OpenAI Secures First Engineering Version of NVIDIA DGX B200 AI Platform to Drive Innovation and Development in AI

NVIDIA Delivers First DGX B200 Engineering Units to OpenAI

On October 9, 2023, NVIDIA announced on X that it has delivered the first engineering units of the DGX B200 to OpenAI. In a tweet, OpenAI expressed gratitude to NVIDIA, saying, “Thank you for bringing the first DGX B200 engineering unit to our office—come check out our latest acquisition!"

Introducing the DGX B200 Platform

The NVIDIA DGX B200 is an integrated artificial intelligence platform designed to provide businesses with a comprehensive solution from development to deployment. Equipped with eight NVIDIA Blackwell GPUs, the DGX B200 boasts exceptional computing power, making it ideal for handling complex AI tasks such as large-scale language models and natural language processing.

Featuring NVIDIA’s B200 Tensor Core GPUs, this system can execute training tasks rapidly and supports a variety of workloads. Additionally, it offers up to 30TB of NVMe SSD storage, ensuring quick data access and streamlined processing.

Microsoft Azure: The First Cloud Service Running NVIDIA Blackwell Systems

On October 8, 2023, Microsoft announced that Azure has become the first cloud service to operate NVIDIA's Blackwell systems. According to their tweet, Azure offers AI servers powered by the GB200, optimized layer by layer to support the world’s most advanced AI models. These servers leverage Infiniband networking and innovative closed-loop liquid cooling. Photos released by Microsoft show that although the cabinet houses only eight GB200 trays, it is equipped with robust cooling technology, including upgraded liquid-to-air heat exchangers that enhance thermal efficiency.

Conclusion

NVIDIA's DGX B200 and Microsoft’s Azure cloud service are set to propel the advancement of artificial intelligence technologies further, providing businesses with enhanced computing capabilities and efficient solutions.