Tower Semiconductor revamps strategy after Intel deal collapses

Intel's decision to walk away from its September 2023 agreement for Fab 11X in New Mexico has prompted Tower Semiconductor to revamp its manufacturing roadmap, scrap the facility from its 2028 model, and redirect customers — marking a strategic reset for the specialty foundry partner.

Tower Semiconductor announced on February 11 a shift in its manufacturing strategy following the breakdown of its agreement with Intel, stating that Intel intends not to proceed under the September 2023 agreement for Fab 11X in New Mexico.

While the two companies are currently engaged in a mediation process, Tower has already begun redirecting customers who were slated for the New Mexico facility to its Fab 7 facility in Uozu, Japan, where the manufacturing flows were originally qualified. As a result, Intel's Fab 11X has been excluded from Tower's newly released 2028 financial model.

Betting big on silicon photonics

To meet surging market demand, Tower is significantly increasing its capital expenditure. The company announced an additional US$270 million investment in equipment, bringing its total planned capital expenditure for capacity and next-generation capabilities to US$920 million. This investment is primarily targeted at expanding silicon photonics and silicon germanium production across its global manufacturing sites, including facilities in Israel, California, Texas, and Japan.

Tower aims to have this new capacity fully qualified by the end of 2026. This expansion is designed to achieve silicon photonics capacity exceeding five times the wafer shipments recorded in the fourth quarter of 2025. Management noted that over 70% of this total SiPho capacity is already reserved or in the process of being reserved through 2028, with many agreements backed by customer prepayments.

Record results fuel bold 2028 targets

Tower's pivot comes as the company reports record-breaking financial performance. Revenue for the fourth quarter of 2025 reached US$440 million, marking an 11% increase quarter-over-quarter and 14% growth year-over-year. For the full year of 2025, Tower reported total revenue of US$1.57 billion, a 9% increase from 2024. Net profit for the fourth quarter of 2025 was US$80 million, while full-year net profit reached US$220 million.

Buoyed by this "value-driven growth strategy," Tower has released an ambitious new financial model for 2028. The company is targeting annual revenue of US$2.84 billion by 2028, an 81% increase from 2025 levels. This model also targets a net profit of US$750 million, more than tripling its 2025 performance. CEO Russell Ellwanger emphasized that these targets are driven by the "exceptionally high adoption rate" of SiPho technology for 1.6T transceivers, where Tower currently serves as a primary provider.

Navigating memory shortages and demand risks

During the earnings call, management addressed potential headwinds, specifically regarding memory shortages and pricing. Tower executives acknowledged concerns about market factors beyond their control, such as high-bandwidth memory (HBM) manufacturers prioritizing data center demand over other sectors. Such a shift could dampen demand from mobile integrators unable to secure the memory components they need for handsets.

However, Tower highlighted its operational flexibility as a mitigating factor. CEO Russell Ellwanger explained that the company's manufacturing capacity is largely "fungible" — meaning if memory shortages create a gap in demand for RF mobile products, Tower can backfill that capacity with power management or imaging products. While these alternative products might carry different margin profiles, this strategy allows the company to continue absorbing fixed costs and maintain high utilization levels across its factories.

Tower maintains an optimistic outlook for 2026, guiding for first-quarter revenue of approximately US$412 million. The company expects sequential increases in both revenue and profitability throughout the year.

Source: Tower Semiconductor, February 2026

Article edited by Jerry Chen