Coherent Breaks Ground on Expanded Texas Facility, Scaling AI's Optical Backbone

The physical bottleneck of artificial intelligence is moving from logic to light. As AI systems scale, the ability to move data between chips and servers depends entirely on the optical components that serve as the infrastructure's backbone. Coherent has broken ground on an expanded manufacturing building in Sherman, Texas, specifically to scale the production of indium phosphide wafers used for this purpose.

This expansion is a concrete step in domestic semiconductor manufacturing. Coherent operates what it calls the world’s first 6-inch indium phosphide fab, producing the lasers and compound semiconductors that wire AI systems together. The project is supported by a $50 million CHIPS Act grant, building on roughly $17 million in earlier support from the Texas CHIPS program and the Sherman Economic Development Corporation.

The significance of this move lies in the shift toward massive-scale GPU domains. When 576 GPUs span eight racks and operate as a single system—as seen in NVIDIA Vera Rubin Ultra NVL576, which links eight NVLink racks of 72 NVIDIA Rubin Ultra GPUs into one 576-GPU domain—copper cannot carry the signal. This creates an absolute dependency on high-speed optical interconnects.

While logic chips dominate the headlines, the domestic supply chains for compound semiconductors like indium phosphide have been thin for years. The expansion in Sherman, supported by public programs like the CHIPS Act, which was funded at roughly $50 billion, aims to close that gap. This manufacturing capacity aligns with NVIDIA’s commitment to produce up to $500 billion of AI infrastructure in the U.S. through industry partnerships.

The move from commitment to construction signals that the "reindustrialization" of the United States is no longer a theoretical policy goal but an active industrial process. As compute density increases, the companies controlling the optical layer will hold the keys to scaling intelligence.

Watch the progress of the Sherman facility to see if domestic supply chains for compound semiconductors can scale at the same rate as GPU demand.