OpenAI inked a deal with Foxconn to co-design and manufacture AI data centre equipment in the US, marking the ChatGPT maker’s latest push to secure the infrastructure needed for its fast-growing models.

Under the agreement, Foxconn will work with OpenAI to create customised server racks optimised for AI workloads. The vendor will also manufacture and supply critical components including cabling, power systems and cooling.

The pair said that jointly engineered rack designs will help bring new systems online faster, while locally sourced architecture would improve reliability and deployment speeds to meet the “demands of AI workloads today and in the future”.

While the deal includes no purchase commitments, OpenAI will also get early access to Foxconn’s hardware and holds an option to buy. Meanwhile, the AI heavyweight will provide insight into evolving model requirements, giving Foxconn early visibility into emerging hardware needs.

OpenAI hailed the tie-up as a win for domestic technology manufacturing. CEO Sam Altman argued AI infrastructure represents “a generational opportunity to reindustrialise America”, calling the partnership “a step toward ensuring the core technologies of the AI era are built here”. He added the deal would “strengthen US leadership and help ensure the benefits of AI are widely shared”.

The move follows a flurry of multibillion-dollar compute deals OpenAI struck with players including Nvidia, AMD and Oracle as it races to scale up capacity.

Foxconn also continues to expand its AI footprint, unveiling a $1.4 billion supercomputing deal with Nvidia earlier today (21 November). According to Reuters, the 27-megawatt Taiwan cluster is set to go live in H1 2026, making it Asia’s first AI data centre built around Nvidia’s powerful GB300 AI platform.