it converts: an Apple collaboration expanding US connectivity/PMIC work, plus engagements management describes across all four US hyperscalers and the top auto OEMs. This is the most durable pillar.

© FD-SOI ecosystem. GFS is the leading-volume FD-SOI (22FDX/12FDX) foundry, a differentiated ultra-low-power platform with an established IP/design ecosystem and a degree of ecosystem lock-in for IoT/RF/auto customers who designed in FDX.

The load-bearing skepticism — moat vs. ROIC. A durable competitive advantage must surface in financial outcomes. GFS’s ROIC of ~5.8% (FY25), ~8.4% (FY23) and a ~10.2% peak (FY22), with a 24.9% gross margin, are the hallmarks of a business with weak economics by Greenwald’s profitability test (sustained 15–25%+ ROIC signals a moat; 6–8% signals its absence). The honest reconciliation: the switching-cost moat is real but narrow and shallow. It protects retention and design-win persistence on specific single-sourced platforms, but it does not confer pricing power at the portfolio level — GFS was forced into one-time price concessions to dual-sourced mobile customers in 2025, the opposite of pricing power. Simultaneously, this is a brutally capital-intensive, depreciation-heavy business (gross PP&E $33.8B), and the moat is too narrow to earn an above-WACC return on that base. GFS is substantially a price-taker on anything dual-sourced or commoditizing, and a price-keeper only on the single-sourced specialty/defense long tail.

Versus TSMC and China. The TSMC contrast is clarifying: TSMC’s moat is its 60% gross margin (scale fused with captivity, still widening); GFS’s “moat” coexists with a 25% gross margin and a 6% ROIC — captivity without scale. Same industry, opposite ends of the advantage spectrum. Against China: the FD-SOI/RF/auto/defense niches are insulated from SMIC/Hua Hong (which compete on subsidized generic-CMOS price), but any GFS revenue overlapping generic mature CMOS is directly exposed to Chinese price-flooding — the 2025 mobile concessions are a live example.

Verdict: a real but narrow and shallow switching-cost moat, reinforced by a genuine, harder-to-replicate geopolitical/Trusted-Foundry moat — but not a moat that yet produces good economics. Defensible, not commoditizing; but the defense currently earns a sub-WACC return. The geopolitical pillar is the most durable and least-priced; the financials say the overall moat is too thin to call this a high-quality business today.

5. Growth History and Forward Opportunities

Historical: the 2022 peak → 2024–25 trough. Revenue ran $4.85B (2020) → $6.59B (2021) → $8.11B (2022 peak) → $7.39B (2023) → $6.75B (2024) → $6.79B (2025, flat). The ~17% peak-to-trough decline was a classic mature-node inventory correction: auto/industrial/IoT (~34% of sales) over-ordered into the shortage then destocked, with Q1-2024 revenue −16% y/y at “the bottom of the cycle.” Automotive was the lone bright spot throughout. The 2025 “recovery” then grew just +1% — and only because +10% volume was offset by ASP erosion (mobile concessions + LTA-payment roll-off). So far the recovery is volume-led and ASP-pressured — a low-quality recovery.

Forward driver 1 — onshoring/capacity (real, funded, capital-hungry). A $16B incremental U.S. commitment (Jun-2025, including $3B for advanced packaging, silicon photonics and GaN R&D) and a €1.1B Dresden expansion (EU Chips Act-supported). This is demand-pull (Apple, hyperscalers and auto OEMs mandating non-China/non-Taiwan supply) co-funded by customer prepayments and government grants. The Marathon caution: this is asset growth into a capital-intensive, sub-WACC-ROIC base. The tailwind is real, but adding rate base at ~6% ROIC is value-dilutive unless accompanied by genuine ASP/mix improvement. The metric to watch is whether incremental US/Dresden capacity earns above WACC.

Forward driver 2 — the AI-connectivity pivot via M&A (real revenue or narrative?). The most-hyped, most-scrutinized vector, assembled through MIPS (RISC-V IP, Aug-2025), Advanced Micro Foundry (silicon photonics, Nov-2025), InfiniLink (optical connectivity, Nov-2025), Tagore (GaN, 2024), a TSMC GaN technology license (Nov-2025) and Synopsys ARC IP (closing ~H1-2026).

• The most concrete datapoint: management guides to a ~$1B silicon-photonics run-rate by end-2028 (up from a prior, smaller target), built on pluggable optics, near-/co-packaged optics (CPO) and SiGe TIA/driver ICs for AI optical networking; Q1-2026 cited Malta NY tape-outs for CPO design wins supporting the OCI standard.
• Evidence it is partly real: Communications/datacenter already grew +29% in FY25 (five straight double-digit quarters) on SATCOM/optical — actual revenue today, not slideware — plus GaN collaborations (Navitas, onsemi, Cirrus Logic) into AI-datacenter power.
• Skeptic’s read: the $1B photonics target is a run-rate three years out (~15% of current revenue) — meaningful but not transformational, and unproven; the MIPS/RISC-V “physical AI / robotics” framing is the thinnest, most narrative-driven leg with negligible disclosed revenue; the string of acquisitions is itself a top-of-AI-cycle capital-cycle tell; and licensing GaN from TSMC (rather than self-developing) marks GFS as a fast-follower, not a leader.

Verdict: mixed-to-low quality, improving at the margin. The historical record is cyclical and the 2025 “growth” was volume-led with ASP erosion. Forward, the mix is improving (auto + datacenter now the engines, mobile shrinking) and the photonics/optical/SiGe vector is partly real revenue (not pure narrative), but the GaN and especially RISC-V legs are early-stage TAM-expansion stories assembled via top-of-cycle M&A, and the onshoring build pours capital into a ~6% ROIC base. The story is getting better and more differentiated — but it must clear the bar of earning an above-WACC return, which it does not yet do.

6. Financial Quality

Revenue and margins. Revenue is below its 2022 peak and roughly flat (FY25 $6.79B). Gross margin has compressed from 28.4% (FY23) and 27.6% (FY22) to 24.9% (FY25) — the combined effect of lower utilization off the peak, ASP concessions, and the loss of LTA underutilization payments. EBITDA margin fell from ~36% (FY22–23) to 31.1% (FY25); operating margin is 11.7%. Management guides gross margin to exit 2026 “at or above 30%” — a recovery thesis, not a demonstrated result.

Cash flow and the capex cut. The most important financial fact is the capex collapse: $3.06B (2022) → $1.80B (2023) → $625M (2024) → $722M (2025), a 79% peak-to-trough cut that flipped FCF from −$435M (2022) to ~+$1.0B (2024 and 2025). This is the source of the “FCF positive” narrative — but note it is FCF earned by starving a depreciation-heavy asset base (D&A was $1.31B in FY25 against $722M capex, i.e., GFS spent ~0.55× depreciation). Under-investing relative to depreciation can sustain FCF for a few years on a young fab base, but it is not a steady state for a foundry that intends to grow — and indeed management guides FY26 net capex back up to 15–20% of revenue (~$1.05–1.4B gross direction, net of grants). The “FCF positive” claim and the “we’re re-accelerating capex” claim are in tension; the reconciliation is government grants and customer prepayments lowering GFS’s net cash outlay.

The FY24 GAAP loss — a one-time, non-cash item. FY24’s headline GAAP net loss of −$265M despite $728M of operating income is a trap for the unwary: it was driven by a $935M impairment of legacy production-capacity assets at Fab 8 (Malta, NY) — the accounting crystallization of the 2021–22 over-build, a PwC critical audit matter, non-cash, and to be normalized out of run-rate (OCF and D&A were unaffected). Adjusted/non-IFRS earnings were positive in FY24; the GAAP loss is not evidence of an operating problem, it is evidence of a past capital-allocation problem.

Returns. ROIC of ~5.8% (FY25), ~8.4% (FY23) and ~10.2% at the 2022 peak — mediocre, hovering around or below a reasonable WACC for a capital-intensive foundry, and negative in FY24. Return on equity is similarly unimpressive once the $935M charge is included. This is the quantitative core of the “narrow moat / price-taker” verdict.

Balance sheet. Strong and a genuine asset. Total equity ~$11.98B (FY25); cash and short-term investments ~$3.10B against total debt (incl. leases) ~$1.71B, i.e., a net cash position (~$1.4B including ST investments). Gross PP&E $33.8B against $26.0B accumulated depreciation leaves net PP&E of $7.8B — a heavily depreciated base, which both flatters near-term FCF (low capex sustainable for a while) and signals a future re-investment need. A $12.4B accumulated deficit on the balance sheet is a legacy of the pre-2022 AMD-carveout losses, not current operations. Liquidity (current ratio ~2.6×) and leverage are not concerns. Dilution is modest: shares crept from ~532M (2021) to ~556M (2025), with SBC ~$186–200M/yr (~3% of revenue) — partly now offset by the new buyback.

Verdict: economics that are stable and cash-generative but structurally mediocre, and that do not clearly improve with scale. The balance sheet is a real strength and FCF is positive — but the FCF is partly a function of under-spending depreciation, margins have compressed, and ROIC sits below the cost of capital. There is no evidence here of the scale-driven margin expansion that defines a high-quality foundry.

7. Capital Allocation

The capex arc — a Marathon case study. GFS over-built at the 2021–22 cycle top (funded partly by customer prepayments and government grants), then impaired $935M of that build in 2024 and cut capex 79% into the downturn. That is, in sequence, classic capital-cycle value destruction: high returns and shortage-era demand drew capital into mature-node capacity industry-wide; GFS added supply at peak prices; demand normalized; utilization and ASPs fell; the impairment crystallized the over-investment; discipline came after the damage. The redeeming features: much of the build was government- and customer-funded (capping GFS’s own cash at risk), and GFS cut faster and harder than peers once demand rolled over.

M&A — a coherent up-the-value-chain pivot, with caveats. Despite the optics of six deals in 18 months, there is a clear through-line — AI connectivity + differentiated power + processor IP: silicon photonics (Advanced Micro Foundry + InfiniLink), processor IP (MIPS + Synopsys ARC), and power GaN (Tagore + the TSMC license). The strategic logic is sound: move from a pure capacity-lessor (sell wafers, earn on utilization) toward selling IP and differentiated technology at accretive gross margin — exactly the right response to being a mature-node price-taker. The caveats are real: (1) every deal price is undisclosed, which prevents any outside ROIC discipline check on a now-serial acquirer; (2) the TSMC GaN license-in signals GFS buys/licenses capability rather than generating it; (3) goodwill is building (~$407M added in FY25) on a balance sheet that already carries a $12.4B accumulated deficit and just took a $935M impairment — future writedown risk if the pivot underdelivers; (4) several targets are small teams/startups, carrying integration and retention risk rather than synergy-driven scale.

Government funding. Up to $1.5B in U.S. CHIPS Act direct funding (Malta NY + Burlington VT) plus eligibility for the 25% CHIPS investment tax credit; NY Green CHIPS and EU Chips Act (Dresden) and Singapore incentives. The $16B U.S. plan (4-Jun-2025) = $13B previously committed + $1.5B CHIPS + $3B new (advanced packaging/photonics/GaN). Government grant proceeds were ~$148M in FY25 with ~$261M of deferred grant income on the balance sheet. This materially de-risks the next build — but it also makes GFS a policy-dependent business whose build economics rely on subsidies continuing.

Shareholder returns — newly initiated, and Mubadala-linked. No dividend (never paid; correct, given the accumulated deficit and reinvestment runway). The first capital return came in 2026: a $500M buyback authorization (Feb-2026), of which $400M was executed in Q1-2026 — but ~$300M of that was a direct repurchase of Mubadala shares (7.34M shares at $40.845) concurrent with the March-2026 $840M Mubadala secondary (20M shares at $42.00). The buyback is accretive (bought below the public clearing price) but is as much a sponsor-liquidity mechanism as an unconstrained value-driven return.

Comp and governance. A controlled company and FPI: Mubadala (~81% at the FY25 20-F, falling) holds tiered consent and board-nomination rights, GFS is exempt from several Nasdaq governance requirements, and comp is disclosed only in aggregate (FY25 aggregate executive+director comp ~$68.2M). PSUs vest on revenue + adjusted-FCF-margin + absolute TSR — reasonable and cash-skewed, but absolute (not relative) TSR can pay out in a sector-wide bull market without outperformance, and a prior ROIC performance threshold was loosened in a 2023 PSU modification — a yellow flag precisely because ROIC is the metric most relevant to this business’s central problem. CEO Tim Breen (effective Apr-2025) was seconded from Mubadala — a sponsor-linked appointment.

Verdict: a poor historical record giving way to more intelligent, defensive allocation — but the proof is pending. Building at the peak, impairing $935M, and running a $12.4B accumulated deficit is a negative historical verdict. The forward setup is better: government/customer-funded capacity, disciplined capex cuts, positive FCF, a coherent IP/connectivity pivot, and the first capital return. The reservations: undisclosed deal prices, building goodwill, GaN licensed not built, a loosened ROIC incentive, and buybacks doubling as a sponsor exit ramp.

8. Changes and Headwinds — Last Two Years

Leadership. A full C-suite transition: Tim Breen (ex-Mubadala/McKinsey) became CEO effective 28-Apr-2025; founder-era CEO Thomas Caulfield moved to Executive Chairman; Niels Anderskouv became President & COO; Sam Franklin is CFO. Mubadala’s fingerprints are on the appointment (Breen’s secondment), reinforcing the controlled-company character.

Strategic pivot. The defining change is the shift from “build wafer capacity” to “move up the value chain”: the MIPS, Advanced Micro Foundry, InfiniLink and Synopsys ARC acquisitions, the Tagore GaN buy and TSMC GaN license, and the explicit $1B-photonics-by-2028 and “Technology Services” framing. The $16B U.S. investment plan (Jun-2025) and €1.1B Dresden expansion anchor the onshoring leg.

Ownership. The persistent Mubadala selldown is the dominant corporate event stream: ~88–89% at IPO → ~81% (FY25 20-F) → further reduced by the March-2026 $840M secondary and a ~$1.91B block reportedly offered in May-2026. Each secondary prices at a discount and caps the float — a structural overhang on the stock independent of fundamentals. Insider Form 4 activity is otherwise routine (10b5-1 sales, tax withholding; no open-market purchases).

Cyclical/financial. The auto/industrial/IoT inventory correction bottomed in 2024; the $935M Fab 8 impairment landed in Q4-2024; LTA underutilization payments wound down through early 2025; and the 2025–26 prints showed a volume-led recovery with stabilizing pricing and a mix shift toward auto and datacenter. The 2026 mature-node pricing inflection (Chinese players and VIS raising prices; TSMC closing legacy fabs) is a recent tailwind.

Verdict: net thesis-neutral-to-modestly-strengthening on fundamentals, offset by the valuation and overhang. The mix shift, photonics traction, pricing inflection and capital return strengthen the operating story; the Mubadala overhang, undisclosed-price M&A and policy dependence are the offsets. None of it justifies the 93rd-percentile valuation by itself.

9. Risk Analysis

Catastrophic-loss assessment. Low. Net cash, positive FCF, a government-backed and DoD-accredited strategic asset, and obvious strategic/takeout value (a scaled non-China/non-Taiwan foundry) make a permanent capital wipeout remote. The dominant risk is price — a high-beta de-rating from a rich multiple — not solvency.

10. Valuation Discussion (Embedded Expectations)

Where the multiple sits. At ~$85.83 (18-Jun-2026) and ~555.9M shares, market cap is ~$47.7B; with ~$1.4B net cash (incl. ST investments), EV is ~$46.3B. Against FY25 results that is ~22× EV/EBITDA ($2.11B), ~6.8× EV/sales ($6.79B), ~54× GAAP P/E ($1.59 EPS) and ~46× P/FCF (~$1.0B). On a forward non-IFRS basis (Q1-2026 non-IFRS EPS $0.40; Q2 guide ~$1.76B revenue; plausible FY26 non-IFRS EPS ~$1.90–2.20), the forward non-IFRS P/E is ~40–45×. The own-valuation-history percentiles are the cleanest single read: P/S 99th, P/B 92nd, P/E 87th, composite 93rd — the richest the stock has been across its entire ~4.5-year public life.

The comp that matters. On a quality-adjusted basis GFS is more expensive than TSMC. TSMC trades around ~20–25× earnings with a ~60% gross margin, 30%+ ROE and structural growth; GFS trades at ~54× GAAP / ~22× EV/EBITDA with a 25% gross margin and ~6% ROIC. Against the analog/mature-node cohort that the factor model identifies as its nearest peers (MCHP, ON, LSCC, NXPI, TXN), GFS carries a premium P/E despite comparable-or-weaker returns; on EV/EBITDA the gap narrows but GFS is at the rich end. The valuation is not anchored to current economics; it is anchored to a recovery-and-re-rating story.

Embedded expectations (what the price requires). To justify a ~$46B EV at a “fair” 12–14× EV/EBITDA for a sub-WACC, cyclical specialty foundry, EBITDA would need to reach roughly $3.3–3.9B — i.e., +56% to +85% over FY25’s $2.11B, and above the 2022 cycle peak of $2.88B. The market is therefore underwriting (1) a revenue recovery well above the prior peak (toward $8–9B+), (2) gross margin durably back above 30% (management’s own 2026 exit target, not yet proven), (3) the silicon-photonics/auto/IP mix ramping on schedule, and (4) the current re-rating holding rather than mean-reverting. A reverse-DCF on ~$1.0B FCF growing into the photonics ramp similarly requires sustained mid-teens FCF growth for a decade at a ~10% discount rate — a lot must go right, in sequence, with no cyclical air-pocket.

Scenario sketch (illustrative; not a price target).

• Bear: China price war resumes / auto-inventory rollover; EBITDA resets toward ~$1.6B; the market applies a sober ~10–12× → EV ~$16–19B, well below today — the high-beta round-trip the −55% prior drawdown warns of.
• Base: Margins recover to ~30%, revenue grinds to ~$7.5–8B, EBITDA ~$2.4–2.7B; a 12–15× multiple → EV ~$29–40B — i.e., today’s price already discounts most of a successful base case.
• Bull: Photonics hits the $1B run-rate, GM exceeds 32–34%, ROIC clears WACC, onshoring fills the new fabs; EBITDA ~$3.5–4B at a 14–16× re-rated multiple → EV ~$49–64B — upside exists, but only if essentially everything works.

Verdict (embedded expectations): the market is pricing the bull case as the base case. The reshoring optionality, the photonics ramp and the government-funded build are fully paid for at today’s price, not free.

11. Variant Perception

Consensus view. GFS is the Western world’s strategic foundry — the onshoring/AI-connectivity winner with Trusted-Foundry status, a record automotive book, a credible silicon-photonics datacenter story, government backing, and improving FCF — and it deserves a premium re-rating. The price action (a 3× run, all-time highs, +131% one-year) embodies this consensus.

Strongest bull case. The geopolitical moat is genuinely un-replicable and structurally under-monetized; the mix shift toward auto and datacenter is real and accelerating; the photonics franchise is a true AI-era growth vector with revenue already compounding (+29%); the build is funded by taxpayers and customers, not GFS’s balance sheet; the IP/connectivity pivot moves the business up the value chain to higher-margin revenue; net cash and positive FCF give optionality; and a tightening mature-node supply backdrop (Chinese price hikes, TSMC legacy-fab closures) supports pricing. If margins recover to 30%+ and the pivot earns its capital, GFS re-rates into the multiple.

Strongest bear case. This is a ~6%-ROIC, 25%-gross-margin price-taker priced like a 60%-margin compounder, at the 93rd percentile of its own valuation, near all-time highs, after a sector-beta-driven run with negative idiosyncratic alpha — i.e., it went up with the AI-semis trade, not because of company-specific quality. The moat is too narrow to confer pricing power (proven by 2025’s forced concessions); the FCF is partly manufactured by under-spending depreciation; the M&A is top-of-cycle, undisclosed-price, goodwill-building bolt-ons with a license-in tell; and a relentless Mubadala selldown caps the upside. When the semis factor turns, a β-1.8 name with a −55% prior drawdown gives it back violently.

The 3–5 assumptions that matter most:

1. Does gross margin durably exceed ~30–34%? (Management’s exit-2026 target; the swing factor between the bear and bull EBITDA paths.)
2. Does ROIC clear WACC (~10%)? The single test of whether the moat is economically real or just narrative.
3. Does the silicon-photonics franchise hit the ~$1B run-rate by 2028? The load-bearing growth assumption.
4. Does China re-flood mature nodes, or does the 2026 pricing inflection hold? The dominant external swing factor.
5. How fast does Mubadala finish selling, and at what price? The overhang’s duration sets the near-term ceiling.

Falsification. Bull falsified if: gross margin stalls in the mid-20s, ROIC stays ~6%, photonics revenue disappoints, or a China price war resumes — the multiple then compresses toward the cohort’s ~10–14× EV/EBITDA. Bear falsified if: GFS prints two-plus quarters of 30%+ gross margin with ROIC trending above WACC and the photonics/auto mix demonstrably re-rates the revenue base — at which point the premium is earned.

Factor-positioning read (overlay). The factor model places GFS firmly in the high-beta AI-semis cohort: β≈1.8, Market loading +1.27, Technology +0.84, Semiconductors +0.82, a negative LowVolatility loading (−0.58), Momentum/Value/Growth zeroed, and negative alpha (−0.25). The leaderboard shows a +131% one-year (Sharpe 2.37) sitting on top of a three-year record of only +11%/yr (Sharpe 0.19) and a −55% max drawdown — the entire risk-adjusted return is the last ~9 months. Nearest factor peers are all semis vehicles (FTXL, CHPS, SMHX) and the analog/mature cohort (MCHP, ON, LSCC, DIOD). This is evidence the consensus is a crowded momentum/sector-beta trade near its richest-ever multiple — exactly the configuration in which positioning, not fundamentals, sets the near-term risk.

12. Fact vs. Interpretation

13. Open Questions

1. Utilization. GFS does not disclose a clean blended fab-utilization figure; the recovery is inferred from “oversubscribed corridors” and the 30%+ GM exit target. What is actual blended utilization?
2. M&A prices. Every recent deal (MIPS, AMF, InfiniLink, Tagore, Synopsys ARC) is undisclosed — what were the prices and implied multiples, and what is the combined annual revenue contribution?
3. Photonics revenue today. How much of the $745M Comms/datacenter line is silicon photonics vs. SATCOM/SiGe, and what is the current photonics run-rate against the $1B-by-2028 target?
4. Incremental ROIC. What return does management underwrite on the $16B U.S. and €1.1B Dresden capacity, net of grants?
5. Mubadala endgame. What is the target ownership and timeline? When does the overhang clear?
6. Normalized earnings power. What is mid-cycle revenue/EBITDA once LTA payments are fully gone and the new fabs fill?

14. What Must Be True

Bull case — what must be true:

• Gross margin recovers durably above ~30–34% (management’s 2026 exit target), proving operating leverage and pricing stabilization.
• ROIC clears WACC (~10%), demonstrating the moat is economically real, not just narrative.
• The silicon-photonics/optical franchise reaches a ~$1B run-rate by 2028 and the auto/datacenter mix re-rates the revenue base above the 2022 peak.
• The 2026 mature-node pricing inflection holds (no renewed China price war), and the Mubadala overhang clears without crushing the price.
• Falsification test: if, over the next 3–4 quarters, gross margin stalls in the mid-20s and ROIC stays ~6% while photonics revenue disappoints, the bull thesis is broken and the 93rd-percentile multiple compresses.

Bear case — what must be true:

• The moat stays too narrow to earn its capital; price concessions recur as China re-floods mature nodes; ASP erosion continues.
• The AI-connectivity pivot underdelivers (photonics slips, RISC-V stays narrative), goodwill is impaired, and capex re-acceleration dilutes returns.
• A cyclical air-pocket (auto/industrial rollover) resets EBITDA toward the low-$1.5B’s while the multiple is still 20×+, triggering the high-beta round-trip.
• Falsification test: if GFS prints two-plus consecutive quarters of 30%+ gross margin with ROIC trending above WACC and a demonstrably re-rated revenue mix, the bear thesis is broken and the premium is earned.

15. Source Appendix

See the Source Appendix below for full citations. Primary sources: GlobalFoundries FY2025 Form 20-F (filed 2026-02-27), FY2024 Form 20-F (filed 2025-03-20, $935M impairment), and Q4-2025 (2026-02-11) and Q1-2026 (2026-05-05) earnings calls; SEC EDGAR XBRL and Form 4 corpus (CIK 0001709048); public market price history; and publicly disclosed TSMC figures for foundry-industry cross-read.

APPENDIX A — Standard Diligence Questionnaire

Supplemental to the research memo. Fact / Interpretation / Assumption labels applied where it matters. As of 2026-06-19.

General

What thoughtful questions have other investors asked about this company? The recurring institutional debates: (1) Is GFS a structurally inferior price-taker permanently capped at ~25% gross margin and ~6% ROIC, or can the specialty/IP mix lift it toward 30%+ and above-WACC returns? (2) Is the silicon-photonics/AI-connectivity pivot real revenue or a narrative bolted on at the top of the AI cycle? (3) How much of the apparent FCF is sustainable vs. a function of under-spending depreciation? (4) When does the Mubadala overhang clear, and at what price? (5) Does the geopolitical/Trusted-Foundry moat justify a structural premium, or is it already over-priced at the 93rd valuation percentile? (6) Is GFS ultimately a strategic takeout candidate (a scaled non-China/non-Taiwan foundry)? (Interpretation.)

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Off the trough but not at a peak. Revenue ($6.79B FY25) is ~16% below the 2022 peak ($8.11B); gross margin (24.9%) is compressed vs. 28% in 2022–23. EBITDA $2.11B is below the $2.88B 2022 peak. So earnings are recovering from a cyclical low, not at a high — but the stock price is at a high. (Fact.)

Driven by external environment or internal actions? Both. External: the auto/industrial/IoT inventory cycle and China-distorted mature-node pricing. Internal: deliberate price concessions to dual-sourced mobile customers, the capex cut, and the mix shift toward auto/datacenter. (Fact + Interpretation.)

How stable are revenues? Cyclical. Foundry volume and ASP both swing with end-market inventory; the 2022→2024 −17% peak-to-trough is the evidence. The single-source design-win base and ~11% non-wafer revenue add some stability. (Fact.)

Outlook for products/services? Mix improving: automotive (record) and communications/datacenter (+29%) growing; smart mobile declining; silicon photonics/GaN/RISC-V the forward bets. (Fact for history; Interpretation for outlook.)

How big will this market be? The specialty/mature foundry segment grows roughly with semiconductor units and content (low-to-mid single digits secularly), with AI-connectivity (optical/photonics) and onshoring as above-trend pockets. Global and capital-intensive. (Interpretation.)

Business Quality & Competitive Moat

Is the industry getting more or less competitive? Generic mature nodes: more competitive (Chinese subsidized capacity). Differentiated specialty niches (FD-SOI, RF, auto/defense, photonics): less competitive due to qualification barriers. (Interpretation.)

How profitable is the business (ROIC, ROE)? Mediocre. ROIC ~5.8% FY25 (below WACC), ~10.2% peak FY22; ROE depressed by the FY24 impairment. Gross margin 24.9%. (Fact.)

How profitable is the industry / barriers to entry? Bifurcated: leading edge is extraordinarily profitable and near-impenetrable (TSMC); mature/specialty is mediocre with low barriers in generic nodes and moderate barriers in specialty niches. (Interpretation.)

Can the business be easily understood? Yes — a capital-intensive contract manufacturer of differentiated chips, paid per wafer/service. (Fact.)

Can it be undermined by foreign low-cost labor? The relevant threat is subsidized capital (Chinese state-backed fabs), not labor. Foundry is capital- and process-intensive, not labor-intensive. (Interpretation.)

Do brands matter? Nature of competition? Switching costs? “Brand” matters as reputation/qualification/Trusted-Foundry status, not consumer brand. Competition is on process differentiation, qualification, capacity reliability and geography. Switching costs are real on single-sourced specialty designs (multi-year, multi-$M requalification) and weak on dual-sourced commodity wafers. (Fact + Interpretation.)

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The Trusted-Foundry accreditation, the FD-SOI ecosystem position, and government-relationship optionality are valuable but not on the balance sheet. The heavily depreciated fab base (net PP&E $7.8B on $33.8B gross) understates replacement value. (Interpretation.)

Off-balance-sheet liabilities? Operating/capital lease obligations are on-balance-sheet (IFRS 16); LTA capacity commitments and capex purchase commitments exist; deferred government-grant income (~$261M) is a liability tied to milestone conditions. No unusual off-balance-sheet exposure identified. (Fact.)

How conservative is the accounting? Mixed. The $935M Fab 8 impairment was a timely, conservative recognition of the over-build. Against that: management leans heavily on non-IFRS metrics, the LTA underutilization payments flattered prior-period ASPs, and “adjusted FCF” framing should be checked against the under-spend of depreciation. (Fact + Interpretation.)

How CapEx-hungry is the business? Very. Capex peaked at $3.06B (2022); the current ~$0.7B is below depreciation ($1.31B) and not a steady state — FY26 net capex guided back to 15–20% of revenue. Government grants and customer prepayments lower GFS’s net outlay. (Fact.)

Capital Allocation & Management

How much FCF, and how is it used? ~$1.0B FCF FY24–25 (partly via low capex). Uses: targeted growth capex, IP/connectivity M&A, and the new buyback (which partly absorbs Mubadala selling). No dividend. (Fact.)

Significant acquisitions recently? Yes — MIPS (RISC-V, Aug-2025), Advanced Micro Foundry (silicon photonics, Nov-2025), InfiniLink (optical, Nov-2025), Tagore (GaN, 2024), TSMC GaN license (Nov-2025), Synopsys ARC IP (closing ~H1-2026). Coherent AI-connectivity/IP pivot; all prices undisclosed; goodwill +$407M FY25. (Fact + Interpretation.)

Buying back shares? Yes, newly: $500M authorization (Feb-2026), $400M executed Q1-2026 (~$300M of it a direct repurchase of Mubadala shares). (Fact.)

Issuing large amounts of stock to insiders? SBC ~$186–200M/yr (~3% of revenue); share count crept 532M→556M (2021→2025). Modest, partly offset by the buyback. (Fact.)

Compensation policy. Controlled-company/FPI aggregate disclosure (~$68.2M FY25 for all execs+directors). PSUs on revenue + adjusted-FCF-margin + absolute TSR; a prior ROIC threshold was loosened in a 2023 modification (yellow flag). (Fact + Interpretation.)

Motivations of management. CEO Tim Breen is Mubadala-linked (seconded from Mubadala). Capital-return and secondary decisions sit at the sponsor/minority intersection — watch for accretion to all holders. (Interpretation.)

Valuation & Market Data

ADR, MLP, or K-1 issuer? None of these. GFS is a Cayman-incorporated foreign private issuer trading ordinary shares directly on Nasdaq (files 20-F/6-K, IFRS). Not an ADR, not an MLP, no K-1. (Fact.)

Dividend policy? No dividend; none expected near-term given reinvestment runway and accumulated deficit. (Fact.)

How profitable is the business? See above — gross margin 24.9%, operating margin 11.7%, ROIC ~6%; structurally mediocre. (Fact.)

Net income vs. cash from operations diverging? Yes, and in GFS’s favor on cash: OCF ($1.73B FY25) far exceeds GAAP net income ($885M) due to heavy D&A ($1.31B) — normal for a foundry. The FY24 GAAP loss vs. positive OCF was driven by the non-cash impairment. (Fact.)

Risks & Downside

What factors would cause the stock to decline? A valuation de-rating from the 93rd percentile (highest-likelihood path given β≈1.8 and a −55% prior drawdown); a China mature-node price war; an auto/industrial cyclical rollover; photonics/pivot disappointment; renewed Mubadala secondaries; a semis-factor reversal. (Interpretation.)

Risk of catastrophic loss? Low. Net cash, positive FCF, government-backed, strategically important, with takeout-floor optionality. (Interpretation.)

Chance of total loss? Remote. The dominant risk is price (multiple compression), not solvency. (Interpretation.)

Recent News & Events

Has the business environment changed recently? Yes — favorably at the margin: a 2026 mature-node pricing inflection (Chinese players/VIS raising prices, TSMC closing legacy fabs), an accelerating auto/datacenter mix, and a recognized AI-connectivity narrative. The news tape itself is quiet/sector-driven (no GFS-specific catalyst in the recent window). (Fact + Interpretation.)

Significant acquisitions / accounting changes / new markets? The IP/connectivity M&A spree (above); the $16B U.S. and €1.1B Dresden expansions; the LTA underutilization-payment wind-down (a revenue-recognition normalization); the C-suite transition (Breen CEO, Apr-2025). (Fact.)

APPENDIX B — Source Appendix

Primary sources prioritized. As of 2026-06-19. GFS is a foreign private issuer: annual report = Form 20-F; interim/material events = Form 6-K (no 10-K/10-Q/DEF 14A).

Primary — SEC Filings (EDGAR, CIK 0001709048)

1. Form 20-F, FY2025 — filed 2026-02-27. End-market revenue table; wafer volume; top-10 customer concentration (63%/65%/72% FY25/24/23); LTA/capacity-reservation & prepayment disclosure; competition section (TSMC/UMC/SMIC/Samsung/Intel/TI); China mature-node risk; Trusted Foundry; specialty-platform portfolio; Mubadala ~81% beneficial ownership and Shareholder’s Agreement; CHIPS $1.5B; government-grant balances; capex; goodwill ($407M); PSU/comp metrics; related-party.
2. Form 20-F, FY2024 — filed 2025-03-20. $935M Fab 8 (Malta NY) long-lived-asset impairment (MD&A + Notes; PwC critical audit matter); restructuring; FY24 GAAP net loss −$265M vs. $728M operating income.
3. Form 20-F, FY2023 / FY2022 / FY2021 — filed 2024-04-29 / 2023-04-14 / 2022-03-31. Multi-year revenue, margin, capex, customer-concentration history; IPO-era disclosures.
4. Form 6-K corpus — earnings releases; the 5-Feb-2025 leadership-transition 6-K (Breen CEO eff. 28-Apr-2025; Caulfield → Executive Chairman; Anderskouv President & COO); the 4-Jun-2025 $16B U.S. investment plan; M&A announcements; CHIPS Direct Funding Agreement (Nov-2024); the 5-May-2026 Q1 + Investor Day 6-K.
5. Form 4 / Form 3 corpus — insider transactions: routine 10b5-1 sales (code S) and tax-withholding (code F); no open-market purchases (code P) identified. SEC EDGAR (CIK 1709048).
6. Form 424B7 prospectuses — Mubadala secondary offerings (incl. May-2024; March-2026 $840M, 20M shares @ $42.00 with concurrent ~$300M GFS buyback at $40.845). SEC EDGAR.
7. SEC EDGAR XBRL — common shares outstanding (532M→556M, 2021–2025); revenue/income concepts cross-checked to the 20-Fs.

Primary — Earnings Calls

8. Q1-2026 earnings call — 2026-05-05. Non-IFRS EPS $0.40 (beat $0.34); Q2 guide ~$1.76B; FY26 net capex 15–20% of revenue; LTA underutilization payments “largely in the rearview mirror”; $400M of $500M buyback executed; gross-margin exit-2026 ≥30%; CPO/OCI design wins; Synopsys ARC referenced.
9. Q4-2025 earnings call — 2026-02-11. FY25 figures; ~2.3M wafers / ~85% utilization; 500+ design wins (95%+ sole-source); $1B silicon-photonics run-rate by end-2028; $16B U.S. + €1.1B Dresden; mobile price-concession explanation; MIPS/AMF/InfiniLink/TSMC-GaN.

Quantitative / Market Data

10. Company financial statements (2020–2025) — income statement, balance sheet, cash flow, per-share data, profitability ratios, enterprise value; reconciled to the 20-Fs. (Aggregator-reported market cap was erroneous — reconciled to ~556M shares × price.)
11. Own-history valuation percentiles (as of 2026-06-18) — P/E 87th, P/B 92nd, P/S 99th, composite 93rd; price $85.83; ttm EPS $1.39; book $20.84/sh; P/B 4.12×.
12. Public market price history (through 2026-06-18) — the five-year event-map price points (IPO $47; all-time-low close $30.37 8-Apr-2025; all-time-high close $89.96 26-May-2026).
13. Factor/risk model (2026-06-18/19) — loadings (β Market +1.27, Tech +0.84, Semis +0.82, LowVol −0.58, Quality +0.39; R²~0.50), risk-adjusted track record (y1 +131%/Sharpe 2.37; y3 +11%/Sharpe 0.19/maxDD −55.4%), beta 1.81 / alpha −0.25 / rs_12m 134.5, factor-similar peers (FTXL/CHPS/SMHX/MCHP/ON/LSCC/DIOD).

Secondary — Industry / Trade Press / Company

14. Publicly disclosed TSMC figures — foundry market-share data (TSMC ~70%, GFS ~3.9%, UMC ~4.4%, SMIC), leading-edge vs. mature bifurcation, China overcapacity, TSMC ~60% gross margin / ~30%+ ROE benchmark.
15. TrendForce / DigiTimes / EE Times (Dec-2025–May-2026) — China mature-node ~10% price hikes; TSMC legacy-fab closures (Fabs 2/5 by end-2027); 2024 mature-node utilization ~70%, op-profit −23%.
16. Company press releases & GlobeNewswire — $16B U.S. plan (2025-06-04); MIPS (Aug-2025); Advanced Micro Foundry & InfiniLink (Nov-2025); Tagore (2024-07-01); TSMC GaN license (semiconductor-today.com, 2025-11-11); March-2026 secondary/buyback; ~$1.91B May-2026 block (Bloomberg, 2026-05-26).

Analytical Frameworks

17. Competition Demystified (Greenwald & Kahn) — moat-type taxonomy (GFS = narrow demand/captivity switching costs, no scale advantage), ROIC/share-stability tests. Capital Returns (Marathon) — the capex-at-the-top → impairment → cut capital-cycle read; asset-growth-into-low-ROIC caution.