GFS Risk Register — Eight Categories

This register codifies the structured-risk view that complements bear case (narrative form) and bull case (positive thesis). Each risk is identified by category prefix (G = Geopolitical, C = Customer concentration, K = Capital structure / Mubadala, T = Technology, R = Regulatory, X = Competitive, M = Market / Macro, S = Supply chain). Likelihood and Impact are scored 1-3 (1 = low, 2 = medium, 3 = high). The detection trigger field describes the public signal an analyst would observe if the risk were materializing. The mitigation field summarizes management’s commentary or structural defenses where applicable.

The register is calibrated against today’s date (2026-04-29), GFS’s current operating posture (FY 2025 revenue $6.79B, net income $888M, ~$200M FY 2025 silicon-photonics revenue, AMF acquisition closed Nov 17 2025, InfiniLink acquired CY 2025, CHIPS Act $1.5B award + $550M New York State support, Mubadala 77.05% stake post-March 2026 sell-down), and a spot price of $59.49 (close 2026-04-28, STOCK_PRICE_DATA.json ✓).

G — Geopolitical risks

ID	Risk	L	I	Detection trigger	Mitigation / management commentary
G1	Taiwan-China military escalation impacting Asian foundry supply chain. A Taiwan-China conflict event would compress global semiconductor supply with knock-on effects across all merchant foundries. GFS’s Singapore + Dresden + US footprint is partially insulated but global-customer-pipeline exposure to Taiwan-domiciled supply chain is real.	1	3	Increased PLA exercises around Taiwan; global-supply-chain-impact analyst commentary; defense-contractor commentary on Taiwan-Strait scenario.	GFS Singapore + Malta + Dresden + Vermont fab portfolio is the explicit hedge. CHIPS Act subsidy is the US-domestic-foundry alignment move. The 2026 acquisition of AMF Singapore further diversifies geographic exposure.
G2	BIS export-control rule expansion further compressing China addressable market. Successive BIS rules (Oct 2022, Oct 2023, Dec 2024) have progressively restricted advanced-node + AI-related semiconductor exports to China. Photonics applications + Chinese hyperscaler customer pipelines are gated.	3	2	New BIS Federal Register notices adding photonics or 28nm-65nm mature-node exports to the Commerce Control List; Entity-Listing of additional Chinese hyperscalers / module OEMs.	GFS’s mature-node + photonics mix is less directly affected than advanced-node-leading foundries. Management commentary on Q4 2025 cited “mid-single-digit” addressable-market compression to date — implies the risk is partly priced in already.
G3	EU industrial-policy shifts compress Dresden expansion economics. EU Chips Act subsidies have been politically contested through CY 2025-2026. German federal-state political uncertainty could delay Dresden permitting / construction milestones.	2	2	German coalition formation outcomes; EU Chips Act funding-cycle reauthorization debates; specific Dresden-project milestone disclosures in 10-K.	Multi-jurisdictional fab footprint reduces single-jurisdiction exposure. Fallback option: shift planned Dresden expansion capacity to Singapore or Malta.
G4	US-domestic-foundry posture shift (post-2028 administration trajectory). CHIPS Act 2.0 / sequel-program negotiations could materially shift the subsidy regime. A future administration could either (a) accelerate subsidies for US-domestic foundries (positive) or (b) pivot to tariff-only regimes that reduce direct subsidy support (neutral-to-negative).	2	2	Bipartisan Senate negotiations on follow-on CHIPS funding; House Energy & Commerce / Science Committee hearings; Department of Commerce policy guidance updates.	Mubadala’s UAE-domiciled control + GFS’s deeply-embedded US-domestic-fab presence make GFS structurally well-positioned regardless of administration trajectory. The 77% Mubadala stake is a manageable diplomatic factor (not a CFIUS-risk factor for incremental investment).
G5	Singapore-region geopolitical disruption (post-AMF acquisition). AMF integration concentrates more of GFS’s photonics-fab capacity in Singapore. A regional disruption (Malaysia-Singapore relations, Indonesia-region political event, broader Indo-Pacific instability) would compress AMF integration timeline.	1	2	Regional defense-policy escalations; Singapore-government policy guidance on foreign-investment / industrial-policy alignment with US-allied posture.	Singapore is structurally one of the most US-allied, supply-chain-secure locations in the Asia-Pacific. Risk is genuinely tail.

C — Customer concentration risks

ID	Risk	L	I	Detection trigger	Mitigation / management commentary
C1	Hyperscaler customer concentration in the SiPh segment. Although the GF SiPh page lists 8+ named customers (Marvell, Broadcom, Cisco, NVIDIA, Ayar Labs, Lightmatter, PsiQuantum, Ranovus + AMF customer-base post-acquisition), the revenue concentration is structurally heavy on a small number of hyperscaler-tier customers. Loss or material reduction of any single Tier-1 customer (Marvell, Broadcom, NVIDIA) would materially affect the $1B end-2028 trajectory.	2	3	Q1-Q4 2026 10-K segment-revenue disclosures show single-customer >10%; named customer (Marvell, Broadcom, NVIDIA) earnings calls cite migration to alternate foundry.	Multi-customer roster with ongoing PDK qualifications (LWLG, NLM, NLM-equivalents) expands the customer-engagement pipeline. AMF acquisition broadens the customer base outside the hyperscaler set.
C2	Hyperscaler in-house silicon-photonics ramp shifts volume away from Fotonix. NVIDIA’s announced SiPh program; Broadcom Tomahawk-Bailly transition risks; Marvell Celestial AI + Polariton end-to-end stack. If hyperscaler programs achieve technical-readiness for in-house production by 2027-2028, merchant-SiPh share is structurally compressed.	2	3	Hyperscaler Q3-Q4 CY 2026 calls cite SiPh production ramp on internal capacity; OFC 2027 / OFC 2028 disclosures detail in-house programs; module-OEM commentary cites multi-source qualification with hyperscaler-internal options.	Fotonix’s customer roster is structurally diversified beyond the hyperscalers (8+ named customers including the CPO startups). Fotonix has 5+ years of yield-learning advantage that internal hyperscaler programs cannot quickly replicate.
C3	Apple / Qualcomm / Tesla legacy customer migration. GF’s mature-node customer base includes Apple-related programs, Qualcomm RF, and automotive primes (potentially Tesla, Stellantis, GM via Tier-1 suppliers). Migration of any single high-profile customer relationship to TSMC / Samsung / SMIC at lower nodes erodes both revenue and signaling.	2	2	10-K customer-concentration footnote changes; competitor (TSMC, SMIC) press releases citing GF-customer wins; mature-node-segment revenue line softens persistently.	95%+ sole-sourced design-win positioning per FY 2025 disclosure (Q4 2025 transcript ✓) provides multi-year revenue lock-in for design-win cohorts.
C4	Module-OEM customer (Innolight, Eoptolink, Coherent, Lumentum, Hisense Broadband) consolidation or vertical integration. The module-OEM layer that buys Fotonix-based components is structurally fragmented but consolidating; vertical integration toward in-house SiPh (some Chinese module OEMs are exploring this) compresses Fotonix demand at the module-component level.	2	2	Module-OEM annual reports cite captive SiPh capacity buildouts; module-OEM consolidation events (M&A); Fotonix-revenue per module-OEM declines.	GF’s hyperscaler-tier customer base is one architectural layer above module OEMs and is structurally less affected by module-layer consolidation.
C5	AMF customer-book attrition during integration cycle. Post-acquisition customer-retention risk: AMF’s existing customer book may include customers who chose AMF specifically because it was independent of GF; integration could trigger 10-25% customer-base attrition.	2	1	FY 2026 SiPh-segment revenue disclosures show net post-AMF revenue contribution materially below the implied AMF stand-alone revenue baseline; AMF customer-side press releases cite alternate-foundry qualifications.	GF-AMF Center of Excellence with A*STAR creates ongoing customer-engagement support framework. AMF customer-base profile (long-haul, LiDAR, sensing) is largely complementary rather than competitive to existing GF customers.

K — Capital structure / Mubadala risks

ID	Risk	L	I	Detection trigger	Mitigation / management commentary
K1	Mubadala 77.05% stake recurring selldown technical pressure. Pattern: ~12-18 months between selldown events. May 2024 ($950M), March 2026 ($840M). Forward expectation: 4-6 more events to reach <50% non-controlling. Each event drives -3% to -7% one-day downside + 30-day drift recovery.	3	2	Underwriter-launch announcements; SEC SC 13G/A amendments; pricing 8-Ks.	Concurrent share-buyback program absorbs ~30-40% of each selldown. Mubadala’s orderly-cadence posture is well-established.
K2	CHIPS Act subsidy clawback risk. $1.5B CHIPS Act direct funding includes capex-deployment-milestone covenants. Failure to meet announced fab-construction or job-creation milestones could trigger clawback.	1	3	NIST / Department of Commerce notices on milestone-tracking; GF capex-deployment commentary in 10-K explicitly addresses subsidy compliance.	The $13B+ “over 10 years” investment framing matches the 2024-disclosed CHIPS subsidy schedule; GF’s project-management track record (Malta IPO-era expansion, Dresden expansion) demonstrates milestone-delivery competence.
K3	Capex peak (2025-2027) vs FCF compression. $2.5-3.5B annualized capex against ~$1B annualized FCF (analyst estimate, ⚠ inferred from Q4 2025 disclosed segment-data). At cycle peak, GFS may need to draw down cash reserves or modestly increase net debt.	3	2	Quarterly 10-Q cash + investments + net-debt disclosures; capex-vs-FCF gap explicitly tracked in 10-K MD&A.	CHIPS Act subsidies offset ~16% of US capex; EU Chips Act subsidies (analyst estimate) similarly offset Dresden capex. Customer prepayments + multi-year supply commitments provide working-capital cushion.
K4	AMF acquisition price + InfiniLink price unknown. Both acquisitions closed without disclosed financial terms (GF press releases ✓); the financial-impact assessment depends on rebuilds from accounting reconciliations in the 10-K. Risk: if either acquisition was overpaid, goodwill-impairment risk emerges in subsequent periods.	1	2	FY 2025 10-K + FY 2026 10-K acquisition-related goodwill + intangibles disclosures; subsequent goodwill-impairment-test outcomes.	AMF was a private foundry with disclosed FY 2024 revenue at industry-typical ranges; price discovery happens through 10-K disclosure.
K5	Form-4 insider selling pattern interpretation. edgar_recent.json ✓ shows 13+ Form 4 filings from 2026-03-20 through 2026-04-27 (a high cadence). These are likely 10b5-1-plan-driven sales rather than discretionary; primary-source verification needed in the underlying Form 4 XML disclosures.	2	1	Form 4 transaction-code field (S = sale, M = exercise-and-sale, etc.); 10b5-1 plan adoption disclosures; insider-trading window dates.	10b5-1 plans are mechanical rather than informational; the high cadence is consistent with a typical post-IPO equity-compensation vesting schedule.

T — Technology risks

ID	Risk	L	I	Detection trigger	Mitigation / management commentary
T1	Fotonix yield-learning curve at next-generation node (45CLO → 22CLO transition). Fotonix is built on the 45nm CLO process node (GF SiPh page ✓); the next-generation node transition (likely 22CLO + 12FDX-photonics integration) carries yield-learning risk.	2	2	OFC 2027 / 2028 process-roadmap disclosures; first-customer-tape-out outcomes on next-node Fotonix; yield-recovery commentary in 10-K MD&A.	GF’s 45CLO Fotonix yield curve has 5+ years of production data; the company has demonstrated multiple successful node-transition cycles in its history.
T2	AMF 200mm → 300mm Singapore upgrade execution risk. The announced “scale to 300mm as market needs grow” path requires fab upgrade investments and process-recipe migration. Execution slippage by >12 months would push out the bull-case Pillar 2 capacity-consolidation thesis.	2	2	Quarterly 10-Q disclosures of Singapore capex + revenue ramp; AMF customer-side commentary on supply-continuity during upgrade.	GF has internal-knowledge transfer from prior 200mm → 300mm migrations (e.g., Vermont legacy to Malta 300mm). A*STAR research-partnership commitment demonstrates Singapore-government alignment.
T3	Next-generation modulator architecture (TFLN, BTO) requires new process flows. TFLN modulators (HyperLight, Polariton/Marvell context); BTO ferroelectrics; metasurface optics — each requires a new process flow that may not be a clean extension of Fotonix’s current SOI-based recipe.	2	2	Hyperscaler module-spec disclosures favor TFLN or BTO; competitor (Tower, Intel) discloses TFLN/BTO process readiness.	Fotonix is modulator-architecture-agnostic at the integration level — TFLN heterogeneous-integration (chip-on-Fotonix) is a viable path. The InfiniLink design IP includes optical-transceiver chipsets that integrate with non-SOI modulators.
T4	CPO architectural shift bypasses pluggable-engine demand. Co-Packaged Optics is advancing rapidly (Tower CPO foundry; Marvell Celestial AI; Broadcom Tomahawk-Bailly). If hyperscalers move directly to CPO at 1.6T+ at faster pace than expected, the pluggable-engine-Fotonix demand subset compresses.	2	3	OFC / ECOC 2026-2028 disclosures show hyperscaler CPO insertion at 1.6T+ ahead of base-case timeline; module-OEM CPO design wins drain pluggable-engine demand.	Fotonix is a foundation process for both pluggable engines and CPO chiplets — GF’s CPO-foundry-roadmap announcements position the platform for the architectural transition. The InfiniLink CPO design IP is part of this positioning.
T5	400 Gbps/lane technology-readiness gap. GF’s A*STAR Center of Excellence targets “400 Gbps speed” data transfer (GF press release ✓) for 3.2T+ optical chiplets. The technology-readiness gap to 400 Gbps/lane is multi-year R&D execution; competitor architectures (TFLN at HyperLight; POH at Polariton/Marvell; EO-polymer at LWLG) target the same window.	2	2	OFC 2027 / 2028 demonstration-system disclosures; first-customer-tape-out outcomes at 400 Gbps/lane; competitor-architecture demonstrated-bandwidth records.	A*STAR research partnership + multi-customer PDK ecosystem (LWLG, NLM, Polariton via Marvell) provides multiple competing modulator-architecture paths to 400 Gbps/lane.

R — Regulatory risks

ID	Risk	L	I	Detection trigger	Mitigation / management commentary
R1	CHIPS Act clawback under capex-deployment-milestone covenants. Same as K2 above; treated as both capital and regulatory risk. Not double-counted in scenario aggregations.	1	3	NIST / Commerce notices; explicit GF compliance disclosures.	See K2.
R2	EU State-Aid framework changes affect Dresden subsidy economics. European Commission state-aid review processes can retroactively modify foundry-subsidy terms.	2	2	Commission decisions; specific German government press; Dresden-project capex-rebate disclosures.	Diversified geographic platform reduces single-jurisdiction dependency.
R3	Singapore foreign-investment / industrial-policy shifts post-AMF. Singapore government policy changes affecting foreign-controlled foundries.	1	2	EDB / Singapore government policy announcements; A*STAR partnership commentary.	A*STAR partnership signals strong government alignment.
R4	CFIUS / national-security review on Mubadala-related transactions. While not currently a live risk (Mubadala is a long-standing US-aligned investor), future Mubadala-related strategic transactions could face CFIUS scrutiny.	1	2	CFIUS clearance announcements; defense-related GF customer commentary.	Mubadala has cleared multiple CFIUS reviews historically; structural risk is well-managed.
R5	Antitrust scrutiny of merchant-foundry consolidation. AMF + InfiniLink + future strategic transactions could face antitrust review (e.g., FTC merger-guideline updates affecting semiconductor M&A).	1	1	FTC / DOJ second-request notices; foreign antitrust authority reviews (Singapore CCS, EU).	AMF + InfiniLink were small enough to clear without significant regulatory friction.

X — Competitive risks

ID	Risk	L	I	Detection trigger	Mitigation / management commentary
X1	Tower Semiconductor merchant-SiPh competitive pressure. Tower’s $300M SiPh expansion + CPO foundry technology positioning (GlobeNewswire 2025-11-12 ✓) + 200mm + 300mm hybrid path.	3	2	Tower Q1-Q4 CY 2026 SiPh segment revenue prints; OFC 2027 customer-roster disclosures.	GF has structural 300mm + scale advantage. Tower’s CPO foundry technology is targeting a different layer (foundry-level CPO) than Fotonix (foundry + customer-PDK).
X2	TSMC SiPh merchant-launch (post-2027). TSMC has signaled potential merchant-SiPh capacity opening on the 2027-2028 horizon. TSMC’s brand premium with hyperscalers is structural.	2	3	TSMC capex guidance includes SiPh-merchant allocation; OFC / OCP 2027-2028 customer-roster disclosures.	GF’s 5+ years of yield-learning + named-customer roster is structurally hard to replicate quickly. TSMC’s existing internal-customer prioritization is a meaningful constraint.
X3	Intel SiPh merchant-pivot (post-IDM-2.0). Intel’s IDM-2.0 transformation has been creating partial-merchant-foundry posture. Intel’s SiPh capability includes deeper CMOS+photonics integration than Fotonix.	2	2	Intel earnings calls cite SiPh-merchant customer wins; module-OEM partner press releases citing Intel SiPh integration.	Intel SiPh has historically struggled with merchant-foundry execution and customer-relationship breadth. GF’s customer roster represents a structural lock-in advantage.
X4	Hyperscaler in-house SiPh production (NVIDIA, Broadcom). Same as C2; treated as both customer and competitive risk. Not double-counted.	2	3	See C2.	See C2.
X5	Marvell + Polariton + Celestial AI end-to-end optical stack. Marvell’s $130B-mkt-cap end-to-end positioning (MRVL bull case ✓). Marvell could in principle qualify Polariton’s POH at Tower or TSMC instead of GF, removing one of GF’s flagship Fotonix customers.	2	2	Marvell Q1-Q4 CY 2026 calls; Polariton process-flow disclosures; OFC 2027 disclosures.	Marvell already has a multi-year Fotonix relationship that creates switching costs. POH integration on Fotonix is the most-derisked path for the post-Polariton Marvell roadmap.

M — Market / Macro risks

ID	Risk	L	I	Detection trigger	Mitigation / management commentary
M1	AI capex digestion 2026-2027. Hyperscaler capex pause / digestion cycle following the 2022-2025 acceleration. Optical interconnect demand softens 6-12 months after compute-capex deceleration.	2	3	Hyperscaler Q4 CY 2025 + Q1-Q4 CY 2026 capex guidance commentary; Big-4 hyperscaler aggregate capex guidance for FY 2026; module-OEM order book guidance.	Multi-customer roster + $1B SiPh trajectory through end-2028 spans the digestion cycle. Even at digestion-cycle pace, the Pillar 7 design-win cohort conversion provides demand floor.
M2	Mature-node ASP recompression. Chinese mature-node fab buildouts (SMIC, Hua Hong, JCET) add structural overcapacity. Macro recession compounds the pressure.	3	2	SMIC / Hua Hong CY 2026 capex guidance; merchant-foundry pricing-survey data; segment-revenue prints.	Diversified segment mix; 95%+ sole-sourced design-win base; CHIPS Act + multi-geo platform supports US-domestic premium.
M3	Recession or IT-spend pullback compresses transceiver-module unit-volume demand. Slower hyperscaler + telco infrastructure spend reduces module-OEM order books and Fotonix volume conversion.	2	2	Module-OEM order book guidance; carrier capex cuts at telco operators; transceiver pricing compresses.	AI-driven demand has been more recession-resilient than traditional enterprise IT spend.
M4	Multiple-compression on photonics-narrative re-rating reversal. GFS trades at a meaningful premium to peer mature-node merchant foundries on the photonics-narrative growth premium. If the narrative cools (e.g., end-2028 $1B target slips), multiple compresses to the peer-foundry baseline.	2	3	Peer-comp trading multiples; explicit analyst-day re-rating commentary; missed quarterly SiPh trajectory prints.	Premium is justified by the actual photonics revenue trajectory + scarcity-of-merchant-300mm-SiPh structural moat.
M5	Dollar-strength compresses non-USD revenue translation. GFS reports in USD; meaningful Singapore-dollar + euro revenue creates translation FX exposure.	2	1	10-K currency translation footnotes; FX-hedging program disclosures.	Standard FX-hedging practice applies.

S — Supply chain risks

ID	Risk	L	I	Detection trigger	Mitigation / management commentary
S1	Equipment vendor (ASML, AMAT, KLA, LRCX, TEL) capacity allocation. Foundry capacity expansion is gated by equipment-vendor allocation. ASML’s photolithography systems are the critical-path bottleneck.	2	2	ASML Q1-Q4 CY 2026 backlog disclosures; equipment-vendor customer-allocation commentary.	GFS’s long-standing equipment-vendor relationships + multi-vendor sourcing for capacity-add cycles is industry-standard.
S2	Wafer (200mm + 300mm SOI) supply concentration. 300mm SOI wafers are sole-sourced from Soitec; 200mm SOI is similar concentration. Soitec capacity is allocated across the global SiPh customer base.	2	2	Soitec quarterly-allocation disclosures; AMF Singapore upgrade timeline gated by 300mm SOI wafer supply.	Long-standing Soitec relationship; capacity planning is multi-year forward-committed.
S3	Critical-materials supply (HBM packaging, advanced substrate, photoresists). Geopolitical events affecting Japan / Korea / Taiwan supply chains for critical foundry materials.	1	2	Japan / Korea export-control updates; Tier-2 supply-chain diligence reports; quarterly cost-of-revenue commentary.	Multi-vendor sourcing standard in foundry-materials supply chain.
S4	AMF Singapore-specific supply-chain integration risk. AMF’s existing supply chain (200mm SOI, photoresists, etch chemistries) needs integration with GF’s central procurement; transition friction could compress AMF segment margins through CY 2026.	2	1	AMF segment-margin disclosures in FY 2026 10-K (analyst estimate ⚠ inferred — GF may not break out AMF separately); supply-chain commentary in MD&A.	Standard post-acquisition integration playbook; AMF’s existing supplier relationships likely retained for continuity.

Aggregated risk-scenario sensitivity

Scenario	Risks materializing simultaneously	Implied valuation impact (vs. spot $59.49)
Single-tail bear (most-likely-near-term)	K1 (Mubadala selldown event)	-5% one-day, -3% drift = ~$56-57 floor
AI digestion combined-stress	M1 + M4 (capex digestion + multiple compression)	-15 to -20% = ~$48-51 floor
Photonics-narrative break	T4 + X1 + X2 + C2 (CPO shift + Tower + TSMC + hyperscaler in-house)	-25 to -30% = ~$42-45 floor
Hard-bear combined	K1 + M1 + M2 + T4 + X1 + R1 (Mubadala + AI digestion + mature-node ASP + CPO + Tower + CHIPS clawback)	-35 to -45% = ~$33-39 (back to 52-week-low region $31.51)
Single-tail bull (most-likely-near-term)	$1B SiPh trajectory ahead-of-target + design-win expansion	+10 to +15% = ~$65-68 ceiling
Hyperscaler-recommitment combined	C2 falsified + customer-design-win expansion + AI-capex-acceleration	+25 to +35% = ~$74-80
Multi-bull combined	All eight bull pillars compound through 2027	+40 to +60% = ~$83-95 (multi-year horizon)

The hard-bear scenario revisits the 52-week low of $31.51; the multi-bull combined scenario reaches into the $80-95 range that anchors the bull-case valuation in valuation ranges. The distribution of outcomes is asymmetric in favor of the bull tail at current spot — but the bear-tail-near-spot risk (Mubadala selldown plus an AI-digestion print) is mechanically active and should be factored into position sizing.