• The new supercomputer developed together with OpenAI is a single system with more than 285,000 CPU cores, 10,000 GPUs, and 400 gigabits per second of network connectivity for each GPU server. 
  • Microsoft writes that compared with other machines listed on the TOP500 supercomputers in the world, their system ranks in the top five.
  • This year's Build event was full of surprises. Check out what is the news from our Microsoft Build 2020 section.
  • The Microsoft News Hub is where you will always find the most important stories about the company.
Microsoft new AI supercomputer in Azure

The partnership that Microsoft announced last year with the artificial intelligence startup OpenAI has been fruitful, to say the least, because, at this year’s Build event, they announced that they have built the fifth most powerful AI computer in the world.

What are the tech specs of the Microsoft AI supercomputer?

The supercomputer developed together with OpenAI is a single system with more than 285,000 CPU cores, 10,000 GPUs, and 400 gigabits per second of network connectivity for each GPU server. 

Microsoft writes in a blog post that compared with other machines listed on the TOP500 supercomputers in the world, their system ranks in the top five.

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. 

, said OpenAI CEO Sam Altman.

What is the Microsoft AI supercomputer going to be used for?

Microsoft announced that it would soon start to offer its Microsoft Turing models in open-source, and recipes for training them in Azure Machine Learning.

If you’re a developer, this will give you access to the same family of language models that Microsoft is using to improve language understanding across all its products.

They will also use the supercomputer with the new version of DeepSpeed.

For those that are not familiar with it, DeepSpeed is an open-source deep-learning library for PyTorch, that reduces the amount of computing power needed for large distributed model training.

The update is a lot better than the version released three months ago.  Now, it allows training models more than 15 times larger and 10 times faster than doing the same thing without DeepSpeed on the same infrastructure.