NVIDIA Unveils Blackwell GPU and AI Robotics Platform at GTC 2024
NVIDIA announces its powerful new Blackwell GPU architecture and NVIDIA Inference Microservices to handle trillion-parameter AI models. The company also reveals a full hardware and software stack designed to power the next generation of physical AI robotics.
NVIDIA CEO Jensen Huang kicks off GTC 2024 by highlighting the shift toward Generative AI and the need for accelerated computing. He explains that AI expands beyond the digital realm into the physical world, requiring systems that handle trillions of parameters rather than just billions. To meet this massive demand, NVIDIA introduces the Blackwell GPU platform, which provides the necessary power to train and run these enormous models.
To make this computing power accessible, NVIDIA unveils NVIDIA Inference Microservices, or NIMs. This library of specialized AI APIs allows users to assemble teams of AI agents to solve complex tasks collaboratively. These capabilities extend directly into the physical world through advanced robotics found in factories, warehouses, and autonomous vehicles.
NVIDIA supports this robotic evolution with a comprehensive stack of hardware and software solutions. NVIDIA DGX drives foundational model training, while OVX powers the Omniverse digital twin environment where thousands of AI models learn to operate simultaneously. Finally, NVIDIA AGX brings this intelligence to the edge, enabling physical robots to perceive, adapt, and function as capable AI companions in the real world.