Tower Semiconductor (TSEM) — Graham Screen

Layer: L5-SpecialtyFoundry | Date: 2026-05-28 | Analyst: Claude (Sonnet 4.6)


PRICE CHECK (ABBREVIATED TRIGGER)

Metric Value
Current Price $274.22
52w Low $37.48
52w High $302.86
52w Range $37.48 – $302.86
YTD Return +125.3% (from $121.74)
Distance from 52w High 9.5%
Trigger YTD > 80% AND within 10% of high → ABBREVIATED

Both conditions met. Abbreviated screen applies. The stock has re-rated massively: from a post-deal-break trough in the low $30s (52w low $37.48) to near all-time highs at $274. This is not a Graham depressed-asset opportunity — it is a momentum/AI re-rating story that has already played out.


ABBREVIATED SCREEN

Valuation Snapshot

Metric Value Graham Signal
Trailing P/E 126x RED — extreme overvaluation
Forward P/E 47.8x AMBER — still expensive; implies ~$5.74 FY2026 EPS
P/B 10.6x RED — Graham limit is 1.5x (defensive)
EV/EBITDA 55.7x RED
Market Cap $32.6B
Enterprise Value $29.9B (net cash: ~$1.3B)

Note on trailing P/E: FY2023 net income was inflated by the $353M Intel termination fee. Stripping this out gives FY2023 adjusted net income of approximately $165M. FY2024 was $208M, FY2025 $220M. The trailing P/E on clean operating earnings is still ~120x+ — not a misread. The $2.17 trailing EPS (yfinance) appears to use recent quarterly annualization; at $220M NI / ~119M diluted shares the run-rate EPS is ~$1.85. Forward EPS of $5.74 uses 2026 consensus — that is the real number to watch.


Graham 7 Defensive Filters

# Filter Threshold TSEM Result
1 Adequate size >$2B revenue (industrial) $1.57B (FY2025) MARGINAL FAIL
2 Financial condition Current ratio >2x 6.5x PASS
3 Earnings stability No deficit last 10 yrs Profitable FY22-25 (FY23 inflated by fee); no deficits visible PASS (conditional)
4 Dividend record Uninterrupted 20 yrs No dividend FAIL
5 Earnings growth +33% over 10 yrs FY25 $220M vs FY22 $265M: NEGATIVE on clean basis FAIL
6 P/E ≤15x trailing 126x FAIL
7 P/B ≤1.5x 10.6x FAIL

Graham Score: 1/7 (PASS only on current ratio)

This is not a Graham value candidate at current price. It is a momentum/growth/AI story fully priced at the upper end of a re-rating cycle.


Graham IV Sensitivity Table

Trailing EPS: $2.17 (distorted; clean op. EPS ~$1.85)

Growth Assumption Graham IV vs $274.22
g = 0% $18.45 -93.3%
g = 3% $31.46 -88.5%
g = 5% $40.14 -85.4%
g = 7.5% $50.99 -81.4%
g = 10% $61.84 -77.4%

Forward EPS: $5.74 (consensus FY2026E)

Growth Assumption Graham IV vs $274.22
g = 0% $48.77 -82.2%
g = 3% $83.20 -69.7%
g = 5% $106.15 -61.3%
g = 7.5% $134.84 -50.8%
g = 10% $163.53 -40.4%

Break-even growth to justify $274 price:

Per shr-003: the large gap between trailing (~58.9%) and forward (~19.6%) break-even reveals the market is pricing a massive earnings inflection (trailing EPS $2.17 → forward $5.74 = ~164% jump), not sustained high growth. This is an inflection-stage pricing pattern. At 19.6% required perpetual growth on forward EPS, the stock is pricing in strong execution of its SiPho ramp, capacity build, and AI tailwinds — with very little margin for error.


Post-Intel Strategic Position (shr-thesis items a–d)

(a) Post-INTC standalone story The Intel acquisition (announced Feb 2022, terminated Aug 2023 after China regulatory failure) resulted in Tower receiving a $353M termination fee — roughly 1.5 years of historical operating income in one payment. The standalone story has re-accelerated: Q1 2026 revenue +15% YoY to $414M; Q2 2026 guidance is "highest revenue in company history" at mid-range $455M ±5% (+22% YoY). The company is executing quarterly sequential growth throughout 2026.

(b) Israel geopolitical discount — status: largely re-rated away The geopolitical discount that compressed the stock in FY2024 (range $22–$40) is gone. Israel-conflict concerns were the primary reason the stock sat at ~$30–$37 for much of FY2024 despite solid operations. The AI/SiPho narrative from Q4 2025 onward has completely overridden that concern. S&P Maalot upgraded the outlook to Positive in May 2026. The geopolitical risk remains real (all major fabs in Israel and Japan Uozu) but the market is no longer pricing it as a discount — it is pricing it as irrelevant given the AI demand pull.

(c) Capacity expansion — $920M SiPho build Tower announced a $920M CapEx program to quintuple SiPho production capacity by December 2026, targeting the Japan Fab7 (Uozu) expansion to 4x current 300mm capacity. An additional $270M incremental investment was announced to support demand. This is the core bull thesis: TSEM as the specialty foundry for silicon photonics used in AI data center optical interconnects.

(d) Intel investment status — CANCELLED, redirected to Fab7 The September 2023 Intel manufacturing agreement (Tower using Intel New Mexico 300mm capacity) has been cancelled by Intel — Intel notified Tower of its intention not to perform and parties are in mediation. The volume being transferred from Intel NM has been redirected to Tower's own Fab7 Japan. This is a risk mitigation success story (no Intel dependency for capacity) but also signals Intel's broader strategic retreat from foundry.

(e) NVIDIA partnership (new catalyst, Feb 2026) Tower announced a collaboration with NVIDIA for 1.6T silicon photonics optical modules for NVIDIA networking protocols (Feb 5, 2026). This is the narrative catalyst that drove the stock from ~$120 YTD-start to $274+. Tower's SiPho platform enables 2x the data rate vs. prior-gen solutions. The partnership is real but its revenue contribution is forward-looking (SiPho ramp targets Dec 2026 completion).


Key Risks

  1. Execution risk on $920M SiPho CapEx: If qualification slips past Dec 2026, the bull thesis breaks. TSEM has never managed a CapEx program of this magnitude before.
  2. Intel mediation outcome: If mediation turns adversarial, there could be legal overhang and potential customer disruption during transition.
  3. Concentration in Israel/Japan: Both operational clusters face geopolitical/natural disaster tail risk with no geographic diversification.
  4. Revenue quality — Intel fee distortion: FY2023 earnings were significantly inflated. Clean EPS trajectory is modest: FY22 $264M → FY23 $165M op-adj → FY24 $208M → FY25 $220M NI. The inflection is Q1/Q2 2026 driven.
  5. No insider buying despite +125% YTD: Zero insider purchases in last 12 months (per shr-002: insider buying is the strong signal; absence at these prices is at minimum neutral-to-cautionary).
  6. Short interest building: +131% YoY increase in short interest (now 3.9% float, DTC 1.8). Informed shorts may see the same valuation stretch I do.
  7. Value trap risk (shr-017): This is NOT a value trap — it's the opposite. It is a growth stock priced for perfection. The risk is paying a 50x forward multiple for a specialty foundry that historically earned 10-15% margins.

Financial Health Snapshot

Metric Value
Gross Margin (TTM) 24.8%
Operating Margin (TTM) 15.6%
Net Margin (TTM) 15.1%
ROE 8.7%
Current Ratio 6.5x
D/E Ratio 5.2% (very low debt)
Total Debt $156M
Net Cash ~$1.3B
Free Cash Flow (TTM) $129M
Operating Cash Flow $812M

Balance sheet is strong. Net cash of $1.3B with only $156M total debt is exceptional. The $920M SiPho CapEx will consume this buffer and likely require financing, but leverage will remain manageable given the scale. ROE of 8.7% on clean earnings is decent but not exceptional for a specialty semiconductor company — margins have room to expand as SiPho scales.


Analyst Consensus

Metric Value
Recommendation Strong Buy (6 analysts)
Mean Price Target $313.83
Current Premium to Target -12.6% (stock below mean PT)
High PT ~$350+
Low PT ~$230 (Benchmark)

Only 6 covering analysts — thin coverage for a $32B market cap company. The mean PT of $314 implies modest upside from $274 and suggests the consensus view is fully priced but not outright expensive at current levels assuming the SiPho ramp executes. This is purely a growth/execution story, not a value screen.


VERDICT

🔴 NOT A GRAHAM VALUE CANDIDATE

TSEM is a high-quality specialty foundry with legitimate AI tailwinds and strong execution, but it is the inverse of what this screen is looking for. The stock has:

The original thesis question was whether geopolitical/strategic uncertainty still kept it depressed. The answer is: No. The discount has been fully and then some eliminated. The AI + NVIDIA partnership narrative has driven the stock from ~$37 (trough) to $274, a ~7x move in roughly 18 months.

For the EUR 3,000 satellite: No allocation. This is a momentum/AI story at momentum/AI prices. It requires a view on whether TSEM can execute a $920M SiPho build to justify a 20%+ perpetual growth assumption — that is a venture capital bet, not a Graham value bet.

Revisit condition: If TSEM retraces to the $100–$120 zone (forward P/E ~17–20x on $5.74E, forward break-even growth drops to 5–7%) following an execution stumble on SiPho or a broader semi selloff — at that level the business quality (net cash, low debt, specialty moat in RF/power/SiPho) would warrant a full Stage 2 screen.


VERDICT LINE: 🔴 PASS — AI/momentum re-rating complete, price fully reflects best-case SiPho execution at 126x trailing / 48x forward P/E; Graham IV requires 19.6% perpetual growth on forward EPS; zero insider buying; revisit below $120.