[News] Intel Reportedly Backs Out of 300mm Wafer Deal With Tower; Production Could Move to Japan

While Tower Semiconductor has just unveiled a silicon photonics partnership with NVIDIA, its ties with another major chipmaker appear to be unraveling. According to Israeli outlet CTech, Tower disclosed that Intel intends to withdraw from a 2023 production agreement under which the U.S. semiconductor giant would have manufactured 300mm wafers for Tower’s clients at its New Mexico facility.

The report, citing Tower, reveals that the companies have since entered mediation, with the production volumes originally slated for Intel’s New Mexico facility rerouted to its Fab7 plant in Japan, where the manufacturing processes were initially certified.

The now-fracturing arrangement between the two parties, signed in September 2023, was reached just weeks after Intel’s proposed $5.4 billion acquisition of Tower fell through, CTech explains. The manufacturing agreement was struck less than a month after the two companies scrapped their merger plans, as the deal failed to secure regulatory clearance in China, the report notes.

Under the 2023 deal, Tower had planned to invest $300 million in equipment at Intel’s Rio Rancho site, securing capacity exceeding 600,000 photo layers per month to support next-generation 300mm wafer production, CTech adds.

Still, Tower Semiconductor shows strong momentum. The company reported fourth-quarter 2025 revenue of $440 million on Wednesday, up 14% from a year earlier. For the full year of 2025, Tower posted a 9% rise in revenue to $1.6 billion, while net income increased 6% to $220 million.

Looking ahead to early 2026, the company provided first-quarter guidance of $412 million in revenue, implying a 15% year-over-year gain.

Teaming up with NVIDIA

Notably, Tower Semiconductor and NVIDIA have joined forces to advance next-generation AI infrastructure through 1.6T optical modules for data centers. As highlighted by EE News Europe, the partnership leverages Tower’s advanced silicon photonics platform, engineered to address the escalating bandwidth and performance requirements of large-scale AI deployments.

According to the report, Tower said its silicon photonics (SiPho) technology will underpin 1.6T optical modules built to support NVIDIA’s networking standards. The company noted that the platform can deliver up to twice the data rate of earlier silicon photonics generations, enabling greater bandwidth density and improved throughput for optical links within AI-driven infrastructure.

Read more

• [News] Intel Reenters DRAM Race? A Closer Look at the Z-Angle Memory Collaboration with SoftBank
• [News] EMIB Challenges Test Intel amid Buzz of NVIDIA Considering 14A/18A and Advanced Packaging

(Photo credit: Intel)