OpenAI is all set to start its mass production of its own AI chips in 2026 in collaboration with Broadcom, aiming to meet the significant surge in demand for computing power and minimise reliance on Nvidia.
Several reports suggested the chip will be co-designed by the ChatGPT manufacturer and used internally to train the high-end artificial intelligence (AI)-powered models such as GPT-5.
Broadcom CEO Hock Tan indicated a major new client with $10 billion in orders, exponentially raising Broadcom shares to 15%, pushing its market value to $1.7 trillion.
Several credible analysts confirmed that OpenAI is the unnamed customer.
This significant move is similar to strategies by Amazon, Google, and Meta, which have created custom chips to streamline the workflow by utilising the advanced AI amid surging demands of infrastructure.
While Nvidia remains the top player in the AI industry, growth in custom chip development is likely to accelerate, aiming to outdo Nvidia by 2026.
OpenAI CEO Sam Altman has stressed the need for enhanced computing capacity to train the advanced AI models.
In August, Altman revealed plans to double OpenAI's compute fleet within five months.