During its annual GTC AI developer conference, Nvidia announced new AI services and infrastructure aimed at providing more computing power for enterprises and cloud providers. The company unveiled its latest generation of Blackwell GPUs, along with Nvidia AI Enterprise 5.0, microservices, and industry-specific applications for its Nvidia Omniverse platform.
The new Blackwell B200 GPU architecture is designed for AI computing, utilizing six technologies. It features two GPU dies connected by a 10 TB-per-second chip-to-chip link. Nvidia CEO Jensen Huang expressed the need for larger GPUs during his keynote.
The Blackwell architecture, according to Gartner analyst Chirag Dekate, offers a compelling option for enterprises looking to build AI systems, enabling them to deliver more compute in a smaller footprint. However, widespread adoption may not happen until later this year, with cloud hyperscalers like Google, Microsoft, AWS, and Oracle being among the first to access these chips.
The introduction of microservices also allows enterprises to create and deploy custom applications on Nvidia platforms. Microservices are built on the Nvidia CUDA platform, providing a computing and programming model that works across all of Nvidia’s GPUs. One notable microservice is Nvidia NIM, which enables optimized inference on more than two dozen AI models from various providers.
Nvidia’s focus on both hardware and software, including microservices, gives it an advantage over its competitors. Cloud providers such as Google, Microsoft, and AWS already offer similar services, but Nvidia’s offerings extend beyond just hardware, making it a prominent player in the AI industry.