TSEM Open Research Questions

This file enumerates the questions about Tower Semiconductor that the company has not publicly answered as of 2026-04-29 — questions that materially affect the bull / bear thesis but cannot be resolved from primary-source materials at this date. Each question gets:

The research path: where the answer would surface (forward filings, industry conferences, partner disclosures, etc.)
Current best-guess with explicit confidence flag (◐ partial / ⚠ inferred / ✓ corroborated)
Thesis-impact mapping to bull / bear pillars and risk register IDs

The list is prioritized roughly by thesis-significance: highest-impact unanswered questions first. Where the analyst genuinely does not have an answer, the entry says so explicitly — speculation is flagged inline rather than substituted for primary-source verification.

Q1. Who are the top-4 unnamed customers in the FY 2024 20-F (27% combined revenue concentration)?

Research path: (a) FY 2025 20-F (April-May 2026 expected filing window) customer-concentration footnote; (b) industry triangulation via supplier-relationship analysis at named publicly-listed semiconductor customers; (c) module-OEM and IDM annual reports citing Tower as a foundry partner; (d) supply-chain disclosures from automotive Tier-1s, Apple supplier disclosures, RF-front-end-module OEMs.

Status update 2026-04-29: Customer-side primary-source verification pass partially resolved this. The full record lives at customer verification 2026-04-29. Verified customer-side primary attestations of Tower / TowerJazz / Jazz Semiconductor as a foundry counterparty:

Vishay Intertechnology (Siliconix subsidiary) ✓ — VSH FY 2022 10-K: long-term Tower foundry agreement since 2004 with multi-year minimum wafer-purchase commitments. Disclosure language streamlined out post-FY 2022; current status unconfirmed.
Maxim Integrated (now ADI by succession) ✓ — MXIM FY 2021 10-K: long-term wafer-supply agreement with TowerJazz Texas (San Antonio Fab 9) executed 2015-11-18. ADI acquired Maxim 2021-08-26; ADI 10-K silent on inheritance.
Impinj ✓ — Impinj FY 2021 10-K: Tower is sole reader-IC wafer supplier since 2008. Genericized to “foundry partners” in FY 2023+; current status unconfirmed.
Inphi (now Marvell by succession) ✓ — Inphi FY 2019 10-K: TowerJazz one of four principal foundries. Marvell acquired Inphi 2021-04-20.
O2Micro ✓ — O2Micro FY 2021 20-F: TowerJazz Panasonic (TPSCo) one of three top foundries.
Skyworks ◐ partial — Tower-side 2012 IR attestation valid; SWKS modern 10-K silent on specific foundries.

Negative customer-side checks (each candidate’s most recent 10-K silent on Tower / TowerJazz / Jazz / Newport):

Qorvo (FY 2025 10-K), Broadcom (FY 2025 10-K, names TSMC), Semtech (FY 2026 10-K), Texas Instruments (FY 2025 10-K), ON Semiconductor (FY 2025 10-K), NXP (FY 2024 10-K), Teledyne: all remain ⚠ aggregator-only.
Murata, Samsung Semi, Infineon, Renesas: foreign issuers with no SEC filing or Tower disclosure in English-language IR. Renesas’s publicly-named foundry alliance is with GlobalFoundries (expanded Feb 2025) — a structural exclusion signal.

Excluded (customer-side disclosure rules them out):

Lattice Semiconductor (LSCC) — FY 2025 10-K names UMC + USJC + Samsung + TSMC + Epson; Tower NOT in roster.
Cirrus Logic (CRUS) — FY 2025 10-K names TSMC; no Tower reference.

Apple and Bosch in the original speculative list remain ⚠ inferred-not-customer-named — Apple’s supplier disclosures do not name Tower; Bosch is a private company with no SEC filing.

Implication for the 27% / 4-customer bracket: the verified set provides candidate names — Vishay, ADI (by Maxim succession), Marvell (by Inphi succession), Skyworks, Impinj, O2Micro — but no individual triangulation to a 3-11% revenue band. The Skyworks-Qorvo merger close (announced 2025) compounds the unknown.

Confidence: ◐ partial — five customer-side primary attestations exist (Vishay, Maxim/ADI, Impinj, Inphi/Marvell, O2Micro) plus Tower-side Skyworks; six candidates from the original aggregator list checked customer-side and silent (Qorvo, Broadcom, Semtech, TI, ON, NXP); two candidates excluded (Lattice, Cirrus Logic). Top-4 concentration band naming requires further customer-side disclosure — most likely from Marvell (Inphi-lineage), Skyworks (post-Qorvo-merger), or a new disclosure event.

Thesis impact: bull case Pillar 1, 2 (customer-roster diversification) + risks C1, C2 (top-5 concentration). The five verified customers materially de-risk the “Tower customers are entirely opaque” framing — at least the structural shape of the customer roster (RF, MOSFET, RFID-reader, high-speed analog, Japanese power-management) is now visible.

Q2. What is the Tower-Intel Fab 11X first-customer-out timing?

Research path: (a) Tower 6-K capex disclosures + Fab 11X commissioning announcements; (b) joint Tower-Intel press releases referencing Fab 11X production status; (c) FY 2025 + FY 2026 20-F segment disclosures showing US-domestic 300mm capacity contribution; (d) supply-chain triangulation via equipment-vendor (ASML, AMAT, KLA, LRCX, TEL) capacity-allocation reports.

Current best-guess ⚠ aggregator-derived: Per the user-prompt research-task framing, “initial wafers H1 2026, productionized H2 2026” — but this aggregator-derived phrasing requires primary-source verification from Tower 6-K disclosures. Fab 11X is the Albuquerque, NM 200mm/300mm specialty-foundry purchased from Intel in 2024 for $300M equipment-investment program. Industry-typical timeline for an acquired-fab-with-process-recipe-conversion: 18-24 months from acquisition to first-customer-revenue — placing first-customer-out in 2026 H2 to 2027 H1.

Confidence: ⚠ inferred — primary Tower 6-K extraction queued. The “$300M equipment-investment” amount appears to be aggregator-derived; primary Tower disclosure of the exact dollar commitment is queued.

Thesis impact: bull case Pillar 3 (capacity expansion); risks P4, K1 (capex envelope). Material for assessing US-domestic 300mm capacity-deployment timing for the FY 2028 model.

Q3. What is the Tower-ST Agrate 65nm BCD ramp curve through 2026-2028?

Research path: (a) ST Microelectronics quarterly earnings commentary on Agrate JV (ST is the JV partner; ST’s quarterly disclosures provide independent triangulation); (b) Tower 6-K commentary on Agrate cost absorption + revenue contribution; (c) FY 2025 20-F segment disclosure of Agrate JV-attributable revenue and gross margin; (d) joint Tower-ST press releases.

Current best-guess ◐ partial: The next 65nm BCD platform was productized December 23 2024 (Tower release ✓). Tower has been absorbing Agrate operational cost since Q4 2024 ahead of meaningful revenue. The +640 bps GM expansion across 2025 (20.4% Q1 → 26.8% Q4 per quarterly_trend.md ✓) reflects partial Agrate flow-through but is also driven by SiPh mix shift and RF Infrastructure ASP improvement. The clean Agrate-attribution of GM expansion is not separately disclosed.

Confidence: ◐ partial — directional inference is well-supported; specific ramp-curve dollar quantum requires primary-source verification.

Thesis impact: bull case Pillar 2.2 (Agrate operating-leverage flow-through) — this is the load-bearing GM-expansion driver in the FY 2028 model.

Q4. What is the unit-cost economics of PH18 (200mm) vs GFS Fotonix (300mm)?

Research path: (a) industry analyst surveys (LightCounting, Dell’Oro, 650 Group) on per-die SiPh cost economics; (b) hyperscaler module-OEM commentary on per-module cost economics for PH18 vs Fotonix sourcing; (c) academic publications (e.g., IEEE Photonics Society, Optica) comparing 200mm vs 300mm SiPh process economics; (d) Tower or GFS investor-day disclosures of per-wafer ASP (neither company has historically disclosed this directly).

Current best-guess ⚠ inferred: Neither Tower nor GFS has disclosed wafer-cost data publicly. Industry-typical inference:

300mm SiPh wafer ASP (GFS Fotonix): $5,000-10,000 per wafer ⚠
200mm SiPh wafer ASP (Tower PH18): $2,500-5,000 per wafer ⚠
Per-die cost: 300mm produces ~2.2× more die per wafer at typical SiPh die size, partially offsetting the ASP difference
Net per-die cost on Fotonix could be 15-30% lower than PH18 ⚠ at scaled production

The $1B end-2028 SiPh revenue implication for GFS would be ~100K-200K wafers per year. Tower’s $228M FY 2025 SiPh revenue at ~$3-5K per wafer ASP implies ~50K-75K wafers per year ⚠ inferred. The implied unit economics is the principal load-bearing variable in the GFS-vs-Tower competitive thesis.

Confidence: ⚠ inferred — direct disclosure unlikely from either company. Customer-side commentary at OFC 2027 / OFC 2028 may provide indirect signals.

Thesis impact: bull case Pillar 1 (PH18 differentiation) + risks X1 (GFS Fotonix at 300mm cost-down threat). The most analytically consequential unanswered question — directly determines whether PH18’s 200mm process can hold its own against 300mm Fotonix as polymer-modulator volumes ramp.

Q5. What is the SiGe-only revenue split from the SiPh+SiGe combined disclosure?

Research path: (a) FY 2025 20-F segment-revenue footnote may disaggregate SiPh from SiGe; (b) Tower IR Q&A at industry conferences may provide directional split; (c) segment_revenue_mix.md ✓ tracks the disclosed numbers.

Current best-guess ◐ partial: Tower discloses combined SiPh + SiGe revenue at $241M FY 2024 → $421M FY 2025 (+75% YoY). Tower also separately disclosed silicon photonics specifically at $106M FY 2024 → $228M FY 2025 (Tower Q4 2025 release ✓). The implied SiGe-only revenue is $241M − $106M = $135M FY 2024, growing to $421M − $228M = $193M FY 2025 (+43% YoY). This split is directional, not officially disclosed.

Confidence: ◐ partial — derivable arithmetic, not separately disclosed segment.

Thesis impact: bull case Pillar 1 — separating SiGe from SiPh allows independent tracking of each growth driver. SiGe driver-IC growth is structural-AI-related but distinct from the PH18 modulator narrative; conflating them would over- or under-state the SiPh ramp.

Q6. What is the CHIPS Act funding probability for Tower-Intel Fab 11X — and what amount?

Research path: (a) NIST + Department of Commerce CHIPS Act award announcements; (b) Tower IR press releases referencing CHIPS Act application; (c) 20-F MD&A commentary on CHIPS Act application status; (d) congressional hearings + bipartisan negotiation commentary referencing US-Israeli specialty-foundry partnerships.

Current best-guess ⚠ inferred: Tower has not publicly committed to a CHIPS Act application. Eligibility criteria fit (US-domestic-fab capacity expansion + specialty-process diversification) at Fab 11X is strong. Probability ⚠ inferred at 25-40% based on (a) eligibility criteria fit, (b) Tower’s documented Fab 11X investment program, (c) post-2026 administration political-economy framing of US-Israeli specialty-foundry partnerships. Likely amount range if granted: $200-500M ⚠ inferred — sized to a fraction of the Fab 11X equipment-investment program plus possible additional capex commitments. The base case is no award, not a risk per se — an award would be upside.

Confidence: ⚠ inferred — no Tower public application disclosure as of 2026-04-29.

Thesis impact: risks K3 (CHIPS Act funding optionality). Material for the valuation ranges bull case multiplier; an award would shift the balance-sheet capex math materially.

Q7. What is the debt issuance probability for the $920M SiPh/SiGe envelope?

Research path: (a) Tower 6-K announcements of bond issuance or bank credit facility expansion; (b) FY 2025 + FY 2026 20-F balance sheet disclosures; (c) Israeli bond-market issuance reports referencing Tower; (d) credit-rating-agency commentary (Moody’s, S&P).

Current best-guess ⚠ inferred: Tower’s FY 2024 YE balance: cash + ST deposits ~$1.22B, total debt ~$0.18B, net cash ~$1.04B ✓. Customer prepayments fund >70% of $920M = $644M+, leaving ~$276M residual. Combined with base-rate maintenance capex ($300-400M/year) over the envelope deployment period, FY 2026-2028 cumulative capex could approach $1.5-2.0B, while FCF generation in 2026-2027 is constrained by capex peak. Debt issuance probability ⚠ inferred at 30-50% for a $300-500M issuance over the 2026-2027 window. Tower’s investment-grade-equivalent profile + clean balance sheet support reasonable cost-of-debt.

Confidence: ⚠ inferred — material future-disclosure event probability assessment.

Thesis impact: risks K1, K2 (capex vs FCF; debt issuance probability). Material for valuation ranges base/bear scenarios — debt issuance dilutes shareholder return marginally and adds interest expense.

Q8. What is the GFS-PH18-equivalent (i.e., GFS launching a 200mm SiPh process competitive with PH18) timeline?

Research path: (a) GFS investor-day disclosures referencing 200mm SiPh strategy post-AMF acquisition; (b) AMF 200mm Singapore capacity utilization + customer-roster expansion; (c) OFC / ECOC 2026-2028 disclosures; (d) cross-reference GFS bull case ✓ for AMF integration roadmap.

Current best-guess ◐ partial: GFS acquired AMF (Advanced Micro Foundry, Singapore, 200mm SiPh) Nov 17 2025 (GFS press release ✓). The AMF 200mm process is now part of the GFS portfolio and is structurally the closest GFS-equivalent to PH18. Singapore 300mm upgrade is the announced trajectory through 2027-2028 (GFS catalysts ✓). The 200mm AMF process operates today and is therefore an immediate competitive presence. The 300mm upgrade extends the GFS competitive footprint to be even more architecturally similar to PH18 + Fotonix combined.

Confidence: ◐ partial — AMF acquisition is verified ✓ ; specific GFS-200mm-customer-roster-vs-Tower comparison requires primary-source customer-side verification.

Thesis impact: risks X1 (GFS Fotonix competitive pressure + AMF post-acquisition expansion). High-impact: AMF integration extends GFS’s process-portfolio coverage to overlap with Tower’s 200mm specialty positioning.

Q9. Will TSMC open SiPh foundry-mode capacity to merchant customers, and on what timeline?

Research path: (a) TSMC capex guidance allocation to SiPh-merchant; (b) OFC / OCP 2027-2028 customer-roster disclosures; (c) hyperscaler module-spec commentary citing TSMC SiPh; (d) TSMC technology-symposium (typically annual) disclosures of SiPh-merchant strategy.

Current best-guess ⚠ inferred: TSMC has signaled potential merchant-SiPh capacity opening on the 2027-2028 horizon (per industry trade press and TSMC technology-symposium commentary; ⚠ aggregator-only). TSMC’s existing internal-customer prioritization (NVIDIA, AMD) is a meaningful constraint through 2027. Probability of material TSMC merchant-SiPh launch by end-2028: ~40-60% ⚠ inferred. If TSMC enters merchant SiPh at scale, both Tower PH18 and GFS Fotonix face share compression at the highest-end customer tier.

Confidence: ⚠ inferred — TSMC has not made formal merchant-SiPh commitment.

Thesis impact: risks X3 (TSMC SiPh merchant-launch). High-impact long-term competitive risk.

Q10. What is SiCarrier’s specialty-photonics roadmap and timeline?

Research path: (a) Chinese trade press + SEMICON China disclosures referencing SiCarrier; (b) Chinese Ministry of Industry and Information Technology (MIIT) industrial-policy announcements referencing SiCarrier; (c) SMIC / Hua Hong commentary referencing SiCarrier process technology; (d) academic publications (Chinese photonics research groups) citing SiCarrier process collaboration.

Current best-guess ⚠ aggregator-only: SiCarrier is a Chinese specialty-foundry equipment + process developer (per industry reporting; ⚠ aggregator-only, no primary verification). The exact strategic positioning (foundry, equipment vendor, IP licensor) is unclear from public sources. SiCarrier’s role in Chinese-domestic SiPh ecosystem is plausibly significant but quantitatively unknown.

Confidence: ⚠ aggregator-only — no primary-source verification available in English-language sources.

Thesis impact: risks X4 (SiCarrier / Chinese specialty-foundry advance under non-US sanctions regime). Probability and timing are both uncertain; impact is structural for the Chinese-domestic addressable market.

Q11. What is Israel Corp / Idan Ofer ownership in Tower as of 2026? — RESOLVED 2026-04-29

Resolution: ✓ verified-primary via EDGAR 13G/13D archive search (sources agent, 2026-04-29). No Israel-Corp- / Ofer-affiliated 13G/13D filings exist in TSEM’s EDGAR archive in the post-2016 window. Combined with Tower’s own FY24 20-F language describing the company as “without a controlling shareholder,” this confirms Israel Corp / Kenon Holdings / Idan Ofer family vehicles are NOT a current 5%+ Tower shareholder. The historical “~25%” framing in legacy aggregator sources is stale — the stake fell below the 5% Schedule 13G threshold somewhere in 2014-2016 (exact year ⚠ pending primary-source extraction from early-2010s 20-F Item 7).

Current ownership reality (per institutional_holders.md ✓):

Israeli pension cohort (~32% combined): Migdal ~7.8%, Harel ~7.8%, Menora 6.17%, Phoenix 5.77%, Clal 4.80%, Meitav Dash ~1.7%
US institutional layer: T. Rowe ~4.2%, Vanguard ~3.9%, Senvest ~3.5%, BlackRock ~3.5%, Point72 2.6%
No single holder >10%; no controlling shareholder

Thesis impact: risks governance-related. The “fully distributed-float post-Intel-deal-collapse” framing in comps valuation is now confirmed rather than inferred. No M&A-overhang or controlling-shareholder discount applies. Material for institutional-investor allocation decisions — TSEM trades on its specialty-foundry/AI-photonics fundamentals without a controlling-shareholder taxes/discount.

Remaining backfill: Exact year Israel Corp’s stake fell below 5% (would clarify the divestment chronology — likely tied to Intel-deal-pendency exit window 2022-2023 or earlier).

Q12. What is the Israeli board independence dynamic — does Tower’s board reflect majority-independent governance per US standards?

Research path: (a) Tower 20-F governance section; (b) Israeli Companies Law disclosures on independent-director composition; (c) ISS / Glass Lewis governance ratings (aggregator-only); (d) shareholder-rights advisory commentary.

Current best-guess ⚠ inferred: Tower follows Israeli Companies Law governance with US-listed-company supplementary requirements (NASDAQ corporate-governance rules). Industry-typical Israeli-listed-NASDAQ-listed semiconductor company maintains ~60-70% independent-director composition. CEO Russell Ellwanger has been in role since 2005 (20+ year tenure) — long-tenured CEO governance is structurally distinct from US-typical (~5-7 year CEO tenure). Specific board-independence percentage is not separately verified.

Confidence: ⚠ inferred — not separately verified from primary disclosures.

Thesis impact: Lower-priority for investment thesis; relevant for governance-focused allocation decisions.

Q13. What customer is the anchor for the Nov 12 2025 CPO Foundry Program?

Research path: (a) Tower joint press release with the CPO customer (typical disclosure path); (b) OFC 2027 / OFC 2028 disclosure venues; (c) customer-side announcement (AI-cluster system vendor, hyperscaler, module OEM); (d) industry-trade-press customer-pipeline commentary.

Current best-guess ⚠ inferred: The CPO Foundry Program announcement (Nov 12 2025 GlobeNewswire ✓) made the technology platform available — the program is positioned for customer engagement. First customer disclosure expected 2026-2027 (analyst expectation ⚠ inferred from industry-typical 12-18 month foundry-program-to-first-customer cycle). Plausible anchor candidates: Marvell + Polariton (POH-on-Tower path), Coherent (extension of OFC 2026 partnership), Broadcom (Tomahawk-Bailly CPO program), Lightmatter / Ayar Labs (CPO startup tier). None individually verified.

Confidence: ⚠ inferred — high-impact pending customer-disclosure event.

Thesis impact: bull case Pillar 1 (PH18 customer-roster expansion); catalysts “first CPO Foundry customer disclosure” entry. The customer name will materially shape the architectural-positioning narrative.

Q14. What is the Tower-Coherent partnership scope at OFC 2026 — specifically Tower-PH18 attribution?

Research path: (a) Coherent’s investor-relations commentary referencing Tower; (b) Tower IR commentary referencing Coherent; (c) OFC 2026 technical presentations and proceedings; (d) joint Tower-Coherent press releases.

Current best-guess ◐ partial: The user-prompt research-task framing references “Tower-Coherent partnership (OFC 2026 400G/lane SiPh demo)” as one of the publicly-disclosed PH18 photonics-customer markers. However, the specific Tower-Coherent-PH18 attribution requires primary-source verification — Coherent has multi-foundry strategy (GFS Fotonix relationship is well-established) and may have demonstrated 400G/lane SiPh on a different process. The Tower-Coherent partnership scope (volume, exclusivity, design-win) is not separately disclosed.

Confidence: ◐ partial — the Tower-Coherent partnership existence is inferred; exclusivity and volume scope are pending.

Thesis impact: bull case Pillar 1 (PH18 customer roster); if Tower-Coherent is a high-volume relationship, it materially de-risks the SiPh ramp.

Q15. What is the OFC 2027 expected disclosure cadence for Tower?

Research path: (a) OFC 2027 program (typically published 2-3 months before conference, expected ~December 2026); (b) Tower IR conference-attendance commentary; (c) historical Tower OFC disclosure pattern (OFC 2025, OFC 2026 disclosures provide pattern for 2027 expectations).

Current best-guess ⚠ inferred: Tower has historically disclosed at major industry conferences (OFC, ECOC, OCP). Expected OFC 2027 disclosures (March 2027): (a) 110 GHz+ tapeout results from LWLG-PH18 development agreement (per bull case Pillar 1.2 timeline), (b) CPO Foundry Program first-customer disclosure if timeline tracks, (c) PH18 roadmap update for 200mm scale, (d) potential 300mm SiPh process extension or platform announcement (analyst speculation ⚠ — Tower has not committed to 300mm SiPh).

Confidence: ⚠ inferred — pattern-based expectation.

Thesis impact: catalysts OFC 2027 entry. High-density disclosure venue for multi-pillar thesis verification.

Q16. What is the NIS/USD exchange rate trajectory through 2026 and Bank of Israel policy cycle 2026?

Research path: (a) Bank of Israel monetary policy committee decisions and minutes; (b) Israeli inflation prints; (c) NIS-USD spot data; (d) Israeli sovereign bond yields.

Current best-guess ⚠ inferred: Recent NIS-USD has ranged ~3.4-3.7 (2025-2026). 2026 trajectory depends on (a) Israeli inflation dynamics, (b) regional-conflict cycle, (c) US Federal Reserve policy. Bank of Israel rate-cycle direction is uncertain — Israeli inflation and regional-conflict dynamics drive divergent policy paths.

Confidence: ⚠ inferred — macro forecast.

Thesis impact: risks F1, F2 (NIS/USD volatility, Bank of Israel policy). Lower-priority for fundamental thesis but material for cost-of-equity premium dynamics.

What this list deliberately does not include

The research-prompt direction explicitly noted that the open-questions file should be honest where the analyst doesn’t know — not fabricate answers. The following types of questions have been deliberately not added because the analyst genuinely has no informed inference yet:

Specific 6-K-equivalent filings frequency analysis for FY 2025-2026 (Tower’s pattern is 4 per year, not separately worth listing as a question)
Detailed cost structure of the Tower-Intel Fab 11X integration (acquired-fab integration economics is multi-variable)
Specific Tower position on EU Chips Act subsidy renegotiation outcomes (Tower operates Agrate JV under EU Chips Act framework but at JV-partner-second-order rather than primary)
Tower’s TFLN / BTO / metasurface roadmap (Tower has not publicly committed to alternative modulator-architecture programs beyond LWLG EO-polymer)

These are open-questions categorically, but listing them with speculative best-guesses would violate the source-validation rule.