Third quarter 2026 earnings call, November 2, 2026

A higher model built on a tripling of silicon photonics capacity

Tower Semiconductor raised its long-term financial model. It now targets $3.6 billion in annual revenue, $760 million above the February 2026 framework, a 27% increase. The company expects the added revenue to carry a gross margin of 45%, up from 39%. The new model also targets $1.63 billion of gross profit, $1.38 billion of operating profit (38% margin) and $1.2 billion of net profit (33% margin). CFO Oren Shirazi said the incremental $760 million of revenue should drop through at 67% to gross profit and 59% to net profit. The model assumes all fabs run at 85% utilization once the $920 million SiPho and 5G investment is installed and qualified. It also includes the repurposing of the Arai facility, formerly Fab 6, for 12-inch silicon photonics wafers, supported by Japan's METI.

The model is notable because capacity is being sold ahead of construction. CEO Russell Ellwanger said silicon photonics start capacity will be about double the Q3 level by year-end. It will reach 3x by the third quarter of 2027 once the Track-1 300-millimeter expansion reaches full production. Q3 silicon photonics revenue was about $240 million. Management said capacity for both 200-millimeter and 300-millimeter production is covered by signed agreements or contracts in advanced negotiation, and customer commitments now extend through 2029. Ellwanger described the company as "engaged in billions of dollars of long-term agreements."

Asked to translate this into revenue, Ellwanger declined to give a formal 2028 target. He offered arithmetic instead. Tripling the $240 million quarterly base implies about $720 million per quarter, or a silicon photonics run rate "a bit above" $2.8 billion on announced capacity alone. That figure measures start-rate capacity, not guidance. It excludes integrated lasers and a further tranche of 200-millimeter expansion at two sites, which Ellwanger said will likely be announced shortly. Stifel's Sahej Singh initially mis-multiplied the figure to $960 million per year, and the CFO corrected it. The discussion nonetheless showed how large the gap is between the current silicon photonics run rate, which the company expects to exceed $1.1 billion annualized in Q4, and the capacity under construction.

Cody Acree of StoneX asked whether the $3.6 billion target might already be too conservative. Ellwanger replied, "No comment. No, I think it's an aggressive target. But certainly, we wouldn't be giving a target to the Street that we don't have very, very high confidence that we will meet." The answer stops short of endorsing higher numbers. The gap between the $3.6 billion model and an implied silicon photonics run rate near $2.8 billion, alongside roughly $1.5 billion of other businesses, will fuel debate about whether the model is deliberately cautious.

Quarterly results and a stronger start to 2027

Third quarter revenue was a record $526 million, up 14% from $460 million in Q2. Year over year, revenue rose roughly 33% from $396 million. The CFO cited 36%, which does not reconcile with the dollar figures. Gross profit was $179 million, up 92% from a year earlier. Operating profit was $127 million, 2.5x the prior-year level. Net profit was $120 million, a 23% margin, with diluted EPS of $1.05 versus $0.47 a year ago. Tower guided Q4 revenue to a midpoint of $600 million, implying full-year 2026 revenue of about $2.0 billion versus $1.57 billion in 2025.

Ellwanger said demand and start plans do not point to the usual first-quarter seasonal decline. Management expects Q1 2027 to begin above a $2.4 billion annual run rate and more than 45% above Q1 2026 revenue. The company expects the new start rates to flow through the fabs and ship in the first and second quarters of 2027, setting what it called a substantially higher sustainable baseline.

Margins: a 62% incremental that will not last

Q3 incremental gross margin reached 62%, the first time it has exceeded 60%. Shirazi cautioned against treating it as a new baseline. Only about 65% of the $920 million tool set, the portion paid for to date, has begun depreciating, and the Track-1 purchases will add further depreciation. Management maintained its 55% to 58% long-term incremental gross margin assumption. Even so, the 45% gross margin in the new model is well above the roughly 34% reported in Q3. The path there depends on utilization, pricing and the timely qualification of new tools. Ellwanger also emphasized a lean cost structure, with only a 7-point drop from gross to operating margin in the model.

Pricing, NPO and the move up the content curve

Ellwanger said the long-term agreements include pricing targets, which set "a bar of a starting position," while acknowledging Tower would be "reasonable" if the market demanded lower prices. He argued that deep discounting usually comes from new entrants who cannot compete technically. He expects near-package optics to reset the pricing curve upward because it is a more advanced platform. Pricing for NPO has not been disclosed. Management also confirmed that mix is moving toward 3.2T, 400-gigabit-per-lane designs, which carry higher dollar content.

Tower is prototyping NPO products with more than 10 customers, and Ellwanger said most NPO tape-outs are on 300-millimeter, in 16-lane and 32-lane configurations. NPO can place both transmit and receive functions in the silicon photonics die, whereas pluggables typically integrate only transmit. NPO also tends to replace DSPs with silicon germanium TIAs and drivers, expanding that opportunity. Ellwanger said all NPO customers are existing pluggable customers. He expects an "appreciable amount" of NPO shipments in the second half of 2027 but conceded the timing "really depends on market adoption." President Marco Racanelli added that Tower holds share in both formats, so slower NPO adoption would still mean pluggable wafers.

Integrated lasers and heterogeneous integration remain outside the model

Tower has begun initial production of silicon photonics die with indium phosphide lasers integrated through die-to-wafer attach. Ellwanger said this opens a laser market that multiplies Tower's serviceable market. The same process is being applied to an indium phosphide 400G modulator and thin-film lithium niobate for next-generation products. He said these forecasts are not in the $3.6 billion model and carry meaningfully higher ASPs. He was candid about uncertainty: "I truly do not have a clear answer" on the eventual attach rate, noting that analysts disagree on how much of the market will integrate lasers. Tower also sees silicon photonics optical circuit switches and coherent products for inside the data center as further expansion areas. Longer term, it is engaged with Tier 1 customers on micro-LED and VCSEL-based wide-and-slow links.

Track-2 and the 300-millimeter ramp

The Track-1 300-millimeter expansion is ahead of schedule, with all tools ordered and full production targeted for Q3 2027. Track-2 would quadruple Japan's 300-millimeter capacity from roughly 10,000 to about 40,000 wafers per month, though Ellwanger said exact numbers are not set because the building has not been constructed. He said a 300-millimeter wafer carries about 2.25x the die area of a 200-millimeter wafer, so 40,000 wafers is close to 100,000 in 200-millimeter equivalents. He said Track-2 will be well underway in 2028, with the contribution expected in 2029. Output could arrive a quarter earlier, but management is not counting on it. Tower plans to self-fund all expansion. Roughly 65% of the $920 million program has been paid, with the remainder due in Q4 2026 and the first half of 2027. Management reiterated a full return on investment within a year.

Competition and share

Analysts from Barclays and Susquehanna pressed on competition as other foundries add photonic IC capacity. Management declined to quantify share. Racanelli said the market is large enough for more than one player and that Tower was first to ramp both 200-millimeter and 300-millimeter at high yields. Ellwanger said he would not forecast share for 2027 to 2029 and likened it to "proctology." He also said the company had already achieved a majority share of the pluggable scale-out market. His main evidence was that customers are committing capacity through 2029, which he said they would not do without confidence in Tower's delivery and road map. He added that what Tower ships to lead customers today "will not be a very big portion" of its 2029 shipments. That claim is hard to verify, and the absence of share data and of named customers is a gap investors will have to accept.

Silicon germanium, RF and analog businesses

Silicon germanium demand is described as unprecedented and is served from three 200-millimeter fabs and one 300-millimeter fab. Tower announced a substantial capacity agreement with Semtech this quarter. Its next-generation H6 technology is already in production for demanding products. In RF SOI, Agrate's 300-millimeter corridor posted its highest quarterly revenue, and Fab 10 set a record for tape-outs. An analyst noted that mobile RF revenue is near a trough in the low $50 millions. Ellwanger acknowledged an earlier dip from requalifying customers while Fab 7 shifted toward optical products. He expects continued growth with a record Q2 2027. He also hinted at an unreleased RF program with a "breakthrough figure of merit" at one lead customer. Capacity here is limited and deliberately focused on the leading edge.

BCD revenue is growing about 10% year over year. Several strategic accounts have signed long-term agreements at higher prices, which Tower says implies blended BCD prices about 20% higher in 2027. In CIS, Tower started production of a high-speed global shutter sensor for a leading DSLR and mirrorless camera maker. It is also working with OLED-on-silicon display customers on an expected steep ramp in 2027. Utilization was 85% to 90% across the 200-millimeter fabs, and the 300-millimeter fab was fully utilized, above the 85% model assumption.

Balance sheet

Tower ended September with $4 billion of assets, $3.2 billion of equity (80% of assets) and a current ratio of about 3.7x. Customer prepayments, including the $290 million received in Q1 2026 and further amounts toward 2027 capacity reservation, sit as liabilities and flow through operating cash flow. That means reported operating cash flow benefits from deposits that will be earned out as wafers ship. Currency exposure is mostly hedged: yen is naturally hedged, and the shekel is covered by zero-cost collars.