NVIDIA expands its U.S. AI supply chain pitch, citing partner factories across chips, optics and data centers

NVIDIA on Tuesday tied its AI boom to a wider U.S. manufacturing effort, arguing that the next wave of artificial intelligence depends as much on physical infrastructure as it does on chips and software. In a blog post headlined “Build in America, for America,” the company said it and a growing network of partners are investing in American manufacturing, supply chains, energy grids, and skilled workforces to “bring the supply chain home.”

The company’s central claim is that AI buildout is not only about semiconductor devices, but about a layered ecosystem of components and services. NVIDIA said the physical stack behind modern AI systems includes advanced semiconductors, packaging, power systems, cooling, cloud capacity, and other elements that must be produced, assembled, and operated in the U.S. Jensen Huang, NVIDIA’s founder and CEO, linked the plan to a “once-in-a-generation opportunity to reinvigorate American manufacturing and supply chains.”

NVIDIA said it has helped onshore the most advanced semiconductor manufacturing in recent years to build and test its Blackwell chips in Arizona. It pointed to production underway at TSMC’s Phoenix facility and described plans for additional U.S.-based AI supercomputer manufacturing plants with Foxconn in Houston and Wistron in Dallas. NVIDIA also said it plans to produce up to $500 billion of AI infrastructure in the U.S. with partners including TSMC, Foxconn, Wistron, Corning, Lumentum, Coherent and Amkor.

The post placed those chip and system moves inside a broader network of regional factory expansion announcements. In Sherman, Texas, NVIDIA highlighted Coherent, a networking supplier across NVIDIA’s AI stack, as breaking ground on an expanded facility it described as the world’s first volume production 6-inch indium phosphide fabrication, with a project expected to create 1,000 jobs. NVIDIA described Sherman as a city of roughly 54,000 people and said those new jobs are only part of the broader employment effects tied to AI buildout.

In North Carolina and Texas, NVIDIA said Corning is expanding U.S. manufacturing of advanced optical connectivity solutions needed for next-generation AI, including new facilities and more than 3,000 jobs. NVIDIA also said Lumentum is deepening its U.S.-based manufacturing and research and development collaboration with NVIDIA on state-of-the-art optics technologies in North Carolina.

On the systems side, NVIDIA said Foxconn is building a state-of-the-art factory in Houston to manufacture NVIDIA AI systems, including NVIDIA GB300 tray modules. It also said Foxconn engineers used digital twins built on NVIDIA libraries and open models to design and validate the physical structure and the AI and robotics systems that help factory workers. Separately, NVIDIA said Wistron and NVIDIA will assemble and test NVIDIA AI systems at a new advanced manufacturing facility in Fort Worth, Texas, developed first as a Wistron digital twin built on NVIDIA AI and Omniverse libraries, using “open libraries, models, blueprints and physical AI ecosystem” to accelerate production.

NVIDIA added examples beyond manufacturing to emphasize data center and energy constraints. It said U.S. companies are involved in designing, powering, cooling, simulating and operating the infrastructure, naming Caterpillar for AI and digital twin integration in construction and industrial innovation, and listing Vertiv, Schneider Electric, Eaton, Jacobs, Siemens, Trane Technologies, and GE Vernova as participating in elements of that infrastructure. NVIDIA also described its “Rubin” generation of AI infrastructure as the world’s first to achieve 100% liquid cooling, and it said NVIDIA and Emerald AI are working with energy companies on flexible data centers that can adjust power use in response to grid conditions.

In healthcare, NVIDIA cited examples of clinical use cases where AI is intended to reduce administrative burden. It said Abridge is building a foundation model for clinical conversations using NVIDIA Nemotron open models and the NVIDIA Blackwell platform, and that it is deployed across more than 300 health systems while processing more than 2.5 million clinical conversations per week. It also described Aidoc’s aiOS platform as deployed across more than 100 U.S. health systems and 1,300 U.S. hospitals, citing more than 130 million patient cases analyzed to date and more than 50 million patient scans in the U.S. alone, with a tool called First Read drafting preliminary reports for chest X-rays designed to be reviewed by radiology teams.

In science and research, NVIDIA said it is working with Oracle and the U.S. Department of Energy to build new supercomputing systems at Argonne National Laboratory. It also referenced “NVIDIA Earth-2” open models for weather and climate forecasting, including faster localized storm predictions and global forecasts. The post also referenced broader employment and productivity claims, including a statement attributed to “Public First” that NVIDIA-driven AI demand could contribute $485 billion to U.S. GDP in 2026 and support over 100,000 jobs, along with a Ramp report that it said found early evidence that high-intensity AI adopters saw faster job growth after adoption.

The post did not provide details on many of the commitments it cited, including timelines, contract values, or whether all projects are already funded and under contract. It also did not break out how much of the employment figure is tied directly to NVIDIA’s spending versus partners’ separate initiatives. NVIDIA did, however, emphasize that responsible scaling requires attention to energy availability, grid reliability, water use, local needs, workforce development, and regulatory requirements. The immediate question for investors and policy watchers is how quickly the described factory plans convert into measurable capacity, what bottlenecks appear in power, cooling, and supply inputs, and whether the U.S. buildout keeps pace with the pace of AI demand.