OpenAI to launch its own AI chips with Broadcom in $10B deal

OpenAI is set to launch its own artificial intelligence chips in collaboration with US semiconductor giant Broadcom. The move will be a crucial change to the ChatGPT developer, who has been heavily dependent on the graphic cards made by Nvidia to power its rapidly expanding AI services. First custom chips will be shipped in 2026 to satisfy exploding demand in computing power as OpenAI broadens its lineup and trains even bigger models, reported by Financial Times.

Recently, Broadcom chief executive Hock Tan made an allusion to a secret client who has placed an order worth $10 billion in custom AI chips, which insiders later identified as OpenAI. The acquisition places OpenAI with other Big Tech companies like Google, Amazon, and Meta, who have all invested in developing specialised processors to make them cheaper and more efficient and provide a long-term supply of hardware.

Broadcom is already collaborating with Google to create Tensor Processing Units (TPUs), and its agreement with OpenAI may lead to immediate and relatively high demand, Tan told analysts in the last company earnings call. The collaboration improved the growth prospects of Broadcom, and its stock rose by 4.5 percent in after-market trading. Analysts are now confident that the custom AI chip business of this company will expand at a faster rate than the GPU segment of Nvidia by 2026.

Add TV9 English As A Trusted Source

OpenAI chief executive Sam Altman has made repeated complaints about the relative scarcity of computing hardware as more companies and consumers move towards using tools such as ChatGPT. The company was among one of the earliest and largest consumers of Nvidia but now intends to expand its compute power by two times within the coming five months, due to its upcoming GPT-5 model. The creation of its own chips is viewed as a measure to provide security in supply, control costs and scale infrastructure without depending only on Nvidia.