Tower Semiconductor
Pure-play specialty foundry: manufactures differentiated analog process technologies (RF-SOI, SiGe/BiCMOS, silicon photonics, power management, CMOS image sensors, MEMS) for fabless and IDM customers on a per-wafer basis across 200mm and 300mm fabs in Israel, the US and Japan. Asset-heavy, capacity-reservation and prepayment-backed.
Earnings, margins, COGS & capex
FY2025 revenue $1,566.1M (+9%) with net profit $220M, on ~26-27% gross margins. Growth accelerated into 2026 as silicon photonics revenue tripled YoY and RF/SiGe demand mix shifted; Q1 2026 delivered $413.6M revenue and $65M net profit (+62% YoY), with record Q2 2026 revenue guided to $455M. Balance sheet is a fortress (net cash >$1.2B), funding a ~$920M capacity program and the buyout of the 300mm Fab 7 in Japan. Management reaffirmed a 2028 model of ~$2.8B revenue / $750M net profit and flagged a possible upward revision.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~73¢ is cost of goods and ~11¢ operating expense, leaving ~16¢ of operating profit (~16¢ net).
Revenue trend
Margins
expanding on mix + utilization toward a ~39-40% 2028 target
up sharply YoY; operating profit +96% YoY
rising
target (with ~31.7% operating margin, ~26.4% net margin)
COGS structure
Dominated by fab operating cost: wafer materials (incl. photonics-grade SOI substrates from Soitec), high fixed depreciation on 200mm/300mm capacity, energy, and skilled labor across Israel, US (Newport Beach, San Antonio, Intel Rio Rancho foundry-in-a-fab) and Japan (Fab 7 Uozu). Gross margin is utilization- and mix-sensitive; the shift of RF-SOI/power to lower-capital-intensity 300mm and the higher-value SiPho mix are the margin levers.
Capex
~$920M SiPho + SiGe program ($650M announced + $270M incremental equipment); SiPho wafer-start capacity targeted at >5x the Q4 2025 monthly run-rate by December 2026 with customer-committed consumption. Separately, restructuring transfers full ownership of the 300mm Fab 7 (Uozu, Japan) to Tower (targeting ~4x current Uozu 300mm capacity via an adjacent expansion, subject to METI subsidy) and 200mm Fab 5 to Nuvoton, targeted close April 1, 2027.
Latest earnings
Beat - adjusted EPS $0.65 vs consensus ~$0.55; revenue $413.6M (+15% YoY); net profit $65M (+62% YoY)
Q2 2026 revenue $455M +/-5% (record, +22% YoY, +10% QoQ); reaffirmed 2028 model of ~$2.84B revenue, ~39-40% gross margin, $750M net profit, with management signaling a possible upward revision
- Q1 2026 revenue
- $413.6M
- Q1 2026 gross profit / margin
- $111M / ~26.8%
- Q1 2026 operating profit
- ~$65M (+96% YoY, ~15.7% margin)
- Q1 2026 net profit
- $65M (+62% YoY)
- Q1 2026 EPS
- $0.58 basic / $0.57 diluted GAAP; $0.65 adjusted
- SiPho 2027 contracted
- $1.3B + $290M prepayments
- Net cash
- >$1.2B
Growth drivers
- Silicon photonics for AI/datacenter optical transceivers — revenue tripled YoY; $1.3B of contracted 2027 revenue plus $290M customer prepayments and 50+ active SiPho customers
- SiGe/BiCMOS (+24% YoY) for high-speed optical, radar and RF
- RF-SOI (+12% YoY over the year) for smartphone front-end modules and antenna tuning, transitioning from 200mm to lower-cost 300mm
- Power management (+10% YoY) on 65nm/300mm at Intel Rio Rancho foundry-in-a-fab
- CMOS image sensors (+9% YoY) for medical, industrial and machine-vision
- 300mm capacity expansion (Fab 7 buyout) lowering capital intensity per wafer
Bull & bear
Tower is a scarce, non-Chinese specialty foundry sitting at the center of the AI-datacenter optical-interconnect buildout - a silicon-photonics ramp with contracted, prepaid demand that is re-rating a formerly sleepy analog foundry into a structural growth compounder toward a ~$2.8B/$750M-net 2028 model.
- Silicon photonics revenue tripled YoY with $1.3B of 2027 revenue already contracted and $290M prepaid - visibility most foundries lack
- AI/datacenter optics is a secular driver: as electrical interconnect hits a copper wall, co-packaged optics and pluggable transceivers pull SiPho volume
- Net cash >$1.2B fully funds the ~$920M capacity program and Fab 7 buyout without dilution or leverage
- Margin expansion is largely mechanical: 300mm migration + richer SiPho/SiGe mix drives gross margin from ~27% toward the ~39-40% 2028 target
- Management reaffirmed the 2028 model and flagged an upward revision - a credibility signal after years of steady execution
- Geographic diversification (Israel/US/Japan) makes Tower the China+1 specialty-foundry of choice for RF and power customers
The stock has re-rated to a multiple (~16-17x sales, ~100x+ trailing earnings) that already prices in flawless execution of the SiPho ramp, into a market where GlobalFoundries, UMC and TSMC are all pushing into silicon photonics - leaving asymmetric downside if AI-optics demand, mix, or margin timing disappoints.
- Valuation is priced for perfection: ~118x trailing earnings ($26B / $220M) and ~35x even the 2028 net-profit target ($750M) - multiple compression is the base-case risk on any stumble
- SiPho is contested: GlobalFoundries bought Advanced Micro Foundry (Nov 2025) to become the largest pure-play SiPho foundry; UMC licensed imec iSiPP300 (risk production 2026-2027); TSMC is entering (COUPE) - Tower is not the only supplier
- Contracted commitments are 2027-dated and hyperscaler optics roadmaps (LPO vs DSP, CPO timing, in-house silicon) can shift and strand capacity
- Gross margins remain sub-30% - the ~39-40% 2028 target depends on utilization and 300mm migration executing on schedule (Fab 7 close not until Apr 2027; Fab 2 utilization ~60% during qualifications)
- Heavy capex compresses free cash flow now; a demand air-pocket during the build would hit both earnings and the growth narrative
- Cyclical/geopolitical overhang: smartphone-RF softness (200mm RF-SOI ~-36% QoQ in Q1 2026, expected down in 2026), Israel operational risk, and US-China export policy
What it is worth
Growth-adjusted multiples cross-checked against the company's 2028 model. At ~$26B market cap: ~16-17x FY2025 sales ($1.57B) and ~118x trailing net income ($220M) - i.e., ~35x the reaffirmed 2028 net-profit target ($750M).
AI-optics demand timing slips, competition (GF/UMC/TSMC) compresses SiPho pricing/share, or margin ramp lags - trailing ~100x+ earnings compresses sharply, driving a material derating well below recent highs.
Company executes roughly to the reaffirmed 2028 model (~$2.8B revenue / $750M net); the stock compounds with earnings but the multiple gradually normalizes as growth is delivered rather than anticipated.
SiPho ramps ahead of plan and the 2028 model is revised up; sustained 20%+ growth and margin expansion toward ~39-40% support the premium multiple and further upside from ~$250.
The stock is valued as an AI-silicon-photonics growth story, not a legacy analog foundry. The multiple is only defensible if the SiPho ramp, 300mm margin expansion, and 2028 model land largely as guided; it already discounts a large step-up in revenue and margins. Contracted/prepaid 2027 demand supports the base case; competition and cyclicality frame the downside. Not financial advice.
SWOT
Strengths
- Differentiated specialty-process portfolio (RF-SOI, SiGe, silicon photonics, power, CIS) that competes on capability, not leading-edge geometry
- Fortress balance sheet: net cash >$1.2B, ~$3.0B equity, ~5.6x current ratio - self-funds the capacity ramp
- Prepayment/reservation-backed SiPho demand ($1.3B contracted 2027, $290M prepaid, 50+ active customers) de-risks the ramp
- Geographic manufacturing diversity (Israel, US, Japan) attractive to customers de-risking China exposure
Weaknesses
- Sub-30% gross margins - structurally below leading-edge foundry (TSMC) and GlobalFoundries in some quarters
- Small vs peers ($1.57B FY2025 revenue vs GlobalFoundries ~$7.4B FY2024, TSMC >$90B) - limited scale/pricing leverage
- Utilization-sensitive earnings; 200mm RF-SOI fell ~36% QoQ in Q1 2026 during the 200mm->300mm transition
- Heavy capex cycle compresses free cash flow through 2026-2027
Opportunities
- AI datacenter optical interconnect — silicon photonics is the structural growth vector as copper hits bandwidth/power limits
- 300mm migration (Fab 7 buyout, Intel Rio Rancho) lowers capital intensity and lifts margin
- Reshoring / China+1 sourcing favors a non-Chinese specialty foundry with US and Japan capacity
- Path to the reaffirmed 2028 model (~$2.8B revenue / $750M net) implies a large step-up from today
Threats
- GlobalFoundries scaling SiPho aggressively (acquired Advanced Micro Foundry Nov 2025, now largest pure-play SiPho foundry by revenue); UMC licensing imec's iSiPP300 (risk production 2026-2027); TSMC entering photonics (COUPE) - competition for the same AI-optics dollars
- Customer concentration in a few AI-optics and smartphone-RF accounts; SiPho commitments could slip if hyperscaler optics roadmaps change
- Rich valuation (~16-17x sales, ~100x+ trailing earnings) leaves little room for a ramp miss
- Cyclical semiconductor demand and geopolitical risk (Israel operations, US-China export policy)
Moats, dependencies & bottlenecks
Moats
Moderate-to-strong Customers co-develop and qualify products on Tower's RF-SOI/SiGe/SiPho/power flows; re-qualifying on another foundry is costly and slow, creating switching costs.
$1.3B contracted 2027 + $290M prepaid + 50+ active customers gives a lead, but GF/UMC/TSMC are investing - durability depends on staying ahead on capacity and yield.
Non-China footprint is a genuine sourcing advantage for RF/power/defense-adjacent customers amid China+1 and export controls.
Net cash >$1.2B lets Tower self-fund capacity when smaller specialty players cannot - but not a durable moat against GF/TSMC-scale balance sheets.
Dependencies
AI-datacenter optical-transceiver demand (hyperscalers via optics OEMs) Demand / end-market The SiPho growth thesis rides on continued explosive AI-interconnect buildout and pluggable/co-packaged optics winning; a roadmap shift (in-house silicon, LPO) could reduce pull.
Soitec is an approved volume supplier of photonics-grade SOI to the industry (Tower, GF, TSMC) - a concentrated upstream input for both SiPho and RF-SOI.
Manufacturing partner 65nm power/RF-SOI on 300mm runs in Intel's New Mexico fab under a foundry-in-a-fab arrangement; capacity ramp depends on the partnership holding.
Partner / counterparty The Fab 7 full-ownership / Fab 5 transfer (target close Apr 1, 2027) hinges on mutual long-term supply agreements with Nuvoton (NTCJ) and regulatory/subsidy approvals.
Smartphone RF front-end demand (Qualcomm/Skyworks/Qorvo ecosystem) Demand / end-market RF-SOI is handset-cycle exposed - 200mm RF-SOI fell ~36% QoQ in Q1 2026 during the 200mm->300mm transition, with management guiding RF mobile down in 2026.
The ~$920M capacity program and Fab 7 buildout depend on timely equipment delivery.
Advantages
- Early scale and prepaid, contracted demand in AI-driven silicon photonics
- Broad differentiated analog process menu that leading-edge foundries deprioritize
- Net-cash balance sheet self-funding a large capacity cycle without dilution
- Non-China Israel/US/Japan manufacturing footprint aligned with reshoring and export-control tailwinds
- Deep customer co-development / qualification relationships that raise switching costs
Weaknesses
- Sub-30% gross margins and small revenue base relative to GlobalFoundries and TSMC
- Earnings sensitivity to utilization and end-market cycles (handset RF softness)
- Free cash flow compressed by the heavy 2026-2027 capex cycle
- Elevated valuation leaves little margin for execution error
- Concentration risk in a handful of large AI-optics and RF customers
Bottlenecks
- SiPho and 300mm capacity availability — capacity is heavily reserved/prepaid through 2028, so incremental upside is gated by the buildout pace (Fab 7 expansion multi-year, close targeted Apr 2027)
- Photonics-grade SOI substrate supply (Soitec) as an upstream constraint
- Fab utilization / mix timing — gross margin is levered to filling 300mm capacity as RF-SOI migrates off 200mm (Fab 2 utilization ~60% during qualifications)
- Skilled fab labor and equipment lead times across three geographies during a simultaneous multi-site ramp
Top signals & trends
Top signals
Contracted, prepaid demand is unusually high visibility for a foundry and de-risks the ramp.
Sequential and YoY acceleration confirms the ramp is landing in reported numbers, not just backlog.
Signals confidence; an upward revision would extend the growth narrative.
Shows near-term lumpiness and handset-cycle exposure even amid the SiPho boom.
Competitive intensity in Tower's key growth vector is rising.
A crowded, richly valued AI-adjacent trade - sentiment-sensitive.
Trends
High positive · The structural driver of Tower's SiPho ramp as electrical interconnect hits bandwidth/power limits.
Lowers capital intensity per wafer and supports the path to a ~39-40% 2028 gross margin; near-term transition dislocation in RF-SOI.
Non-Chinese Israel/US/Japan footprint wins share from customers de-risking China.
GF, UMC, TSMC all investing - validates the market but pressures Tower's early lead.
RF-SOI demand swings with the handset cycle; a source of quarter-to-quarter lumpiness.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Approved volume supplier of photonics-grade and RF SOI substrates to the industry (Tower/GF/TSMC) - key upstream input for SiPho and RF-SOI.
Lithography equipment for the 300mm capacity buildout.
Deposition/etch/process equipment supplier for fab expansion.
Counterparty in the Japan fab restructuring - NTCJ receives 200mm Fab 5, provides mutual long-term supply agreements (target close Apr 1, 2027).
Optical/networking silicon and transceiver ecosystem - representative of the AI-optics demand pulling Tower's SiPho (specific attribution not disclosed by Tower).
Optical DSP / interconnect silicon for AI datacenters - part of the SiPho transceiver value chain (attribution not disclosed).
Optical transceiver / photonics component maker in the datacenter-optics chain that consumes SiPho foundry capacity (attribution not disclosed).
Datacenter optical components / transceivers - representative SiPho end customer (attribution not disclosed).
Smartphone RF front-end supplier - representative of Tower's RF-SOI/SiGe handset demand (attribution not disclosed).
RF front-end / handset customer segment for RF-SOI and SiGe (attribution not disclosed).
Closest comparable specialty foundry (~$7.4B 2024 revenue); competes directly on RF-SOI, SiGe and silicon photonics; acquired Advanced Micro Foundry (Singapore, Nov 2025) to become the largest pure-play SiPho foundry by revenue.
Large mature-node/300mm foundry pressuring Tower on power and display drivers; licensed imec's iSiPP300 silicon-photonics process (Dec 2025) with risk production targeted 2026-2027.
Scale leader entering silicon photonics / co-packaged optics (COUPE, verification complete); an approved Soitec photonics-SOI customer - a long-term threat in the highest-value SiPho tier.
Taiwan specialty foundry competing in power management and mature analog; building 300mm capacity (VisionPower JV with NXP).
China context only - state-backed mature-node/analog foundry pressuring price-sensitive segments; named for landscape, not a buy/own call.
China context only - specialty analog/power/embedded-NVM foundry; competitor in mature analog, named for landscape, not a buy/own call.