Earnings Call Highlights: Tower Semiconductor (TSEM) | 2026 Q1
Executive Takeaway
SiPho monetization opens up: $1.3B in 2027 contracts is locked in, 2028 will be "substantially larger," and this is only part of customer commitments, not the ceiling. $290M in prepayments has been received, a sign that silicon photonics has truly entered the "cash-backed booking" stage.
The 300mm strategic axis shifts to Japan: taking full ownership of Fab 7 (Uozu) and planning a 4x capacity expansion is Tower's most important decision for the next three years. Using the closed Arai fab as a bridge in between is a bet on not missing any 2027 upside.
Product mix upgrade enters its second phase: gross margin jumped from 20% to 27%, with management's model targeting 39%. The momentum comes from high-value SiPho and SiGe, not price hikes (except the 13% "value reassessment" on 200mm BCD).
1. The numbers: good, but guidance is the real story
Q1 2026 revenue was $414M, up 15% YoY; net income was $65M, up 62% YoY; net margin jumped from 11% a year ago to 16%. That is the surface. Two things matter more:
Q2 guidance of $455M ±5%: a company record, up 10% QoQ and 22% YoY.
27% gross margin: up a full 7 percentage points from 20% in Q1 last year, and management stressed this curve will move linearly toward the "39% model."
CFO Oren Shirazi explained it clearly on the call: incremental revenue will fall through to the P&L at a 59% incremental gross margin. This is a structural shift driven by new-generation platforms (SiPho, SiGe, 300mm BCD) replacing older products, not a cyclical rebound.
The long-term model targets were also reaffirmed: annual revenue of $2.8B, gross profit of $1.12B, operating profit of $900M and net income of $750M.
2. SiPho: from "outlook" to "cash in advance"
Tower's SiPho story used to be stuck at "impressive growth rate, small base." This quarter the company gave two unprecedented numbers for the first time:
Committed 2027 contracts: $1.3B
Customer prepayments (received): $290M
By comparison, Tower's full-year 2025 SiPho revenue is estimated at around $230M, so the contracted 2027 portion alone is 5.6x 2025. CEO Russell Ellwanger added: the $1.3B is not total SiPho 2027 revenue, and the company will "forecast substantially higher".
"The $1.3B contract value does not represent the full demand of these customers, nor the full shipments Tower plans." — Russell Ellwanger, CEO
The weight of that statement: the company is not just taking orders, customers are actively lining up. The $290M in prepayments means customers are betting real cash on Tower's capacity.
Capacity is moving just as aggressively. This quarter, four fabs are ramping SiPho in parallel:
Fab 2 (Migdal Haemek, Israel): first SiPho shipments
Fab 3 (Newport Beach, US)
Fab 9 (San Antonio, US)
Fab 7 (Uozu, Japan, 300mm): first SiPho lots at 95% yield
Total SiPho capacity will expand 5x from the Q4 2025 base by the end of 2026. And this is only the first wave.
3. The strategic axis: 300mm in Japan is the real long-term story
If you take away only one signal, it is that Tower is betting its long-term SiPho production on 300mm Fab 7 (Uozu).
The company announced:
Full ownership of Fab 7
Adjacent land reserved, enabling a 4x expansion
Awaiting subsidy approval from METI (Japan's Ministry of Economy, Trade and Industry)
The company stressed: the fab is already profitable at current capacity
But there is an awkward timing gap: the new building takes 1.5 years, so tools can go in at the earliest in 1H 2028, while 2027 orders are already piling up at the door.
Russell's bridge solution is interesting: install specific tools in Nuvoton's closed Arai fab (part of the TPSCo system) to squeeze out capacity first, and these tools can later be moved seamlessly to the new Uozu fab. This is classic capital-efficient ramp thinking: it neither disrupts customers' ramp schedules nor buys equipment that will become useless.
"Our plan is to not miss any 2027 upside." — Russell Ellwanger, CEO
In addition, Fab 5 (Tonami) will sign a long-term supply agreement with Nuvoton to ensure manufacturing continuity for 200mm customers. This shows the company explicitly treats 200mm as a legacy business and is concentrating all resources on 300mm.
4. The modulator race: Tower bets on every horse
As optical modules move from 200G/lane to 400G/lane, the biggest variable is what material the modulator uses. This quarter Tower is advancing four approaches at once:
Approach | Partners | Status |
All-silicon Mach-Zehnder | Coherent | 400 Gb/lane demonstrated |
Heterogeneously integrated InP (EAM) | OpenLight | 400 Gb/lane on PH18DA demonstrated |
Thin-film LiNbO3 (TFLN) | Lightwave Logic, NLM Photonics | Moving toward volume production |
InP integrated lasers | OpenLight, Scintil | Ramp begins in 2026 |
The CEO's personal forecast: TFLN will last one generation, then give way to InP. The reasoning is that as channel counts rise, InP's form-factor (size reduction) advantage will overtake TFLN, which directly affects how downstream transceiver makers (Coherent, AAOI, Eoptolink, Innolight) choose.
Another under-the-radar but important development: together with Scintil Photonics, Tower unveiled the "world's first" heterogeneously integrated DWDM laser source, designed specifically for NPO and CPO. It fills the toughest light-source integration gap in CPO architectures.
5. Pluggable → NPO → CPO: slower than many expect
The market keeps hyping CPO, but Tower's timeline is sober:
Pluggables remain mainstream until 2030 (citing LightCounting data)
NPO (Near Package Optics) volume ramps starting in 2027
CPO won't ramp at scale until at least after 2030
SiPho port volume curve (LightCounting):
2025: 30M ports (90M total ports, 33% SiPho penetration)
2028: 137M ports (205M total ports, 67% SiPho penetration)
This view is key to the long-term investment logic: make money on pluggables in the short term, NPO in the mid term, and only later CPO. Tower believes it will be a leader in all three phases.
The CEO also specifically called out Arista's XPO (Extreme Dense Pluggable), the form factor Andy Bechtolsheim championed at OFC to extend the life of the pluggable generation. Tower SiPho had two strong demos at OFC, and "we are the clear leader in XPO" is an area where Russell was unusually willing to make a firm claim.
As for CPO, Tower's bet is not on "doing CPO integration" but on "making Tower PICs the reference design for CPO integrators". Russell put it plainly:
"There is no reason TSMC would not buy our PICs, if our PICs outperform everyone else's." — Russell Ellwanger, CEO
In other words, Tower does not plan to fight TSMC head-on in the CoWoS arena. Instead, it wants to build a moat around the PIC itself so that every CPO integrator depends on it.
6. SiGe, RF SOI, BCD: the other growth lines overshadowed by SiPho
SiPho is the headliner, but the other platforms posted equally strong numbers:
SiGe +24% YoY: drivers/TIAs for transceivers; active copper cables for short-reach scale-up; LNAs ramping on a Tier-1 handset platform; defense radar/satellite communications (made in the US)
RF SOI +12% YoY: a full shift from 200mm to 300mm. Full-year 2026 is expected to dip slightly, but 2027–2028 will bring record growth (handset design wins are mostly for 2028 models, with orders landing in Q3 2027)
Power Management +10% YoY: Gen3 BCD pushes Ron×Coff to 1.5 mΩ·mm²; 13% price increase on 200mm BCD
Russell's stance on pricing is worth quoting, because it reflects Tower's philosophy on customer relationships:
"We are not a company that raises prices across the board just because supply is tight. We would rather put customers on allocation than break their trust." — Russell Ellwanger, CEO
The company stressed that the 13% increase on 200mm BCD is a "reassessment of platform value," not opportunistic gouging, a framing worth noting for BCD peers. The power delivery opportunity tied to 800V DC bus rack architectures is BCD's next clear demand driver.
7. The NVIDIA thread: a 1.6T development partner
During Q&A, Cody Acree (Benchmark) asked directly about NVIDIA 1.6T collaboration. The CEO's answer was restrained, but he left this line:
"We are a development partner of NVIDIA." — Russell Ellwanger, CEO
Compared with TSMC's SiPh-on-CoWoS approach, Tower's strategy is clear: be the best maker of the PIC itself, so that TSMC and integrators are willing to attach Tower PICs to their own CoWoS.
At the end, the CEO added a "clarification," emphasizing that Tower aims to be "the reference design for major integrators", a public pitch. NVIDIA 1.6T is the first litmus test of this strategy.
8. Utilization: which fabs are full and which are ramping
Fab | Utilization | Interpretation |
Fab 2 (Israel) | ~60% | SiPho and SiGe qualification underway, which will absorb some capacity |
Fab 3 (Newport Beach) | 80% | New SiPho/SiGe processes starting up |
Fab 5 (Tonami) | 75% | Long-term supply agreement with Nuvoton |
Fab 7 (Uozu, 300mm) | >85% (fully loaded) | Awaiting expansion |
Fab 9 (San Antonio) | 80% | — |
9. CapEx and balance sheet
Of the $920M investment plan, ~40% has been paid, with the remaining 60% to be completed in 2026–2027
Total assets of $3.7B and shareholders' equity of $3B (a record high)
Current ratio of 5.6x (very healthy)
S&P Maalot raised its rating outlook from stable to positive
The strength of the balance sheet gives Tower the confidence to expand Uozu without relying heavily on external financing.
10. Three metrics to track next quarter
METI subsidy approval progress: determines when Fab 7 expansion breaks ground
Disclosure of 2028 contract size: the CEO hinted the long-term model will be updated "within 2–3 quarters," meaning the current $2.8B model will soon be replaced by new numbers
InP laser ramp speed: with starting material supply tight, who gets capacity first; Tower says it has prepared its supply chain
Conclusion
The most important thing about Tower this quarter is not the Q1 numbers or the Q2 guidance, but the combination of $1.3B in contracts + $290M in prepayments + full ownership of Fab 7. Together they mean that the silicon photonics supply chain is moving from "showcasing technology" to "locking capacity, paying deposits and fighting for allocation," and Tower is at the front of that queue.
The key over the next 12 months is not "will demand come" but "can Tower stand up its 300mm capacity." That is why the piece of paper from Japan's METI is the thing most worth watching from this call.
[Disclaimer] This article is for technology and industry trend analysis only and does not constitute investment advice.
Comments