Microsoft unveiled a new supercomputer it has constructed that is hosted in its Azure cloud platform and it claims is one of the top five publicly disclosed supercomputers in the world. The supercomputer also banks on a $1 billion investment the computing giant made last year in artificial intelligence (AI) research startup OpenAI.

The Microsoft deployment was announced at the computing giant’s Build 2020 conference. It is a single system with more than 285,000 CPU cores, 10,000 GPUs, and 400 Gb/s of network connectivity for each GPU server. It aims to allow users to run large AI models in remote environments.

The supercomputer was built for OpenAI, which is a nonprofit research firm focused on developing artificial general intelligence that would essentially make machines as intelligent as humans. It was founded in 2015, and set up a for-profit corporation last year to seek additional outside investors.

Microsoft last July stepped up to that investment plate, plunking down $1 billion as part of a plan to develop a hardware and software platform to work in Azure. That investment also tied OpenAI’s cloud work to Microsoft.

Microsoft CTO Kevin Scott explained that by leveraging its Azure infrastructure it was able to deploy the system in six months “from start to finish.” He added that the systems processing capabilities allows it to be used to train “big self-supervised deep neural networks that we're using as platforms for language technologies and vision and speech technologies.”

As for the system's supercomputer ranking, Scott said that Microsoft has not officially gone through the TOP500 supercomputers process “so like we can't make definitive claims but we believe that it's probably right around the fifth most powerful machine in the world in Azure.”

Microsoft this week also added a “link” component to its Azure Synapse Analytics platform that provides a connection between the analytics platform and an organization’s databases. The Azure Synapse Link uses a cloud-native implementation of hybrid transactional analytical processing (HTAP) that allows users to glean real-time insights from data stored in their databases through a centralized process. Users can manage which data sets are run through the system and exported through the Azure Synapse Analytics platform.

In a pre-event presentation, Scott Guthrie, EVP for Microsoft Cloud and Enterprise, explained that the system “allows you to take your analytics system with Azure Synapse and pointed directly at your operational database and be able to do the exact same queries without having to copy the data, and that enables real-time analytics without impacting your online systems or your operational systems at all. That's kind of a game changer, especially when you're dealing with lots of data at big scale.”

Raghu Ramakrishnan, CTO for data and a technical fellow at Microsoft, noted in a blog post that the open-source nature of the HTAP implementation provides for a lower cost structure compared with traditional HTAP models.

The system is initially available to link with Microsoft's Cosmos DB database, but it plans to add support for other database platforms later this year.