On November 21, 2025, Foxconn Technology Group and OpenAI revealed a sweeping new partnership aimed at designing and eventually manufacturing the next generation of AI data-center hardware—a move that signals a profound shift in how artificial intelligence’s physical backbone is built, and who controls it. The deal, announced at a time of mounting geopolitical tension and tightening supply chains, represents not just a business arrangement but a marker of how the AI industry is colliding with the hard realities of energy, manufacturing, and materials.
Under the agreement, OpenAI will share critical insight into the hardware requirements for both its current and future AI models. Foxconn, in turn, will focus on engineering and preparing its U.S. facilities for potential manufacturing, expanding its already significant role as a supplier of AI servers and networking equipment. The companies have outlined three main areas of collaboration: designing multiple generations of data-center racks tailored to OpenAI’s evolving model roadmap; expanding sourcing, testing, and assembly within the United States to strengthen domestic supply chains; and manufacturing key data-center components such as cabling, networking, cooling, and power systems.
Sam Altman, CEO of OpenAI, emphasized the broader significance of the partnership, stating, “The infrastructure behind advanced AI is a generational opportunity to reindustrialize America. This partnership is a step toward ensuring the core technologies of the AI era are built here. We believe this work will strengthen U.S. leadership and help ensure the benefits of AI are widely shared.” According to Capacity Media, Foxconn’s chairman Young Liu echoed this sentiment, saying, “As the world’s largest manufacturer of AI data servers, Foxconn is uniquely positioned to support OpenAI’s mission with trusted, scalable infrastructure that accelerates innovation and broadens access to transformative AI capabilities for businesses and users worldwide.”
This collaboration is set against a backdrop of unprecedented infrastructure investment by OpenAI, which is involved in an estimated US$1 trillion worth of projects this year alone. These include a $300 billion cloud deal with Oracle and the $500 billion Stargate project led with Oracle and SoftBank. Such numbers are staggering, but they also reflect the scale of the challenge: as AI models become more powerful, their demands on physical infrastructure—especially energy and hardware—have grown exponentially.
Foxconn’s expansion of its U.S. manufacturing footprint is not just about business growth. It’s a calculated response to geopolitical risk, particularly concerns over the Taiwan Strait, and to pressure from technology firms to diversify their production locations. While shifting production to the U.S. brings higher labor and operating costs, several major AI hardware buyers have signaled a strong preference for U.S.-based manufacturing to reduce vulnerabilities in the supply chain. This trend is likely to accelerate, given the growing recognition that control over the physical aspects of AI—energy, fabs, and materials—now defines competitive advantage.
But the Foxconn-OpenAI partnership is only one part of a broader transformation. Foxconn also announced a joint venture with Intrinsic, Alphabet’s robotics subsidiary, to develop AI-enabled automation tools for its U.S. facilities. The initial focus will be on assembly, inspection, machine tending, and logistics applications, with an ambitious long-term goal: full factory orchestration. Intrinsic’s web-based developer platform, Flowstate, will allow engineers from both companies to implement AI-driven solutions, including computer vision and intelligent robotics, across manufacturing lines.
Young Liu, speaking about the collaboration with Intrinsic, said, “In working with Intrinsic, we are able to tap their deep expertise in AI-driven robotics. This synergy complements our global manufacturing leadership, enabling us to collaboratively unlock the factory of the future.” Wendy Tan White, CEO of Intrinsic, added, “Together we’re working to bring the value of AI into the physical world. By marrying Intrinsic’s expertise in AI-driven robotics software—as well as Alphabet’s deep expertise in applied research and platform development—with Foxconn’s longstanding expertise with worldwide production, world-class facilities, and vision for the future of manufacturing, we will accelerate the adoption of AI where it is most needed and valuable today.”
Dr. Zhe Shi, Foxconn’s chief digital officer, highlighted the transformative goal: “This partnership will help revolutionize our factory operations, making them even more flexible, adaptable, skill-based, and scalable. From individual tasks at the robot level, to full production lines and plant management, we’re excited to be building the factory of the future together.”
Yet, as Capacity Media and other industry observers note, the promise of AI-driven transformation is now running up against the limits of physics and geopolitics. For two decades, the software industry thrived on the fantasy of infinite, weightless scalability—where marginal costs approached zero and distribution was instantaneous. But AI, especially at scale, is a different beast. Foundation models are energy-hungry, compute-intensive, and capital-heavy. Their performance is bounded not by clever code, but by the limits of electricity, manufacturing, and rare earth materials.
AI data centers now require between one to five gigawatts of power—comparable to a nuclear reactor. Building such power infrastructure is not a quick job; it takes 10 to 15 years, and the intermittency of solar and wind energy makes them insufficient for the 24/7 inference loads that AI demands. Semiconductor manufacturing is another chokepoint. Taiwan produces about 90% of the world’s advanced chips, and ASML ships only 40 to 50 EUV machines per year. Constructing a new semiconductor fab takes five to seven years and costs tens of billions of dollars. These are not constraints that can be bypassed by software iteration or clever business models.
Rare earth materials, too, are a critical vulnerability. China controls 70% of production and 90% of processing—there are no short-term substitutes or scalable alternative supply chains. As a result, the economic model of AI now resembles heavy industry more than the asset-light world of SaaS. Multibillion-dollar capital expenditures, decade-long payback periods, and irreversible infrastructure commitments are the new normal.
The competitive advantage in AI is shifting from who can iterate fastest to who can secure energy, manufacturing, and materials. As the analysis published on November 21, 2025, put it: “The companies that win the AI era are not the ones that write the best model. They are the ones that control the physical world the models depend on.” AI moats are no longer built with product velocity and ecosystem control; they’re built with secured energy, guaranteed compute, and protected supply chains.
This is why partnerships like the one between Foxconn and OpenAI matter so much. They reflect a new understanding that the future of AI is as much about atoms as it is about bits. The companies that can build, protect, and maintain the physical stack—energy, fabs, rare earths, and logistics—will shape the next decade of technological progress. For Foxconn, OpenAI, and their partners, the challenge is clear: to turn the promise of AI into reality, they must first master the world of atoms.
As the landscape shifts, it’s no longer just about writing better code. It’s about building the grid, laying the cable, running the factories, and securing the minerals. That’s where the future of AI will be decided.
