Revenue 2025
$1.57B
+9.0% Y/Y
Net Income 2025
$220.5M
+6.1% | EPS $1.94
Operating Margin 2025
~14.1%
2024: 14.5%
Cash + Deposits
$1.15B
Net Cash ~$1.0B
FCF 2025
($48.9M)
Heavy CAPEX ($444M)
Debt/Equity
4.6%
minimal leverage
Market Cap
$23.79B
StockAnalysis · 27/04/2026
P/E Ratio
108.77
TTM · StockAnalysis
EV/EBITDA
45.04
TTM · StockAnalysis
Net Debt
($990M)
Net Cash · StockAnalysis 27/04/2026
Gross Margin
23.2%
FY25 · Earnings release 11/02/2026
1 Company Profile
Tower Semiconductor is an Israeli chip manufacturer specialising in third-party manufacturing (Pure-Play Foundry) for analog and specialty semiconductors (High-Value Analog Specialty). The company is traded on NASDAQ and on the Tel Aviv Stock Exchange and is a constituent of the TA-35 index. Tower is regarded as one of the "Big Four" of Israeli semiconductors — together with Nova (NVMI), Camtek and AudioCodes — and is the largest Israeli semiconductor manufacturer by revenue in the foundry segment. Headquartered in Migdal HaEmek. Unlike TSMC, which focuses on leading-edge digital logic, Tower specialises in specialty analog/RF/sensor/power processes — a different competitive space and margin profile.
| Target Market | Strategic Role |
|---|
| Automotive (EV + ADAS) | Sensor content, Power Management, communications |
| Data Centers / AI | SiPho and SiGe for optical communications |
| Industrial / IoT | Sensors, BCD, Mixed Signal |
| Mobile / Infrastructure | RF, wireless communications |
| Aerospace & Defense | Specialty chips (Rad Hard, SiGe for radar) |
| Medical | Medical sensors, portable instrumentation |
Source: SEC EDGAR — 20-F FY2025
2 Key Financial Observations
This summary is not a recommendation. It is a factual list of key financial metrics.
Performance — 3 Years ($M)
| Year | Revenue | Net Income | EPS | Operating Margin |
|---|
| 2023* | ~1,423 | 518.5 | $4.66 | — |
| 2024 | 1,436 | 207.9 | $1.85 | 14.5% |
| 2025 | 1,566 | 220.5 | $1.94 | 14.1% |
Cash Flow & Balance Sheet ($M)
| Metric | 2024 | 2025 |
|---|
| OCF | 448.7 | 395.5 |
| Capex | (436.2) | (444.4) |
| FCF | 12.5 | (48.9) |
| Cash + Deposits | — | 1,151.9 |
| Total Assets | 3,080.5 | 3,322.3 |
| Shareholders' Equity | 2,653.3 | 2,919.3 |
| Long-Term Debt | 132.4 | 133.4 |
Missing data: Backlog, revenue breakdown by customer/platform, normalisation for 2021-2022. *2023 includes $353M Intel contract termination fee (one-time).
Capex and Capex Intensity
3 Industry & Competitive Context
Analog/Specialty Foundry — a different segment from pure digital Foundry (TSMC). Emphasis on differentiated technology, mature nodes (130nm+), and technological value. Cyclicality more moderate than Leading Edge. Supportive trends: vehicle electrification + ADAS, AI/Data Centers (SiPho), IoT, government incentives (CHIPS Act), rising defence budgets.
| Competitor | Country | Difference vs. Tower |
|---|
| TSMC | Taiwan | The world's largest Foundry — Tower does not compete on leading edge |
| GlobalFoundries (GFS) | USA | Foundry without leading edge — partial direct competitor |
| UMC | Taiwan | Mature foundry 28nm+ — direct competitor |
| SMIC | China | Mature nodes — rising competitor with export restrictions |
| X-FAB | Germany | Dedicated analog Foundry — closest competitor |
| Vanguard (VIS) | Taiwan | 200mm Foundry (TSMC subsidiary) |
4 Risk Factors
| Risk | Context |
|---|
| Intel contract cancellation (Capacity Corridor) | Intel announced it will not execute the New Mexico capacity agreement — Italy provides a partial alternative |
| Negative FCF 2025 | ($48.9M) — a result of heavy Capex; if the investment cycle does not moderate, liquidity will come under pressure |
| Dependence on a low tax rate | Effective tax rate of 9% depends on continued "preferred enterprise" benefits — a policy change in Israel would have a material impact |
| Capex Intensity | 28-31% of revenue — sustained heavy Capex cycle limits the ability to distribute value to shareholders |
| Competition from X-FAB and Vanguard | In the analog/specialty niche, specific competitors can take market share |
| TPSCo Japan — negative NCI | Subsidiary (51%) accumulating losses — point to monitor |
| Export restrictions to China | May reduce the potential Target market |
| FX exposure | Israeli + Japanese cost base, revenue in USD |
5 Analytical Lens — The Questions We Ask
In professional company analysis, the question is not "is this good?" but rather "through which lenses must this company be examined so that we do not miss what matters most?" At Bakshi Finance, every analysis passes through six lenses.
This framework is intended to structure analysis, not to produce an investment conclusion.
Growth
Revenue grew 9% in 2025. How much is the industry recovery versus Tower-specific growth drivers (SiPho, automotive)? What is the impact of the Intel cancellation on 2026-2027 outlook? How will Italy fab expansion translate into sales?
Profitability
Operating margin 14% — low versus leading foundries (TSMC ~40%). What is the reasonable ceiling in Tower’s specialty-foundry model? How much of negative FCF is temporary CAPEX versus a structural issue?
Leverage
Debt/Equity 4.6% — minimal leverage, $1B excess cash. Why doesn’t the company deploy cash for buybacks or dividends? Is it future M&A? Heavier CAPEX? What are the ROE implications of idle cash?
Competitive Position
Tower leads in SiPho — critical for AI. How meaningful is the advantage? How does SMIC (with government support) affect the picture? What is the geopolitical advantage of 4 fabs across different countries (geographic diversification)?
Management Quality
Management handled the failed Intel acquisition (2023) in an orderly way and completed 300mm qualification. How correct is the decision to launch heavy CAPEX now (at the cost of negative FCF)? What is the post-CAPEX-cycle strategy?
Business Complexity / Risk
Tower operates in 4 countries, 8 technology platforms, and dozens of customers. How should an investor value a company whose results depend on both the global semi cycle and individual fab performance? What is the leading indicator of a cycle shift?
6 Scenario Framework
Scenarios are descriptive, not predictive. They outline possible conditions, not expected outcomes.
These scenarios carry no probability assessment, no preferred direction, and no expectation regarding which, if any, will materialise.
Constructive Scenario — if the following conditions hold:
The investment cycle moderates (CAPEX falls from $440M to $300M over 2-3 years), the Italy fab expansion completes on schedule and becomes profit-accretive, SiPho demand continues to grow with AI, the automotive market continues to demand more sensors, and the Israeli tax rate remains low. Under these conditions, FCF returns to materially positive ($100-150M+), revenue grows 8-12% annually, operating margins rise toward 16-18%.
Base Scenario — if current trends continue:
Revenue grows 5-8% annually, CAPEX remains elevated (>$400M), FCF around 0 or slightly positive. Operating margins stable around 14-15%. The Intel cancellation is addressed through alternative customers but shortens the planned capacity ramp.
Adverse Scenario — if the following risks materialise:
A global semi-cycle downturn, AI demand moderates and pressures SiPho, the Intel cancellation leaves capacity underutilised in Italy, the effective tax rate rises to 15-20%, or FCF remains negative requiring debt issuance. Under these conditions, net income contracts and equity growth slows.
Scenarios describe conditions, not forecasts. There is no preferred direction and no probability assessment expressed in this framework.
7 How to Think About This Company
Tower is not a standard "semiconductor equity" — it is a foundry where technological intelligence matters more than scale. While TSMC competes on the smallest nodes (3nm, 2nm) and on massive scale, Tower competes on domains that require specialised know-how — SiPho for optical communications, SiGe for radar, BCD for power management. This distinction is fundamental: one must not compare Tower’s margins (~14%) to TSMC’s (~40%) — these are different business models. The real question in analysing Tower is not "is the model profitable" (it is), but rather "can Tower capture a growing share of AI/Data Centers through SiPho, or will it be squeezed by larger suppliers?"
The critical variables to monitor are three. First, results of the Italy (300mm) expansion. The investment is large relative to the company, and the Intel cancellation left a capacity void. The question is how the 300mm migration in Italy progresses and which customers take up the capacity. Second, the SiPho growth rate. Tower is one of the global leaders, and AI-infrastructure demand is growing quickly. If Tower’s growth tracks the market (15-25% annually), it will reshape the profitability profile. Third, the FCF path. Negative FCF in 2025 ($-49M) is a result of heavy CAPEX, not a bad business. The question is when the CAPEX cycle moderates: if by 2027-2028, FCF returns materially positive; if it persists to 2029-2030, value distribution is deferred.
Where the analysis may go wrong. First error — comparing 2023 EPS ($4.66) to 2025 EPS ($1.94) and concluding "dramatic decline". In reality, 2023 included a $353M Intel termination fee — a one-time income that does not recur. Normalised 2023 EPS was similar to 2025. Second error — valuing the balance sheet only by cash ($1.15B). A large portion of cash is earmarked for future CAPEX; it is not "excess cash" available for distribution. Third error — assuming a 9% tax rate will continue. It depends on "preferred enterprise" benefits in Israel; a policy change can raise the rate to 15-23% and reduce EPS by 7-15%.
What distinguishes professional analysis of Tower from headlines. Headlines on Tower speak of "the Intel cancellation" or "Israeli company on NASDAQ". Professional analysis addresses three things: (a) the ratio of utilised capacity to installed capacity — and how long until high utilisation; (b) the difference between revenue from sophisticated products (SiPho, SiGe) and commodity-like revenue — the difference between 30% and 5% margins; (c) the path back to positive FCF. These are not what one buys or sells — they are what one asks before deciding.
The difference between surface-level analysis and professional thinking often lies in the variables that are not immediately visible.
The difference between surface-level analysis and professional thinking often lies in the variables that are not immediately visible.
8 Sources & Data
| # | Source | Date | Type |
|---|
| 1 | SEC EDGAR — 20-F FY2025 (TSEM) | March 2026 | Official — SEC |
| 2 | SEC EDGAR — 20-F FY2024, FY2023 | March 2024-2025 | Official — SEC |
| 3 | towersemi.com — IR | Quarterly | Official — company website |
| 4 | maya.tase.co.il — TSEM | April 2026 | Official — Stock Exchange |
| 5 | NASDAQ.com — TSEM quote | April 2026 | Official — Stock Exchange |
Missing: Backlog, revenue breakdown by customer/platform, normalisation for 2021-2022. *2023 includes $353M Intel contract termination fee (one-time).
