Skip to content
primarysourced Photonics sector Tower
TSEM
~12 min read · 2,870 words ·updated 2026-04-29 · confidence 52%

Public-comp valuation — specialty-analog foundry peers

As of: 2026-04-29. Spot prices reflect 2026-04-28 close. Tower at $191.54 × 111.76M shares = $21.41B market cap.

Confidence legend: ✓ verified-primary (10-K / 20-F / 6-K) · ◐ aggregator-derived (Yahoo / StockAnalysis / Bloomberg / market data) · ⚠ inferred / cycle-distorted

1. The peer set — which companies are valid Tower comparables

Tower sits in the most concentrated specialty-foundry comp position of any merchant foundry: it is the only pure-play 200mm specialty-foundry-with-300mm-Agrate at scale that is also publicly listed in the US (NASDAQ:TSEM). The closest peer set:

CompTickerGeographyProcess focusComp validity
Vanguard International Semi (VIS)5347.TWTaiwan200mm specialty (RF SOI, BCD, MEMS, image sensors)Strongest comp — direct overlap on 200mm specialty, RF SOI, image sensors
DBHiTek000990.KSKorea200mm/300mm specialty (BCD, CIS, RF)Strong comp — Korean equivalent with similar mix; smaller scale
GlobalFoundries (GFS)GFSUS/Germany/Singapore200/300mm specialty (RF SOI, FDX, Fotonix SiPh)Comp at scale — direct SiPh competitor (Fotonix vs PH18)
United Microelectronics (UMC)UMCTaiwan28nm + trailing-edge 200mmComp for trailing-edge mature; less specialty overlap
Semiconductor Mfg Intl (SMIC)0981.HKChina14nm + trailing-edgeComp with China-export-restriction discount
X-FAB Silicon FoundriesXFAB.PAGermany / France200mm analog/mixed-signal (auto, SiC)Comp for European analog specialty
Hua Hong Semiconductor1347.HKChina200mm + 12-inch trailingComp; China-discount distortion
Powerchip Semi (PSMC)6770.TWTaiwan200/300mm DRAM + foundryPartial comp
TSMCTSMTaiwanLeading-edge dominantLimited (mix differs materially); only small specialty-fab overlap
Tower SemiconductorTSEMIsrael/US/Italy/Japan200mm specialty + 300mm Agrate (PH18 SiPh / SiGe)self

VIS and DBHiTek are the cleanest pure-play comps for Tower. Both are 200mm specialty-foundry operators at sub-$2B revenue scale; both serve overlapping end-markets (RF SOI, image sensors, BCD power). GFS is the direct silicon-photonics competitor (Fotonix vs PH18) but operates at 4× the revenue scale. UMC / SMIC introduce trailing-edge-CMOS mix distortion. TSMC is not the right primary comp — its leading-edge dominance and $110B revenue scale produce structurally different multiples.

2. Trading multiples — peer-table snapshot

CompSpot (2026-04-28 ◐)Mkt cap ($B ◐)Net debt / (cash) ($B ◐)EV ($B ◐)TTM revenue ($B)EV / Revenue (TTM)EV / EBITDA (TTM)P / E (TTM)P / B
Tower (TSEM)$191.54 ✓21.4 ✓(1.04) ✓20.41.57 ✓13.0× ✓~50× ⚠~97× ⚠~8× ⚠
Vanguard Intl Semi (VIS)NT$112 ◐~5.5 ◐(1.0) ◐~4.5~1.4 ◐3.2× ◐8.0× ◐18× ◐2.5× ◐
DBHiTekKRW ~80,000 ◐~$2.5 ◐(~0.5) ◐~2.0~$1.0 ◐2.0× ◐6.5× ◐12× ◐1.4× ◐
GlobalFoundries (GFS)$59.49 ✓33.1 ✓(2.8) ✓30.36.79 ✓4.5× ✓12.9× ✓37× ⚠~2.0× ⚠
UMC~$8.50 ◐~21 ◐(3.5) ◐~17.5~7.5 ◐2.3× ◐5.5× ◐11× ◐1.4× ◐
SMICHK$53 ◐~$57 ◐(~3) ◐~54~9.0 ◐6.0× ◐13.0× ◐50× ⚠1.4× ◐
X-FAB Silicon€4.50 ◐~0.7 ◐0.1 ◐~0.8~0.95 ◐0.8× ◐4.5× ◐14× ◐0.6× ◐
Hua Hong SemiHK$36 ◐~$8 ◐(~2) ◐~6~2.2 ◐2.7× ◐9.0× ◐35× ⚠0.6× ◐
TSMC~$215 ◐~1,115 ◐(15) ◐~1,100~110 ◐10.0× ◐14.5× ◐28× ◐7.5× ◐

EV-bridge for Tower: Market cap $21.41B + Total debt $0.18B − Cash + ST deposits $1.22B ≈ EV $20.37B (using FY 2024 YE balance; FY 2025 exit balance pending).

TTM EV/Revenue calc: $20.4B / $1.57B (FY 2025 actual) = 13.0×.

TTM EV/EBITDA: Adjusted EBITDA estimate FY 2025 ⚠ = Op profit $194M + D&A ~$300M ≈ $494M EBITDA. EV/EBITDA = $20.4B / $0.49B ≈ ~41×. ⚠ — depending on D&A assumption, range is 35-50×.

TTM P/E: $191.54 / $1.94 diluted EPS = 98.7× trailing.

The Tower premium is enormous and analyst-grade. At 13.0× TTM EV/Revenue, Tower trades at approximately 4× the multiple of its closest specialty peer VIS (3.2×) and 6-7× the multiple of DBHiTek (2.0×). The premium is even more extreme on EBITDA/PE bases. The market is pricing Tower as if the FY 2028 financial-model target ($2.84B revenue, 39-40% GM, 31.7% Op margin) is partially de-risked — i.e., as if the SiPh growth narrative + envelope-tracker recovery is mostly underwritten by the design-win backlog and customer prepayments.

3. Where does Tower trade vs. comps?

3.1 EV / Revenue — Tower at 13.0× is at a structural AI-photonics premium

Tower’s 13.0× EV/Revenue is:

  • +9.8 turns above VIS (3.2×) — VIS is the cleanest comp; the premium is essentially the AI-photonics + SiPh growth-narrative re-rating
  • +11.0 turns above DBHiTek (2.0×) — DBHiTek lacks SiPh exposure; trades on cyclical specialty-foundry multiples
  • +8.5 turns above GFS (4.5×) — both Tower and GFS have AI-photonics exposure (PH18 vs Fotonix); the Tower premium reflects lower revenue base + steeper implied growth trajectory
  • +10.7 turns above UMC (2.3×) — UMC is structurally lower-margin / lower-growth
  • +7.0 turns above SMIC (6.0×) — SMIC commands its own mature-node premium from China-domestic-supply demand

The “AI-photonics premium” embedded in Tower’s multiple. At a $21.4B mkt cap, applying the VIS comp multiple (3.2×) to Tower’s $1.57B revenue would imply a market cap of approximately $5.0B — i.e., $45 per share. The current $191.54 share price reflects an incremental ~$16.4B / ~$147 per share of “AI-photonics revaluation” above the specialty-foundry cyclical baseline. The bull thesis must justify that $147/sh of incremental valuation through SiPh-revenue ramp + margin expansion + design-win-cohort conversion.

3.2 EV / EBITDA — Tower at ~41× is in growth-multiple territory

Tower’s TTM EV/EBITDA of ~41× ⚠ (high end of estimate range) is:

  • +33 turns above VIS (8.0×) — extreme premium; reflects the FY 2028 EBITDA inflection that the SiPh ramp is supposed to deliver
  • +34.5 turns above DBHiTek (6.5×)
  • +28 turns above GFS (12.9×) — even GFS, which has the same AI-photonics narrative, is ~3× lower EBITDA multiple
  • +26.5 turns above TSMC (14.5×) — Tower commands a higher EV/EBITDA than TSMC, which is structurally questionable

Forward EV/EBITDA at the FY 2028 model. If Tower hits the FY 2028 target of $2.84B revenue × 31.7% Op margin = ~$900M Op profit + ~$500M D&A = ~$1.4B EBITDA. At current EV $20.4B / FY 2028 EBITDA $1.4B = 14.6× FY 2028 EV/EBITDA — in line with TSMC and modestly above the specialty peers. The current multiple is only justified if FY 2028 model materializes mostly intact.

3.3 P/E — Tower at 98.7× trailing / ~30-35× FY 2027E is the growth-stock premium

Trailing P/E of ~98.7× is clearly impractical as a value-anchor — it reflects the depressed FY 2025 EPS ($1.94) ahead of the FY 2026-2028 ramp.

Forward EPS estimates:

  • FY 2026E EPS (sell-side consensus, ⚠ aggregator-only): $2.50-3.00
  • FY 2027E EPS: $4.00-5.00
  • FY 2028E EPS at the model target: $900M Op profit × ~85% (taxes/interest) / 114M shares ≈ $6.70 ⚠

At $191.54 current price:

  • FY 2026 fwd P/E: ~70-75× (still very high)
  • FY 2027 fwd P/E: ~38-48×
  • FY 2028 fwd P/E: ~28-30×

For Tower’s current valuation to be reasonable, the FY 2028 model EPS of ~$6.70 must be realized, AND the market must pay 28-30× FY 2028 EPS at that point — which assumes (a) SiPh revenue + margin expansion delivers, (b) RF Mobile/specialty cyclicality doesn’t compress earnings during ramp, (c) Israel-geopolitical-risk doesn’t increase the cost-of-equity, and (d) competing SiPh ramps (GFS / Intel / TSMC) don’t compress merchant SiPh ASPs faster than volume scales. All four conditions are plausible but none is de-risked.

3.4 P/B — Tower at ~8× is the highest in the comp set ex-TSMC

MetricValue
Total equity (FY 2024 ✓)$2,662M
Shares outstanding111.76M
Book value per share$23.82
Spot ($)$191.54
P/B8.04×

For context:

  • TSMC: ~7.5× (leading-edge premium)
  • Tower: 8.04× (AI-photonics premium)
  • GFS: ~2.0×
  • VIS: ~2.5×
  • UMC: ~1.4×

Tower trades at the highest P/B in the specialty-foundry comp set, slightly above even TSMC. This reflects the asset-light + growth-narrative combination. The structural justification: Tower is generating ~8% return on equity (FY 2025 net income $220M / equity $2.66B ≈ 8.3%) — implying a P/B of ~8× requires the market to expect ROE expansion to 20%+ by FY 2028 (consistent with the operating-margin target of 31.7%). That ROE expansion is the implicit valuation underwrite.

4. Why does Tower trade at a 4-7× turn premium to specialty peers?

The Tower premium of approximately +9-11 EV/Revenue turns vs VIS / DBHiTek is justifiable only on aggressive interpretations of:

  1. AI-photonics positioning — PH18 silicon-photonics process is the ONLY 200mm SiPh foundry process with announced 1.6T transceiver volume + customer prepayments + LWLG EO-polymer modulator integration. The announced December 2026 capacity > 5× Q4 2025 actual with 70%+ reserved through 2028 is a unique forward-revenue-visibility data point.
  2. FY 2028 financial-model target — Tower has publicly committed to $2.84B revenue, 39-40% GM, 31.7% Op margin (vs FY 2025: $1.57B / 23.2% / 12.4%). That is a management commitment, not analyst projection — and Tower’s CEO Russell Ellwanger has been in role since 2005, with multi-cycle execution credibility.
  3. Customer prepayments funding the capex — the $920M SiPh/SiGe envelope is being partially pre-funded by customer capacity-reservation deposits, which materially de-risks the capex-vs-FCF math (see balance sheet).
  4. NASDAQ-listed liquidity premium vs Asian-listed peers (VIS, DBHiTek, SMIC). Worth approximately 2-3 EV/Revenue turns of “geography premium” across other US-listed semis.
  5. No M&A overhang — unlike GFS (Mubadala 77%) or SMIC (state-controlled), Tower is fully distributed-float post-Intel-deal-collapse. No selldown technical pressure.

The combined premium of +9-11 turns is ambitious but not unreasonable — IF the FY 2028 target materializes. Disaggregating: ~3 turns for AI-photonics positioning, ~3 turns for FY 2028 model + customer-prepayment de-risking, ~2 turns for NASDAQ-listed liquidity, ~1-3 turns for “no M&A overhang / sole-sourced design-win quality.” If any of these fails, the premium compresses by 2-4 turns, implying $130-160 per share rather than $191.

5. Why does Tower NOT trade closer to GFS?

The “GFS discount” of approximately −8.5 EV/Revenue turns vs Tower is justifiable on:

  1. Scale. GFS is 4.3× Tower’s revenue ($6.79B vs $1.57B). Larger-scale foundries trade at lower revenue multiples on average (mean-reversion of growth rate).
  2. Mubadala overhang. GFS faces recurring secondary-offering pressure from Mubadala’s 77% stake — see GFS bear case Pillar 1 ✓. Tower has no equivalent overhang.
  3. Mature-node mix. GFS has heavy 22FDX / 12LP+ mature-node CMOS exposure, which compresses growth profile.
  4. CHIPS Act subsidies. GFS’s CHIPS Act award ($1.5B + NY $550M) is a structural subsidy advantage that GFS itself doesn’t fully reflect in multiples — implying GFS is actually undervalued relative to Tower on a subsidy-adjusted basis. This is a relative-value mispricing observation.

Cross-thesis read. If Tower trades at 13× EV/Rev and GFS trades at 4.5× EV/Rev, GFS is materially the cheaper specialty-AI-photonics merchant foundry on a per-dollar-of-revenue basis. The bear-case interpretation of Tower’s premium is that the market is over-paying for SiPh narrative purity at 200mm-PH18 scale relative to the established 300mm-Fotonix line at GFS. The bull-case interpretation is that Tower’s smaller revenue base means SiPh ramp is more concentrated and more accretive — i.e., the same $1B SiPh revenue is 64% of Tower’s FY 2025 base vs 15% of GFS’s — so Tower has higher beta to the SiPh thesis.

6. Specialty-foundry transaction comp anchor — Tower-Intel terminated deal

In August 2023, Intel terminated its planned $5.4B all-cash acquisition of Tower at $53/share (Intel press release ✓; Tower press release ✓) after failing to obtain Chinese antitrust regulatory approval. Tower at that time had:

  • FY 2022 revenue: ~$1.68B (peak cycle)
  • FY 2022 EBITDA: ~$420M (estimated from operating profit + D&A)
  • Implied transaction EV: ~$5.4B − Tower’s net cash ~$0.9B = ~$4.5B
  • Implied EV/Revenue: ~2.7×; EV/EBITDA: ~10.7× — meaningful specialty foundry M&A anchor multiple

Tower at $53/sh is the August 2023 floor anchor. Intel paid the $353M termination fee and walked away. Tower’s stock collapsed to ~$24/sh in the weeks after termination — establishing a double-bottom range of $24-53/sh through 2023-2024.

Tower broke above $53 in mid-2024 as the silicon-photonics narrative began compounding. The current $191.54 represents ~3.6× the Intel deal price — meaning the market now values Tower at ~3.6× what Intel was willing to pay just 2.5 years ago, despite revenue being only modestly higher ($1.57B FY 2025 vs $1.68B FY 2022).

Applied to current Tower: At the Intel-deal multiple of 2.7× EV/Revenue, current Tower at $1.57B revenue → EV $4.2B + $1B net cash = $5.2B mkt cap = $46.5/sh ⚠. The current $191.54 price is therefore ~4× above the Intel-deal-anchor implied price.

Read. The Intel deal anchor is highly stale as a primary valuation reference for the AI-photonics-narrative-priced equity. It serves better as a floor anchor: in a downside scenario where the AI-photonics thesis fully de-rates (SiPh ramp fails, GFS Fotonix wins all the merchant volume, Israel-risk premium expands, FY 2028 model is reset), Tower could compress to $50-70 per share without breaking technical floor support. That implies a downside of approximately −65 to −75% from current in the worst-case bear scenario.

7. Sell-side coverage / consensus snapshot

Per aggregator consensus reads (◐ — primary-source extraction queued):

  • Mean price target: ~$190-220 ◐ — implies ~0% to +15% upside vs current ($191.54)
  • High target: ~$260-280 ◐ — bull-case SiPh + envelope-tracker monetization
  • Low target: ~$120-140 ◐ — bear-case SiPh ramp slippage + RF Mobile compression
  • Coverage: Citi, Needham, Stifel, JPMorgan, Susquehanna, Wells Fargo ⚠
  • Rating distribution: Likely 60% Buy / 30% Hold / 10% Sell ⚠

A representative price-target read: 24/7 Wall St “Could Hit $230” column (24/7 Wall St ◐) sized the bull case at $230, implying ~+20% upside to current. This is consistent with a 12-month sell-side target distribution.

Sell-side primary-source extraction (analyst notes / firm reports) is queued for analyst coverage (not yet written). Aggregator consensus is mid-cycle reasonable but should be triangulated with at least 3 firm-specific notes.

8. Open items / backfill queue

  1. Per-cell primary-source citation table — every TTM revenue / EBITDA / cash / debt figure for the comp set needs to be sourced from the most recent 10-K / 20-F / 6-K. Aggregator-only data is currently the dominant input. TODO ⚠.
  2. VIS / DBHiTek 6-K-equivalent extraction — Vanguard files Taiwan TWSE quarterly disclosures; DBHiTek files Korean KOSPI quarterlies. Direct primary-source extraction queued.
  3. Tower TTM EBITDA precise calculation — depends on D&A breakout from Q4 2025 6-K + FY 2025 cash-flow statement; current $494M is ⚠ inferred.
  4. Forward EPS consensus — FY 2026E / FY 2027E / FY 2028E sell-side consensus distributions; currently aggregator-only.
  5. Tower-Intel deal multiple sensitivity — if Intel had used a more recent FY 2024 or FY 2025 Tower revenue base, the implied transaction EV/Rev multiple would be ~4× rather than 2.7×; useful for downside-floor recalibration.
  6. NVIDIA / Broadcom / Marvell SiPh customer commitment dollar quantum — would materially affect the SiPh-revenue trajectory and the Tower premium justification.

Sources

  • Tower Semiconductor: FY 2024 20-F (acc. 0001178913-25-001537, 2025-04-30); Q4 2025 6-K release (2026-02-11); spot price [companies/tsem/data/STOCK_PRICE_DATA.json].
  • GlobalFoundries: FY 2025 20-F (acc. 0001709048-26-000022, 2026-02-27); Q4 2025 6-K release (2026-02-11). Cross-references in GFS comps_valuation ✓.
  • VIS / DBHiTek / SMIC / UMC / TSMC / X-FAB / Hua Hong: Aggregator data via Yahoo Finance, StockAnalysis. Primary-source 10-K/20-F extraction queued.
  • Tower-Intel terminated deal: Intel press release Aug 16 2023 ✓; Tower release ✓; EE Times $353M fee ✓; Manufacturing Dive ✓.
  • 24/7 Wall St $230 target column — 247wallst.com ◐. Aggregator/opinion source.
  • Tower FY 2028 financial-model target — TipRanks ✓ summarizing Q4 2025 release.

Cross-references