Antonio Bordunovi/iStock Editorial via Getty Images
Microsoft (NASDAQ:MSFT)-backed artificial intelligence startup OpenAI has started using Google (NASDAQ:GOOG) (NASDAQ:GOOGL)-made AI chips to build its products, including ChatGPT, signaling its first significant move away from an exclusive reliance on Nvidia (NASDAQ:NVDA) chips, The Information reported.
The deal for GOOG’s tensor processing units (TPU) marks a supplier diversification for OpenAI, which has relied on NVDA chips in the past to train its AI models and for inference computing, a process where model execution takes place after the training is complete.
One of the biggest buyers of Nvidia’s (NASDAQ:NVDA) graphics processing units ((GPUs)), the Sam Altman-led company expects Google’s (NASDAQ:GOOG) TPUs rented through Google Cloud will help it drive down costs related to inference, according to the report.
However, the Alphabet (NASDAQ:GOOG) unit is reportedly not renting the most powerful TPUs to OpenAI, signaling its plans to use its most advanced versions to power the company’s internal projects, including its own Gemini large language model.