Coherent to quadruple Texas wafer output for AI optics

Coherent plans to spend $650 million at its Sherman, Texas, plant to quadruple indium phosphide wafer output as Nvidia and other artificial intelligence infrastructure builders push optical interconnects into larger GPU systems.

Coherent would double the factory's footprint, add wafer fabrication equipment and expand cleanroom capacity. The company expects the project to create about 1,000 jobs, including about 550 advanced manufacturing and engineering roles.

Coherent has not given a completion date.

Regulatory action

The U.S. Department of Commerce signed a letter of intent June 16, 2026, to provide Coherent with up to $50 million in direct funding under the CHIPS and Science Act. President Joe Biden signed that law Aug. 9, 2022, and Congress gave the Commerce Department $50 billion for semiconductor manufacturing and research programs.

Commerce has not issued a final award. The agency and Coherent still need to sign definitive documents before Coherent can count the federal money as committed project capital.

Coherent also plans to use $20 million from the Texas Semiconductor Innovation Fund and the Sherman Economic Development Corp. Nvidia invested $2 billion in Coherent in March to support optics capacity, according to the supplied report.

Requirements

Legal and finance teams should treat the federal funding as conditional until Commerce closes the award. CHIPS awards can require spending controls, milestone reporting, job documentation, audit records and security commitments tied to the funded facility.

Coherent's compliance staff should separate private capital, state incentives and federal funds in project accounting. That matters because Commerce can attach payment timing to equipment purchases, construction progress and production capacity. Coherent should also document how the Sherman expansion supports U.S. semiconductor supply capacity, since Commerce framed the proposed award around domestic indium phosphide production.

The Texas and local funding create a second compliance track. Coherent should map hiring commitments, capital spending thresholds and reporting dates for each incentive source before the company starts claiming credits or reimbursements.

Technology impact

At rack scale, engineers face a distance and power problem. Copper links carry high-bandwidth electrical signals across short runs, but signal loss and power draw rise as servers span more trays and racks. Optical links carry data with light, which helps operators connect GPUs, switches and memory across longer paths while managing heat.

Engineers use indium phosphide for laser sources and signal-control devices inside optical interconnects. Those components help move data between processors and memory across AI systems. As customers build clusters with hundreds or thousands of accelerators, optical capacity becomes a procurement constraint.

Coherent runs eight U.S. wafer fabs and uses them to make semiconductors for laser light sources and optical modules. The Sherman plant gives Nvidia a U.S. source for a material stack that AI networking vendors need as they move beyond copper-heavy designs.

Nvidia has placed similar bets across the optics supply chain. The company put $2 billion into Lumentum this spring and another $2 billion into Marvell to accelerate silicon photonics work. At Computex in June, Nvidia CEO Jensen Huang said optics could help make Marvell the next trillion-dollar company.

AI operators should focus on lead time. If Coherent completes the Sherman expansion on schedule, buyers could gain more domestic capacity for InP wafers and fewer bottlenecks in optical module supply. Procurement teams should still avoid assuming relief before Coherent publishes a construction schedule, equipment ramp plan and customer allocation model.

Compliance timeline

The first date to track is June 16, 2026, when Commerce signed the letter of intent. The next checkpoint comes when Commerce and Coherent sign a definitive award that sets payment milestones and reporting duties.

Coherent should publish or disclose the completion window before customers treat the added wafer output as available supply. Until then, AI infrastructure buyers should model the Sherman project as future capacity, not current inventory.

Operations staff should also prepare supplier-risk reviews for optics. Buyers who depend on Nvidia systems may need to verify Coherent, Lumentum and Marvell exposure across lasers, transceivers and photonic switching components. A single GPU cluster can consume many optical paths, so a wafer delay can affect server delivery dates, network qualification and data center commissioning.