WHY THIS MATTERS IN BRIEF
Robots building servers to train AI to create better robots – the virtuous cycle of AI-Robot self improvement is starting.
Love the Exponential Future? Join our XPotential Community, future proof yourself with courses from XPotential University, read about exponential tech and trends, connect, watch a keynote, or browse my blog.
As people worry that Artificial Intelligence (AI) is going to take everyone’s jobs other people worry that Humanoid Robots will take the jobs AI doesn’t take. And they might be mostly right. So far we’ve seen Figure robots taking over jobs at BMW’s manufacturing plants, and now Foxconn and Nvidia are reportedly discussing plans to deploy humanoid robots at a new AI server production facility in Houston – ironically to assemble the servers that will run new AI training models that will make AI better and make both AI and robots better at replacing even more jobs.
If implemented, this would be the first time humanoids are used in the assembly of Nvidia products and Foxconn’s first AI server line to incorporate them. Sources familiar with the matter, who declined to be named, told Reuters the deployment could be finalized soon, with operations possibly beginning in the first quarter of 2026.
Houston was considered a strategic choice due to its new facility design and greater available space compared to existing AI server manufacturing sites. However, neither Nvidia nor Foxconn has officially confirmed the plan, but the reports align with both companies’ growing interest in automation and robotics, particularly in areas that demand speed and flexibility, like AI server production.
Foxconn has been developing its own humanoid robots through its subsidiary Foxconn Industrial Internet and is also training robots to do basic tasks like cable insertion, object placement, and light assembly work.
At a recent event in Taipei, Leo Guo, general manager of the company’s robotics division, revealed that two types of robots will be showcased in November 2025. One version has legs, while the other is mounted on a wheeled autonomous mobile base, which Guo said “would cost less than the version with legs,” though he declined to share specific figures.
Nvidia, for its part, has a strong stake in the humanoid robotics field, as it supplies the AI platforms used by many developers to create humanoid machines. At its recent Nvidia GTC 2025 event, CEO Jensen Huang stated the widespread use of such robots in manufacturing was “less than five years away.”
Nvidia’s broader plans in Texas include a second partnership in Dallas with Wistron, and both sites are expected to begin production within 12 to 15 months. Still, one unanswered question lingers: If Nvidia is pushing for humanoid robots in the US, where exactly will these robots be built?
Foxconn may turn to Taiwan, where its robotics unit is based, or to China, where it operates large-scale factories and has previously collaborated with UBTech. Vietnam and India are also plausible choices, given Foxconn’s ongoing expansion in those countries to reduce production costs.
Alternatively, the company could outsource components to established robotics manufacturers in Japan or South Korea, where the technology is more advanced and well-established.
While the United States could be considered for localized production, this is likely part of a longer-term strategy. This key detail will determine how scalable the approach is and whether the future of automated manufacturing will rely on global supply chains or more localized development.
