OpenAI and Hon Hai Technology Group, better known as Foxconn, have agreed to co design and manufacture key hardware for artificial intelligence data centers in the United States. This move, the companies say, will support a rapid expansion of domestic AI infrastructure.
U.S. made racks, cabling and power systems
Under the agreement, Foxconn will work with OpenAI to design several generations of AI data center racks and related components, then build them in its U.S. factories. Company statements say the products will include cabling, networking equipment and power systems for high density AI clusters, produced at sites in Wisconsin, Ohio and Texas.
Foxconn told investors it will design and engineer the racks and components, while OpenAI provides detailed performance and power requirements from its large language models. OpenAI will gain early access to test complete systems and will have the option to buy them, but the initial contract carries no purchase commitments or financial obligations for either side.
The hardware is aimed at the infrastructure that surrounds specialised AI chips, such as the racks, cables and power distribution that let accelerators run at full capacity. Foxconn already supplies AI servers for Nvidia and assembles Apple’s iPhone, giving it experience at scale in both data centers and consumer electronics.
Foxconn leans harder into AI server manufacturing
Foxconn has been shifting its business mix toward electric vehicles and AI computing as growth in smartphones slows. The group reported that profit in its most recent quarter rose about 17 percent year on year, helped by demand for AI servers, and it expects AI related growth to accelerate in 2026.
At a recent briefing, chairman Young Liu said Foxconn can currently produce around 1,000 AI server racks per week and plans to increase that capacity next year, backed by annual AI investments of 2 to 3 billion dollars.
The OpenAI contract gives Foxconn a closer look at how power hungry frontier models are shaping data center design. Reuters reported that Foxconn sees the tie up as a way to refine products for high performance computing customers and stay central to the AI data center boom.
Part of OpenAI’s broader Stargate infrastructure push
The deal sits inside a much larger hardware strategy at OpenAI. The company has already partnered with Broadcom to co develop custom accelerator chips, targeting roughly 10 gigawatts of dedicated AI hardware capacity.
In public comments and filings, CEO Sam Altman has outlined a long term plan to invest about 1.4 trillion dollars to reach roughly 30 gigawatts of computing power, a scale he has compared to the electricity use of tens of millions of U.S. homes.
That effort is anchored in Stargate, a U.S. data center program backed by OpenAI, SoftBank and Oracle that is targeting 10 gigawatts of capacity and 500 billion dollars of investment over four years. Recent updates from OpenAI and Reuters say five new Stargate sites in Texas, New Mexico, Ohio and a yet to be named Midwest location will bring planned capacity to nearly 7 gigawatts and more than 400 billion dollars of committed spending.
SoftBank’s side of the build includes heavy manufacturing in the U.S. industrial belt. The firm is investing up to 3 billion dollars to convert a former Foxconn electric vehicle plant in Lordstown, Ohio, into a factory for modular data center units that will serve Stargate sites. Foxconn is expected to operate that facility under separate agreements.
Supply chain and policy angles
OpenAI and Foxconn say building AI hardware in Wisconsin, Ohio and Texas will shorten supply chains and limit exposure to potential tariffs on technology made in China.
Sam Altman has framed the partnership as a chance to revive manufacturing at scale in the United States. In comments reported by the Associated Press, he called the deal “a step toward ensuring the core technologies of the AI era are built here” and argued that it can help spread the benefits of AI more widely.
