Microsoft and OpenAI Unveil Custom Mega Supercomputer on Azure

Microsoft announces a massive new Azure-based supercomputer built exclusively for OpenAI to train large AI models. The dedicated system features over 285,000 cores and ranks among the top five fastest supercomputers globally.

Microsoft reveals a powerful new supercomputer built exclusively for OpenAI on its Azure cloud infrastructure. The machine features 285,000 CPU cores and 10,000 GPUs, which places it securely in the top five of the TOP500 supercomputer rankings. Although Microsoft keeps the exact performance metrics under wraps, a system of this caliber requires more than 23,000 teraflops per second of processing power.

This dedicated resource serves as the physical foundation for OpenAI's ambitious goal of training massive artificial intelligence models. OpenAI pays for the entire system, which remains completely isolated from other Azure users while utilizing Microsoft's cloud network. The partnership stems from a previous $1 billion investment and makes Azure the exclusive cloud provider for OpenAI.

OpenAI CEO Sam Altman explains that researchers designed their dream system by identifying the exact limits of existing components, and Microsoft successfully built that exact vision. The tech giant declines to share the total cost of the project or specific benchmark numbers. The supercomputer currently runs continuously as OpenAI continues its pursuit of advanced artificial general intelligence.

Read More at the original source →