OpenAI and Microsoft have teamed up with Oracle to increase computational capacity for ChatGPT. As part of a cooperation announced this week, the three firms will collaborate to enable OpenAI to leverage the Microsoft Azure Al platform on Oracle’s infrastructure.
OpenAI CEO Sam Altman has not shied away from admitting that his organization need significantly more infrastructure to run its services. He has even been in talks about raising billions of dollars for an AI chip firm. In the Oracle purchase news release this week, he stated that Oracle’s chips will “enable OpenAI to continue to scale.”
To date, OpenAI has entirely relied on Microsoft for its compute requirements. In exchange, Microsoft has paid $13 billion for a 49 percent ownership in OpenAI’s for-profit company and the sole right to commercially license its technology. However, as the Oracle purchase demonstrates, OpenAI need more computation than Microsoft alone can provide if it is to meet demand and avoid further ChatGPT disruptions.
Microsoft and OpenAI are clearly concerned about how the Oracle purchase is perceived. On Wednesday, OpenAI published a follow-up statement, stating that “our strategic cloud relationship with Microsoft remains unchanged” and that the new alliance “enables OpenAI to use the Azure AI platform on OCI infrastructure for inference and other needs.”
OpenAI also stated that the pre-training of its frontier models “continues to happen on supercomputers built in partnership with Microsoft.”
Another problematic aspect of OpenAI and Oracle’s collaboration is that Oracle also offers infrastructure to xAI, Elon Musk’s OpenAI adversary.