Tower Semiconductor and Semtech have signed a multi-year manufacturing agreement for AI connectivity semiconductors, with production distributed across Tower's fabrication facilities in the United States, Israel and Japan.

The deal pairs a specialty foundry with a company whose connectivity chips sit inside the data-center interconnects that carry traffic between AI accelerators. For Tower, the agreement adds a multi-year volume commitment from a customer in one of the fastest-growing segments of the chip market.

What does the agreement cover?

The two companies did not disclose capacity volumes or financial terms in the announcement. What the agreement establishes, according to the companies, is a long-term supply relationship for Semtech's AI connectivity products manufactured on Tower's specialty process platforms across three countries.

Semtech's product families in this market include silicon used in high-speed links between servers and between racks in AI data centers. Demand for such interconnect silicon has risen sharply as hyperscalers build out clusters of GPUs and other accelerators, where the network between chips often limits system performance as much as the compute silicon itself.

Tower, headquartered in Israel, operates fabs in Migdal Haemek, in San Antonio, Texas, and in Japan through its manufacturing footprint there. The specialty foundry does not compete at the leading edge; it focuses on mature and differentiated nodes — analog, power management, RF and image sensors, plus specialized platforms such as SiGe and photonics-related processes that matter for high-speed connectivity.

Why three countries matter

Running the same product across fabs in the U.S., Israel and Japan gives Semtech geographic redundancy in its supply chain. That structure reduces exposure to disruption at any single site, a lesson the industry drew from the shortages and logistics shocks of 2020 through 2022.

It also fits the procurement policies of large data-center customers, which increasingly ask suppliers to identify non-Taiwan manufacturing options. Most advanced logic remains concentrated at TSMC in Taiwan, and connectivity silicon fabricated in the U.S., Israel and Japan offers a way to diversify that risk without changing the product design.

For Tower, the agreement strengthens utilization at fabs that sit outside the leading-edge investment race. Specialty foundries live on steady, multi-year demand rather than on flagship-node transitions, and AI interconnects are one of the few mature-node segments currently growing rather than shrinking.

How the deal fits the AI supply picture

AI infrastructure buildouts have strained supply not only for GPUs but for the surrounding components: optical modules, retimers, transceivers and the analog front ends that move data at high speed. Semtech sells into exactly that layer of the stack.

By securing a multi-year foundry commitment, Semtech locks in wafer supply for products whose lead times and pricing have become less predictable as data-center demand grows. The agreement gives Tower a customer whose volume tracks AI capital spending, which most analysts expect to keep rising through the coming years.

The announcement did not state when volume production begins, which process platforms each site will run, or what share of Semtech's connectivity output Tower will carry. Those details will determine whether the deal is a routine second-source arrangement or a structural shift in Semtech's manufacturing base.

If AI cluster buildouts continue at their current pace, expect Tower and other specialty foundries with U.S., Israeli and Japanese capacity to sign more multi-year agreements of this kind, as connectivity suppliers move to secure wafer supply ahead of demand.