OpenAI has entered a new collaboration with Hon Hai Technology Group (Foxconn) focused on developing the next generation of hardware required for high-performance AI data centres. The partnership aims to strengthen US manufacturing capabilities at a time when demand for physical AI infrastructure is rising sharply. While the agreement does not include purchase commitments, OpenAI will have early access to test new systems and may choose to adopt them depending on performance.
Co-Designing Multiple Hardware Generations in Parallel
Under the partnership, OpenAI and Foxconn will co-design and engineer several generations of AI data centre racks simultaneously. This approach aligns OpenAI’s long-term infrastructure roadmap with Foxconn’s manufacturing scale and engineering depth. Both companies intend to optimise rack architecture for US-based production and create a more resilient domestic supply chain for AI computing.
The collaboration also focuses on onshore sourcing, faster assembly, and improved end-to-end testing processes. These measures are expected to reduce deployment delays, increase supply chain independence, and boost the reliability of large-scale AI infrastructure.
Strengthening US-Based Manufacturing and Supply Chain Capabilities
Foxconn will manufacture key hardware components in the United States, including power systems, cooling equipment, cabling, and networking gear. The company already holds a major presence in global server manufacturing and expects the partnership to accelerate its role in supplying AI-specific infrastructure.
OpenAI CEO Sam Altman said the initiative represents an opportunity to rebuild industrial capability in the US. He noted that enabling domestic production of AI hardware is essential as the industry enters a phase of scaling compute infrastructure at unprecedented levels.
Foxconn chairman Young Liu stated that the company’s experience in manufacturing AI servers positions it well to support OpenAI’s requirements and broader industry expectations. Both sides intend for the collaboration to set a new benchmark for high-performance AI data centre builds.
Context: OpenAI’s Broader Infrastructure Expansion Plans
This partnership follows OpenAI’s earlier collaboration with Broadcom to co-develop AI accelerators and networking systems capable of scaling up to ten gigawatts of deployment. Under that initiative, OpenAI will design custom accelerators while Broadcom provides connectivity solutions and manufacturing support.
The new co-designed racks with Foxconn are expected to be deployed across OpenAI’s own environments and partner facilities beginning in the second half of 2026, with rollout continuing through 2029.
Together, these moves signal OpenAI’s long-term plan to secure compute supply, diversify its hardware partnerships, and build an infrastructure base capable of supporting increasingly complex AI model development.
