Microsoft and start-up OpenAI are developing a supercomputer that runs on the public cloud platform Azure. The goal of this supercomputer is to train massive AI models.
The supercomputer has approximately 285.000 cores, 10.000 GPUs and 400 Gbps of network connectivity per server. This should rank the supercomputer among the top five supercomputers in the world. Microsoft has not yet disclosed exact performance numbers.
The supercomputer is hosted in Azure but serves as a dedicated source of OpenAI for its work. The supercomputer environment is therefore not available to other users or applications. As a result, OpenAI also has access to all the possibilities of the public cloud infrastructure, the provided services, and the various data centers of the tech giant.
Accelerated development of AI models
With the supercomputer, OpenAI wants to focus on training massive AI models. The AI start-up not only wants to achieve multiple research results quickly but also develop new powerful AI technology that is available for everyone to use. The supercomputer should further accelerate the development cycle of this AI technology.
Last year, Microsoft announced the collaboration on the supercomputer and invested 1 billion dollars (918 million euros) in the project.