Skip to content
primarysourced Photonics sector Tower
TSEM
~13 min read · 2,958 words ·updated 2026-04-29 · confidence 43%

Margins and pricing — specialty-analog ASP, fab utilization, mix sensitivity

As of: 2026-04-29 (data through Q4 2025 6-K release filed 2026-02-11; FY 2025 20-F expected April-May 2026)

Confidence legend: ✓ verified-primary 6-K / 20-F · ◐ partial / aggregator · ⚠ inferred / estimate

1. Gross margin trajectory — FY 2022 peak → FY 2025 trough → FY 2026+ inflection

MetricFY 2022 ✓FY 2023 ✓FY 2024 ✓FY 2025 ✓FY 2024→FY 2025 Δ
Net revenue ($M)1,6791,4221,4401,566+9%
Gross profit ($M)~480354339364+7%
Gross margin~28.5%24.9%23.5%23.2%−30 bps
Operating profit ($M)~350~190191194+2%
Operating margin~21%~13%13.3%12.4%−90 bps

Source: FY 2022-2025 from Tower 6-K releases (Q4 2025 ✓, Q4 2024 ✓); FY 2023 from prior-period comparatives; FY 2022 ⚠ aggregator-derived (StockAnalysis ◐).

The FY 2022 → FY 2025 GM compression of ~530 bps (28.5% → 23.2%) reflects: (i) post-COVID specialty-analog cyclicality + RF mobile destocking through FY 2023, (ii) Tower-ST Agrate startup cost absorption beginning Q4 2024, (iii) capex-driven depreciation drag on PP&E base. The FY 2023 → FY 2025 stabilization at 23-25% range is the cyclical floor — Tower’s GM has not gone below 20% in any quarter of this cycle.

Quarterly progression — the FY 2025 sequential climb

QuarterRevenue ($M)GMYoY ΔQoQ Δ
Q1 202432722.3%—(vs Q4’23: −1.6 pp)
Q2 202435124.8%—+250 bps
Q3 202437125.1%—+30 bps
Q4 202438722.5%—−260 bps (Agrate cost absorption begins)
Q1 202535820.4%−190 bps−210 bps (worst quarter — Agrate ramp + seasonality)
Q2 202537221.5%−330 bps+110 bps
Q3 202539623.5%−160 bps+200 bps
Q4 202544026.8%+430 bps+330 bps

The Q1 2025 → Q4 2025 GM expansion of +640 bps over 4 quarters is the load-bearing operating-leverage data point in the entire Tower thesis. It is the proof-of-concept that:

  1. Agrate cost-absorption drag is transitional — not structural
  2. Mix shift toward higher-ASP RF Infrastructure / SiPh flows through to consolidated GM
  3. Utilization improvement on existing capacity flows through at high incremental margin
  4. Management’s FY 2028 model GM of 39-40% requires ~+1,300 bps of additional expansion from Q4 2025’s 26.8% — a slower-pace continuation of the FY 2025 trajectory

Q1 2026 guide — typical seasonal step-down, structurally higher base

The Q1 2026 revenue guide of $412M ± 5% is below Q4 2025’s $440M peak, reflecting standard Q1 seasonality (handset OEM model-year transitions + industrial-customer holiday pullback). Tower has not provided explicit Q1 2026 GM guidance ⚠ — but the implied trajectory per management’s “sequential revenue and profitability growth throughout 2026” (Investing.com ✓) suggests Q1 2026 GM probably in the 24-26% range — a sequential step-down from Q4 2025 26.8% but materially above Q1 2025 20.4%.

2. ASP and wafer-shipment dynamics — the specialty-foundry-pricing story

Tower does not publicly disclose ASP-per-wafer or wafer shipment counts at the level of detail GFS provides quarterly. The closest available framework comes from:

  • Annual revenue + estimated wafer-out volume (analyst-derived ⚠)
  • Process-mix commentary at earnings calls
  • End-market mix (Q4 2025 disclosure: RF Infrastructure 32% / RF Mobile 24% / Sensors-Displays 15% / Power Mgmt 15% / Other 14%)

Tower’s ASP structure — by process complexity

ASP per wafer at Tower varies materially by process tier:

Process tierWafer sizeTypical ASP per wafer ⚠End-market
65nm SiPh (PH18) — premium200mm$8,000-15,000AI optical-interconnect transceivers (1.6T)
130nm/180nm SiGe BiCMOS — driver IC for SiPh200mm$6,000-10,000SiGe driver IC for optical transceivers
65nm BCD (Tower-ST Agrate) — power300mm$5,000-8,000Premium handset PMIC, EV power
180nm BCD — power200mm$3,000-5,000Industrial / automotive analog
180nm RF SOI — RF mobile200mm$2,500-4,500Smartphone front-end RF
180nm CMOS image sensor200mm$2,500-4,000Industrial machine vision, ADAS imaging
130nm Discretes / Other150mm/200mm$1,500-2,500Legacy industrial, discrete power

Source: Analyst working assumptions ⚠ — not Tower-disclosed. Triangulated from comparable specialty-foundry public ASPs (GFS, VIS, X-FAB) and Tower’s known process-mix.

The structural premium for SiPh (PH18) is the central economic basis for the $920M capex envelope. At ~$10K-15K per wafer (vs ~$3K-4K for legacy 200mm CMOS), each SiPh wafer generates 3-5× the revenue of a comparable legacy specialty wafer. Combined with gross margin in the 35-45% range on SiPh (vs corporate average ~25%), each SiPh wafer is 5-8× as gross-profit-accretive as a legacy wafer. This is what justifies the capex per wafer-out — even though SiPh capacity is expensive ($1M+ per wafer-start of capacity ⚠), the per-wafer revenue + margin economics support the unit ROIC.

ASP comparison to peer foundries

CompAvg ASP per wafer (TTM)Source / methodology
Tower (TSEM)~$3,500-4,500 ⚠Implied from FY 2025 $1,566M revenue / ~400K wafers FY 2025 ⚠; precise wafer count not disclosed
GlobalFoundries (GFS)~$2,800 ✓Q3 2025 $1,688M / 602K wafers (per GFS margins_and_pricing)
VIS ◐~$2,000 ◐Aggregator-derived; lower-mix specialty foundry
DBHiTek ◐~$2,500 ◐Aggregator-derived
TSMC ◐~$5,000 (advanced node) / $2,500 (mature) ◐Bimodal by node
UMC ◐~$1,800 ◐Lower-mix mature node

Tower’s blended ASP at ~$3,500-4,500 is the highest among 200mm specialty-foundry comps — reflecting the SiPh + SiGe + Tower-ST Agrate mix that pulls average pricing up. The premium-process-tier wafers (PH18 SiPh) command 2-3× the average; legacy 150mm Fab 1 discretes drag the average down. Discontinuation of legacy 150mm flows + acceleration of SiPh mix is structurally raising blended ASP at ~+5-8% YoY through 2025-2027 ⚠.

3. Customer concentration and ASP dynamics

Tower’s customer concentration disclosure (per FY 2024 20-F):

  • Top customer at 13% of revenue ⚠ — widely understood to be Nuvoton Technology Corporation Japan (NTCJ) via the TPSCo-related volume and image-sensor flows
  • Multiple customers in the 5-10% range ⚠
  • Long-tail customer base of approximately 200+ active accounts

NTCJ at ~13% has a long-term volume agreement with Tower (associated with the TPSCo joint venture restructuring). The volume is price-protected — i.e., NTCJ’s wafer volume comes at negotiated pricing typically below blended-ASP on a per-wafer basis. Smaller / smaller customers pay higher ASP for capacity-reservation premiums.

Customer-prepayment ASP implications

The “firmly backed with customer prepayments” language at Q4 2025 release (per TipRanks ✓) introduces a new ASP dynamic:

  • Hyperscaler / module-OEM customers reserving SiPh capacity through 2028 are providing upfront cash
  • In exchange, customers receive price-locked capacity allocation at potentially below-spot ASP (capacity-reservation discounts)
  • Tower receives certainty of revenue + working-capital benefit, but at slightly compressed per-wafer ASP vs spot
  • Net economic effect: GM may compress modestly on the prepaid-customer cohort but utilization risk is transferred to customers — customers pay regardless of whether they take all wafers

Read. Customer prepayments trade modest ASP compression for utilization-risk transfer. This is a structurally favorable swap for Tower because the principal margin-compression risk in a specialty foundry is under-utilization (40-65% utilization causes severe GM erosion); locking in 70%+ capacity at slightly compressed ASP eliminates that downside scenario.

4. Utilization rate — the load-bearing margin lever

Tower does not consistently disclose fab utilization rates in 6-K releases. Available data points:

  • Q3 2025 utilization “in 80%+ range” — per management commentary ⚠ paraphrased from Q3 2025 call
  • FY 2025 average utilization ~75% ⚠ — analyst working assumption based on revenue-recovery trajectory + capex run-rate
  • Cycle trough utilization ~55-60% ⚠ — FY 2023 H2 estimate during the post-Intel-deal-collapse + RF-mobile destocking trough
  • 65% utilization threshold for margin compression ⚠ — typical specialty-foundry break-even level below which fixed-cost absorption becomes punishing
Utilization bandEstimated GM impact ⚠Cycle context
<60%<18% GMSevere under-utilization (FY 2023 H2 trough)
60-70%20-23% GMSoft — typical Q1 seasonality
70-80%23-27% GMHealthy — Tower’s FY 2025 average
80-85%27-30% GMStrong — Q4 2025 / Q3 2025
85-90%30-33% GMCycle peak
90-95%33-37% GMConstrained — capacity expansion required
95%+37%+ GMCapacity-constrained pricing power

Q4 2025 utilization specific quote — analyst gap. The Q4 2025 6-K release and earnings call did not state a specific Q4 2025 utilization-rate number. Management commentary on the Q3 2025 call cited “80%+” range, and the Q4 2025 26.8% GM print is consistent with utilization climbing toward the 82-85% range ⚠. The lack of a precise Q4 2025 utilization quote is one of the load-bearing analyst-modeling gaps — flag for FY 2025 20-F MD&A extraction (TODO ⚠).

FY 2025 average utilization implication for FY 2026 forecasting

If FY 2025 average utilization was ~75% and Q4 2025 climbed to ~82-85%, then FY 2026 average could realistically reach 80-85% — which corresponds to GM in the 27-30% range at constant mix. A sustained shift to higher-ASP SiPh mix (+5-8% mix shift) plus utilization at 80-85% would deliver FY 2026 GM ~28-30% — consistent with management’s “sequential profitability growth throughout 2026” framing.

5. Margin levers — what’s driving the recovery and the FY 2028 inflection

5.1 Mix shift to RF Infrastructure (SiPh + SiGe)

The single largest GM driver. RF Infrastructure mix shifted from 17% (FY 2024) → 27% (FY 2025) → ~32% Q4 2025. SiPh + SiGe carry structurally higher GM than corporate average:

  • SiPh: estimated 35-45% GM ⚠ at scale (vs corporate ~25%)
  • SiGe driver IC: estimated 32-40% GM ⚠
  • The combined “AI optical-interconnect stack” (27% of FY 2025 revenue) likely contributes **+3-4 percentage points to consolidated GM** above what a homogeneous-process mix would produce ⚠

Mix-shift contribution to FY 2025 vs FY 2024 GM trajectory is partially masked by the Agrate cost drag — without Agrate startup costs, FY 2025 GM would have been ~25-26% rather than 23.2% per management’s qualitative framing.

5.2 Tower-ST Agrate utilization fill-out

Agrate startup costs absorbed in FY 2024 H2 + FY 2025 H1 are reversing as utilization climbs through 2026-2027. Each step-up in Agrate utilization translates to ~+50-100 bps of consolidated GM ⚠ as the fixed-cost absorption improves. Management has framed Agrate as production-ready at end of 2025, with 2026-2027 utilization climbing to capacity as productized 65nm BCD volumes ramp.

5.3 Discontinuation of legacy 150mm Fab 1 flows

Tower has been transferring 150mm Fab 1 lower-margin flows to 200mm Fab 2 (Migdal Haemek). The 150mm Fab 1 has historically run at sub-20% GM; transferring volume to 200mm Fab 2 lifts the per-wafer GM by 5-10 percentage points. Tower has not committed to a Fab 1 closure date but continued runoff is the implicit strategy — likely full closure by 2027-2028.

5.4 SG&A leverage and operating-margin expansion

Tower’s SG&A has scaled sub-linearly with revenue. FY 2025 SG&A ~$60M on $1,566M revenue = ~4% of revenue, vs FY 2022 SG&A ~$70M on $1,679M revenue = ~4.2% of revenue. Modest operating-leverage at the SG&A line; the bigger leverage is at the gross-profit / cost-of-revenue line.

5.5 R&D investment intensity

Tower’s R&D spend is approximately 6-8% of revenue — typical for a specialty foundry with continuous process-development activity. The FY 2025 R&D run-rate is sufficient to fund:

  • PH18 SiPh process generations + EO-polymer integration with LWLG
  • Tower-ST Agrate 65nm BCD evolutionary roadmap
  • Newport Beach Fab 9 SiGe BiCMOS extension
  • Sensors / image-sensor process refinement

R&D is not a forward GM compression risk — Tower’s R&D/revenue ratio is stable through the FY 2025-2028 model period.

6. Mature-node ASP pressure — the structural headwind

For non-SiPh / non-SiGe processes, Tower faces the same ASP-compression dynamic as broader specialty foundry industry:

  1. Chinese capacity additions — SMIC, Hua Hong, Nexchip add ~$10B+ of cumulative trailing-edge specialty-foundry capacity through 2024-2026, putting price pressure on commodity 200mm flows
  2. TSMC mature-node positioning — TSMC’s 28nm/40nm capacity remains a price-leadership benchmark
  3. Customer consolidation — smartphone front-end RF customers (Qualcomm, Skyworks, Qorvo, MediaTek) negotiate aggressively on RF SOI ASPs
  4. Inventory destocking — through 2024 forced customers to underutilize LTAs and renegotiate where possible

Tower’s strategic response (per management commentary):

  • Pivot from mass-LTA pricing to single-source design wins in differentiated specialty processes
  • Concentrate investment in hard-to-replicate processes: PH18 SiPh, SiGe BiCMOS, Tower-ST Agrate 65nm BCD
  • Discontinue legacy 150mm flows; consolidate to 200mm + 300mm
  • Accept modest RF Mobile / Discretes ASP compression in exchange for higher-margin RF Infrastructure ramp

7. Datacenter / AI uplift — the bull-case GM lever

The largest forward-margin opportunity is the RF Infrastructure expansion:

  • $423M FY 2025 (+73% YoY)
  • Includes PH18 SiPh wafer revenue for 1.6T pluggable transceivers
  • Includes SiGe driver IC complementary stack
  • Includes future CPO foundry capacity for co-packaged optics
  • Includes LWLG EO-polymer modulator integration under the March 11 2026 development agreement (commercialization 2027-2028)

If RF Infrastructure reaches 40% of revenue by FY 2028 ($1.15B at $2.84B FY 2028 model) and that segment carries ~40% gross margin, then RF Infrastructure alone contributes ~16 percentage points of consolidated GM — vs the rest of the business at ~30% GM contributing ~18 percentage points. This produces consolidated GM of ~34% on a steady-state basis, with the additional 5-6 percentage points to reach the 39-40% management target coming from utilization tightness + ASP escalation as capacity becomes constrained.

8. Margin sensitivity to FY 2026 revenue scenarios

Simplified margin-sensitivity table (analyst-built ⚠):

FY 2026 revenue scenarioImplied utilizationEstimated GMEstimated Op margin
Bear: $1,650M (+5%)~75%23.5%12.5%
Base: $1,800M (+15%)~82%27.5%15.5%
Bull: $1,950M (+25%)~88%30.5%18.5%
Stretch: $2,050M (+31%)~92%32.5%20.5%

Each ~$200M of incremental revenue at structural mix flows to gross profit at ~75-80% incremental margin (i.e., $150-160M incremental gross profit), expanding consolidated GM by approximately +90-110 bps. This is the core operating-leverage relationship and matches the Q1 2025 → Q4 2025 actual trajectory (revenue +$82M Q1 to Q4; gross profit +$45M Q1 to Q4; +55% incremental margin — slightly below structural rate due to Q1 trough seasonality + Agrate absorption).

9. Open items / backfill queue

  1. Specific Q3/Q4 2025 utilization-rate quote — Q3 2025 was at “80%+” per management commentary ⚠; precise Q4 2025 utilization rate not directly captured in 6-K or call transcripts. TODO ⚠ for FY 2025 20-F extraction.
  2. End-market gross margin disclosure — Tower does not disclose GM by end-market category. Industry triangulation suggests RF Infrastructure 35-45%, RF Mobile 22-26%, Sensors-Displays 25-30%, Power Mgmt 28-32%, Discretes 18-23%. TODO ⚠ for sell-side disaggregation extraction.
  3. Quarter-by-quarter wafer-shipment trail — Tower does not publish wafer shipment counts. Triangulation from revenue / blended ASP would require ASP precision that is not publicly available.
  4. Customer-prepayment ASP impact disclosure — qualitative only; specific dollar quantum and per-wafer ASP impact not disclosed. FY 2025 20-F deferred-revenue note expected.
  5. Tower-ST Agrate utilization rate — not separately disclosed. Joint-venture financial statements should disclose; TODO for FY 2024 + FY 2025 audited 20-F extraction.
  6. PH18 SiPh ASP per wafer — not publicly disclosed. Industry-channel triangulation queued; estimated $8K-15K range based on comparable advanced-process 200mm specialty wafers.
  7. Tower-Intel Fab 11X capacity utilization — not publicly disclosed. Tower’s $300M equipment investment is in installation/qualification flows; first material revenue contribution expected 2026 H2 / 2027 H1.

Sources

  • Q4 2025 6-K release (filed 2026-02-11) — Tower IR ✓ and GlobeNewswire ✓. Q4 2025 GM 26.8%; FY 2025 GM 23.2%.
  • Q3 2025 6-K release (filed 2025-11-10) — Tower IR ✓ and GlobeNewswire ✓. Q3 2025 GM 23.5%; SiPh +70% YoY commentary.
  • Q4 2025 earnings call slides — Investing.com slides ✓. Application-category percentages.
  • Q4 2025 earnings call transcript — Investing.com ✓ and Seeking Alpha ◐. Forward-utilization commentary.
  • FY 2024 20-F (acc. 0001178913-25-001537, filed 2025-04-30) — SEC EDGAR ✓. Customer-concentration baseline.
  • TipRanks Q4 2025 capacity-target summary — TipRanks ✓. Customer-prepayment qualitative disclosure.
  • GFS comparable margins / ASP — GFS margins_and_pricing ✓. Q3 2025 ASP $2,803/wafer; cross-comp benchmark.
  • VIS / DBHiTek / X-FAB ASP comparables — aggregator data (StockAnalysis, Yahoo Finance, Bloomberg ◐).

Cross-references