OpenAI is partnering with Microsoft and Oracle to extend the Azure AI platform to Oracle Cloud Infrastructure (OCI), providing additional capacity for AI development and deployment.
This move aims to help OpenAI keep up with the growing demand for its generative AI services like ChatGPT, which boasts over 100 million monthly users.
OpenAI CEO Sam Altman stated, "OCI will extend Azure's platform and enable OpenAI to continue to scale."
The OCI Supercluster can scale up to 64,000 NVIDIA Blackwell GPUs or GB200 Grace Blackwell Superchips connected by an ultra-low latency cluster network and a choice of HPC storage for training large language models (LLMs), according to Oracle.
OpenAI's frontier models continue to be trained in the Microsoft cloud
Following Oracle's announcement, OpenAI clarified that its strategic cloud relationship with Microsoft remains unchanged. The partnership with OCI allows OpenAI to use the Azure AI platform on OCI infrastructure for inferences and "other needs," while all pre-training of OpenAI's frontier models will continue on supercomputers built in partnership with Microsoft.
Microsoft is OpenAI's most significant infrastructure partner and investor, having supported the AI pioneer with billions of dollars. In return, Microsoft gains access to OpenAI technologies, which it integrates into its products like Copilot.
Training the massive AI models requires specialized high-performance infrastructure provided by Microsoft's customized supercomputers, and the company likely wants to maintain control over this crucial part of the value chain.