Coherent Expands Texas Facility to Scale Optical Components for AI Infrastructure

AI infrastructure increasingly depends on optical connections, and Coherent is expanding its manufacturing capacity in Texas to meet that demand.

The company broke ground on an expanded manufacturing building in Sherman, Texas. Coherent produces lasers, optical components, and compound semiconductors used to connect AI systems, and operates what it describes as the world's first 6-inch indium phosphide (InP) fabrication line.

NVIDIA founder and CEO Jensen Huang and Coherent CEO Jim Anderson attended the ceremony, alongside Sherman Mayor Shawn Temann and Adriana Cruz, executive director of Texas Economic Development and Tourism.

The expanded facility will increase production of InP wafers that carry data between chips, servers, and data centers using optical signals. These components form part of the optical backbone supporting modern AI infrastructure.

“AI is the ultimate general-purpose technology,” Huang said during a conversation with Anderson at the groundbreaking. “Because intelligence is fundamental, the ability to process information, to reason and solve problems, it affects every single industry.”

The project is receiving public and private support. The CHIPS Act, funded at roughly $50 billion, was created to encourage semiconductor manufacturing in the United States. Coherent announced a $50 million CHIPS Act grant to help finance the Sherman expansion, building on approximately $17 million in earlier support from the Texas CHIPS program and the Sherman Economic Development Corporation.

NVIDIA has also committed to producing up to $500 billion of AI infrastructure in the United States through industry partnerships and new sites in Arizona and Texas.

“Coherent is a world-class company, and the work you do is vital to our future, vital to the future of artificial intelligence and vital to reindustrializing the United States,” Huang said.

Optical connections for large AI systems

Compound semiconductors such as indium phosphide and gallium arsenide support the high-speed networking and optical interconnects used by modern AI systems. Their role is less visible than that of logic chips, but their supply chains have historically been concentrated outside the United States.

The need for optical connectivity grows as AI systems combine more processors. NVIDIA's Vera Rubin Ultra NVL576, for example, links eight NVLink racks, each containing 72 NVIDIA Rubin Ultra GPUs, into a single 576-GPU domain.

Copper connections become increasingly inefficient over the distances involved. Connecting processors separated by hundreds or thousands of feet across a data center requires signal conditioning and retimers, which consume power. Silicon photonics offers a more efficient way to carry signals over these distances, Huang explained.

Optical links require an initial conversion from electrical signals to light, but their energy cost is less sensitive to distance. At the scale of systems such as NVL576, optical communication is the more power-efficient option.

NVIDIA and Coherent have worked together for roughly two decades. In March, the companies announced a multiyear strategic partnership under which NVIDIA is investing $2 billion in Coherent to support research and development, future capacity, and U.S.-based manufacturing. The agreement also includes a multibillion-dollar purchase commitment for advanced laser and optical networking products.

Building optical hardware in Sherman

Sherman, a city of roughly 45,000 people about an hour north of Dallas, is becoming part of the manufacturing expansion surrounding AI infrastructure.

“When we get to full capacity, this site will support more than 550 direct jobs, and thousands of jobs, direct and indirect,” Anderson said.

The facility will produce lasers, transceivers, and pluggable optical modules. These products perform different connectivity functions within NVIDIA networking systems.

“As AI systems grow larger and more powerful, connectivity is just as important as compute,” Anderson said. “AI runs on compute, but it scales on connectivity, and Sherman is where that connective tissue gets built.”

Before the groundbreaking, visitors toured Coherent's existing fabrication facility and viewed equipment planned for the expanded building. An NVIDIA rack was displayed on the factory floor as one of six tour stops.

The tour was followed by a discussion between Huang and Anderson about their partnership and the role of domestic optical manufacturing in the expansion of AI infrastructure.

“Today marks an important milestone, not just for Coherent, but for American manufacturing and for the future of AI infrastructure,” Anderson said.

Why 6-inch indium phosphide wafers matter

The semiconductor laser was developed in U.S. laboratories. Bell Labs demonstrated a room-temperature version in 1970, but much of the technology's manufacturing later moved overseas.

“We were founded as a manufacturing company in 1971. We’ve always been a U.S. manufacturing company, and after 50 years, the most advanced 6-inch indium phosphide line in the world is right here in Sherman,” Anderson said.

Wafer size is an important factor in manufacturing efficiency. Silicon fabs commonly use 12-inch wafers, while much of the world's InP production still uses 3- and 4-inch wafers. Smaller wafers provide less usable area and produce fewer components in each manufacturing cycle.

A 6-inch wafer has approximately four times the usable area of a 3-inch wafer because wafer area increases with the square of its diameter. The larger format can therefore improve manufacturing efficiency and support higher production volumes.

Inside the facility, the basic processes resemble those used in other semiconductor fabs: lithography, photoresist application, material deposition, and etching performed layer by layer. The difference is the substrate. On an InP wafer, engineers grow compound-semiconductor layers and tune them for optical properties that allow the chip to emit and modulate light.

Coherent uses this InP technology in pluggable optical modules. These transceivers, roughly the size of a USB stick, fit into the front of NVIDIA networking switches and transmit data between racks across a data center. Each module contains an indium phosphide laser.

The modules also help enable NVIDIA Spectrum-X Photonics and Quantum-X Photonics switches with co-packaged optics. In those systems, Coherent supplies the external laser module that connects to the front plate of the switch.

As NVIDIA develops networking systems intended to prevent optical links from becoming a bottleneck, demand for these lasers is expected to increase.

“Ten years from now, I think we’ll look back and realize AI is what made it possible to invest in sustainable energy, upgrade our energy grid and reconstitute a workforce,” Huang said. “You can’t have only information workers in an economy, you also have to have builders. We have an opportunity over the next 10 years to reshape our communities and be much more balanced.”