Microsoft today announced the availability of a new powerful supercomputer built on Azure. The cloud-based machine, which is designed to train massive distributed AI models, has been built in collaboration with OpenAI as a part of Microsoft's multiyear partnership with the San Francisco-based company.
To build the supercomputer for OpenAI, Microsoft's engineers used more than 285,000 CPU cores. The single system has 10,000 GPUs to offer 400 gigabits per second of network connectivity for each GPU server, and it ranks in the top five on the list of top 500 supercomputers in the world.
Microsoft added that today's announcement is the first step toward developing the next generation of very large AI models. The company's goal is to make its large AI models, training infrastructure, and supercomputing resources available through Azure AI services and GitHub for other developers and organizations to build upon. Microsoft's Chief Technical Officer Kevin Scott explained that its potential benefits extend far beyond "narrow advances in one type of AI model."
"The exciting thing about these models is the breadth of things they’re going to enable," Scott said. "This is about being able to do a hundred exciting things in natural language processing at once and a hundred exciting things in computer vision, and when you start to see combinations of these perceptual domains, you’re going to have new applications that are hard to even imagine right now."
The new supercomputer is ideal for companies with the most demanding AI workloads, and it will be able to help them in accelerating AI development. In addition, Microsoft promises that its services and products will offer new experiences and increased productivity for customers by taking advantage of the new AI supercomputing technologies and large ML models.