Tower Semiconductor Lifts 2028 Revenue Target 27% to $3.6 Billion as Silicon Photonics Demand Outstrips Capacity, Unveils Fourfold Japan Expansion
Q2 2026 earnings call, August 4, 2026 — record margins across every line item as AI datacenter optical demand accelerates
Tower Semiconductor delivered a quarter that management itself called a turning point, and then used the earnings call to substantially raise the multi-year targets it gave investors just six months ago. The Israeli specialty foundry posted record revenue, gross margin, operating margin and net margin in the second quarter of 2026, and used that momentum to lift its 2028 model to $3.6 billion in revenue, $1.63 billion in gross profit at a 45% gross margin, and $1.2 billion in net profit at a 33% net margin. That compares with a February 2026 model that called for $2.84 billion in revenue, a 39% gross margin and a 26% net margin. CFO Oren Shirazi noted that the incremental $760 million of revenue in the new model flows through at 67% incremental gross margin and 59% incremental net margin, underscoring how favorable the silicon photonics product mix has become for the company's overall profitability.
A second Japan mega-project changes the growth algebra
The most consequential new disclosure was the scope of Tower's expanding Japan footprint, structured as a two-track investment with the backing of Japan's Ministry of Economy, Trade and Industry, or METI. Track 1, announced three weeks before the call, repurposes the former Fab 6 site in Arai for 300-millimeter silicon photonics and advanced packaging capacity while maximizing output at Fab 7 in Uozu, with full production readiness targeted for the fourth quarter of 2027. This is the investment underpinning the newly raised 2028 model. Track 2 is a separate, larger commitment: a new 300-millimeter fab to be built adjacent to Fab 7 that CEO Russell Ellwanger said will "more than quadruple" Tower's Japanese 300-millimeter capacity, weighted toward silicon photonics and silicon germanium, with a target of having tools installed and functioning by the fourth quarter of 2028.
Ellwanger was careful to note that Track 2's financial contribution is not yet built into any published model, and that the company is still finalizing negotiations — not on price, but on construction timing and equipment procurement schedules. He indicated Tower intends to update its long-term model again as those details firm up, likely in early 2027. Importantly, he said the company plans to fund the buildout entirely from internally generated cash. "We're focused on covering everything with internal cash creation," he said. "There's no dilution within the plans." When pressed by Craig-Hallum's Richard Shannon for capacity specifics, Ellwanger offered a partial data point: Tower intends to add at least 20,000 to 25,000 wafers per month of incremental silicon photonics capacity in Japan, with the potential to go considerably higher depending on final facility design.
Silicon photonics is now the central growth engine
Silicon photonics revenue rose more than 60% quarter-over-quarter and over 270% year-over-year, reaching a second-quarter annualized run rate above $680 million, with management targeting a $1 billion run rate by the fourth quarter of 2026. Tower disclosed $1.3 billion of customer contracts for 2027 silicon photonics revenue specifically, with growth expected to accelerate further into 2028 — though Ellwanger clarified on the call that this figure had not been updated from prior disclosure and represents committed contracts rather than the full capacity now being built, which he said is "spoken for" by lead customers even where not yet formally booked.
Management framed the growth as a structural shift in AI infrastructure economics rather than a cyclical upswing. "The data center industry is undergoing a fundamental transformation as AI performance is no longer defined solely by compute," Ellwanger said. "It is increasingly determined by how efficiently data moves between processors." He pointed to near package optics, or NPO, as the next architectural inflection point beyond pluggable transceivers, which he said have largely already displaced copper for scale-out connections outside the rack. NPO is aimed at scale-up connectivity within and across racks, and Ellwanger told analyst Richard Shannon that it should represent "not single digits, but tens of a percentage" of silicon photonics shipments in the second half of 2027.
Tower also disclosed a multiyear epitaxial wafer supply agreement with IQE to secure indium phosphide material for integrated lasers, addressing what Ellwanger described as a prior supply crunch on that specific input; he said the company is otherwise in "very good shape" on materials availability. Separately, the company's coherent optical module program with Marvell crossed into multimillion-unit cumulative volumes, which management pointed to as proof of manufacturing scale in one of the industry's most complex photonic IC categories.
Record quarter, and a much stronger third-quarter guide
Second-quarter revenue came in at $460 million, up 11% sequentially and 24% year-over-year. Gross profit hit a record $138 million for a 30% gross margin, up from roughly 19% a year earlier. Operating profit was $90 million, more than double the prior-year period, and net profit was $91 million for a 20% net margin, up from $47 million and a lower margin base in the second quarter of 2025. Diluted earnings per share were $0.79, against $0.41 a year ago. Management guided third-quarter revenue to a $520 million midpoint, which implies an annualized run rate above $2 billion entering the second half of the year — a threshold Ellwanger described as "turning the page into multiple new exciting chapters for the company."
RF mobile is the soft spot, power and sensors are steady
Not every segment is firing. RF mobile revenue, 12% of the quarter's total, saw 300-millimeter RFSOI revenue decline 14% year-over-year as Tower consolidates that manufacturing into Fab 10 to free up Fab 7 capacity for silicon photonics and silicon germanium. Ellwanger pushed back on the idea that this business is merely stabilizing, pointing to design win momentum in premium smartphones that he expects will drive a threefold increase in 300-millimeter RFSOI wafer starts by mid-2027 relative to current shipment levels. Power management, 14% of revenue, grew on strength in high-performance computing power delivery, while sensor and display, 12% of revenue, was roughly flat year-over-year in image sensors overall but showed a sharp pickup in machine vision demand tied to semiconductor inspection tools for HBM and DDR memory assembly, along with EV battery inspection applications.
Utilization data reinforced the capacity-constrained narrative in photonics: Fab 7 in Japan is running "well above" the company's 85% utilization model, essentially full, while Fab 2, Fab 3 and Fab 9 ran between 80% and 85%, and Fab 5 in Japan sat at 75%.
Balance sheet gives management room to self-fund the buildout
Tower's balance sheet remains a source of strategic flexibility. Total assets stood at $3.8 billion as of the end of June 2026, including $1.6 billion of net fixed assets and $2 billion of current assets, with a current ratio of 4.9 times and shareholders' equity at a record $3.1 billion. The company had already received $290 million in customer prepayments in the first quarter, largely tied to 2027 capacity reservations, which it booked as short- and long-term customer advances. Of the previously announced $920 million capital investment plan spanning Israel, Newport Beach, San Antonio and Uozu, roughly half has been spent to date, with the remainder expected across the second half of 2026 and into 2027. Currency exposure in yen and shekel is largely hedged through zero-cost collar structures, which management says has limited the margin impact of recent currency swings.
Why Japan, and why now
Ellwanger offered an unusually candid rationale for concentrating expansion in Japan rather than expanding further in the U.S., where Tower also operates fabs in Newport Beach and San Antonio that it intends to keep growing. Beyond the METI government backing, he cited geopolitical neutrality as a selling point to customers, alongside a talent base inherited from Panasonic's former semiconductor R&D operations. "There's no issues for somebody to be supplied out of Japan," he said, calling the quality culture of Japanese manufacturing a genuine differentiator rather than a soft factor, drawing on his own experience living in the country to illustrate the point.
Competitive positioning and the risk investors should watch
Asked directly how Tower's expanding silicon photonics capacity stacks up against announced buildouts from GlobalFoundries, STMicroelectronics and Samsung, Ellwanger declined to handicap industry-wide supply growth, saying his focus is on retaining share with existing lead customers rather than tracking competitors. He pointed to insertion loss as the key technical figure of merit that determines competitive standing, since better insertion loss reduces the number and cost of lasers a customer needs to integrate into a module. He also flagged that Tower is working on next-generation modulator technology in both thin-film lithium niobate and indium phosphide, positioning the company for architectures one to two generations out.
On the model's sensitivity, CFO Oren Shirazi was direct about where the biggest swing factor sits: wafer selling prices, more than utilization or cost assumptions, are what drive variance in the model's bottom line. "The selling price is just 100% reflection over the margin," he said, a reminder that the newly raised 2028 targets carry real pricing risk if photonics economics soften as more capacity — from Tower and its competitors alike — comes online over the next two years.