NVIDIA and Coherent have broken ground on a $2 billion optical interconnect manufacturing facility in Texas, a move that underscores the growing infrastructure demands of AI-era computing. The project, backed by federal CHIPS Act funding, targets one of the most pressing bottlenecks in large-scale AI deployments: the optical connectivity between GPUs and servers in data centers. As NVIDIA's GPU compute capabilities continue to expand—particularly with the Blackwell architecture and forthcoming generations—the ability to move data between chips at sufficient speeds has become a critical limiting factor for training and inference workloads.
The partnership addresses a real technical challenge: while NVIDIA has dominated GPU supply for AI infrastructure, the optical components that connect these processors haven't scaled proportionally. Data center operators building the compute clusters needed for foundation model training require high-speed, low-latency interconnects that current supply cannot fully support. Coherent's specialization in optical technologies makes it a natural partner, while CHIPS Act backing reflects government recognition that semiconductor ecosystem depth—beyond just chip design—is essential to maintaining U.S. competitiveness in AI infrastructure.
The Texas facility represents a strategic bet on the longevity of the AI infrastructure boom. By securing both optical supply and manufacturing capacity domestically, NVIDIA ensures that GPU deployments won't be constrained by interconnect availability as data center buildout accelerates globally. This vertical integration mirrors how the company has cultivated its CUDA ecosystem and partnerships—making sure the full software-to-hardware pipeline supports rapid scaling. For the broader AI infrastructure sector, the project signals that bottlenecks beyond raw compute are now receiving serious capital investment.