OpenAI Could Launch Its First AI Chip in 2026

The Financial Times compiled a source from comments made by the CEO of Broadcom — a company that is said to have partnered with OpenAI to design AI chips — suggesting that OpenAI could launch its first AI chip as early as 2026.

Broadcom's recent announcements mentioned a 'mystery' customer on an earnings call, one that has committed to placing orders worth more than $10 billion. Furthermore, the Financial Times reports that multiple sources familiar with the partnership have confirmed that the chip will ship in 2026, but have yet to provide any details on what to expect from the AI ​​chip.

However, there are also many outside sources that OpenAI plans to use the chip for internal work, rather than sell it outside the company.

So far, both OpenAI and Broadcom have declined to comment on the matter. However, The Wall Street Journal later confirmed the information with a report of its own.

Why is OpenAI making AI chips?

OpenAI CEO Sam Altman says it takes more power to run and process ChatGPT and similar models. As GPT and other models become more powerful, they will require more power to process calculations and queries.

This is part of the reason Google and other big tech companies are going all-in on nuclear power, and it's also why some people worry that people's electricity bills could rise, even if they don't use AI.

With newer chips designed specifically to handle AI more efficiently, these companies can hope to ease some of the burden, especially as they work to increase the number of models they offer to customers and train future models. This is especially important as companies like OpenAI continue to work on AI-powered browsers, with some like Perplexity's Comet browser.

Nvidia has had a pretty solid hold on the AI ​​hardware industry so far, but with this deal, Broadcom could make a big move, especially considering it has also partnered with Google on its custom AI chips.

David Pac

Update 08 September 2025