OpenAI and Taiwanese tech giant Foxconn announced on Friday an agreement to design and build AI data centre hardware, marking the latest in a growing series of infrastructure deals for the US creator of ChatGPT.

The companies will collaborate on hardware to be manufactured at Foxconn facilities in the United States, with OpenAI receiving early access to evaluate the systems and an option to purchase them.

The deal comes as global demand for AI infrastructure far outstrips supply.

“Demand for critical components in AI infrastructure is already far outpacing supply, and we expect that will only continue,” OpenAI head Sam Altman said by video link at a Foxconn event in Taipei.

OpenAI has signed an estimated $1 trillion dollars in infrastructure agreements in 2025, including a $300 billion deal with Oracle and the $500bn Stargate project with Oracle and Japan’s SoftBank.

Foxconn, also known as Hon Hai, has seen profits surge as it shifts from low-margin iPhone assembly to artificial intelligence servers, which are in huge demand as firms invest hundreds of billions of dollars into the technology.

Optimism over AI has pushed valuations sharply higher, raising fears of a bubble and contributing to recent stock market volatility.

Foxconn and Intrinsic, a subsidiary of Google’s parent company Alphabet, also announced a deal on Friday to develop and deploy intelligent robotics solutions across Foxconn facilities in the United States. No headline figure was provided.

Bloomberg Intelligence analyst Steven Tseng said Foxconn has been expanding its US capacity as a major supplier of AI servers and networking gear.

A formal partnership with OpenAI is “a strong validation that Foxconn is now firmly among the leaders in AI infrastructure build-out”, he told AFP.

Tseng added that the risk of a potential invasion of democratic Taiwan by China has prompted many AI server companies to favour production in the United States despite higher costs.

“Still, Hon Hai’s shipment scale, vertical integration, and experience in server business should help it manage the margin impact effectively,” he said.