According to the Financial Times, OpenAI is preparing to begin mass production of its own AI chips next year. The move is intended to secure the massive computing power needed for its services and to reduce its dependence on NVIDIA.
Read: Bandwidth Blog & Smile 90.4FM Tech Tuesday: Smart shopping trollies!
The custom chips were reportedly designed in collaboration with U.S. semiconductor manufacturer Broadcom. The CEO of Broadcom recently announced a new, unnamed client had placed a staggering $10 billion in orders, which sources confirmed was OpenAI. The company plans to use the chips internally and does not intend to sell them.
OpenAI has been exploring the possibility of making its own chips since as early as 2023, after CEO Sam Altman cited GPU shortages as a reason for its API’s speed and reliability issues. The decision to move forward with a custom chip design aligns with Altman’s recent announcement to double OpenAI’s compute fleet over the next five months to keep up with increased demand.
See also
By designing its own chips, OpenAI hopes to mitigate potential GPU shortages and lower costs. The Times reports that custom AI chips, or “XPUs,” like the one OpenAI is reportedly developing, are expected to capture a significant portion of the AI market. Currently, NVIDIA remains the dominant player in the industry, recently reporting a 56% increase in quarterly revenue.
