NVIDIA founder and CEO Jensen Huang on Monday introduced the company's new Blackwell computing platform, designed to unleash real-time generative AI on trillion-parameter large language models. Huang unveiled the platform at NVIDIA's GTC conference, outlining how accelerated computing has reached a tipping point to drive the next wave of AI innovation.
"Accelerated computing has reached the tipping point — general purpose computing has run out of steam," Huang told attendees. "We need another way of doing computing — so that we can continue to scale so that we can continue to drive down the cost of computing, so that we can continue to consume more and more computing while being sustainable. Accelerated computing is a dramatic speedup over general-purpose computing, in every single industry."
Named after David Harold Blackwell, a pioneering mathematician, the Blackwell platform succeeds NVIDIA's Hopper architecture. It delivers 2.5x Hopper's performance for training and 5x for inference. Blackwell features the new NVIDIA GB200 Grace Blackwell Superchip, which combines two Blackwell GPUs and a Grace CPU with high-bandwidth interconnects.
To scale up Blackwell, NVIDIA introduced the NVLink Switch chip and the liquid-cooled, rack-scale NVIDIA GB200 NVL72 system, offering 720 petaflops of AI training performance and 1.4 exaflops of AI inference performance in a single rack. NVIDIA also announced its next-generation AI supercomputer, the DGX SuperPOD powered by GB200 Grace Blackwell Superchips, for processing trillion-parameter models.
Huang emphasized that Blackwell is being adopted by major cloud providers, AI companies, system vendors, and others worldwide. "The whole industry is gearing up for Blackwell," he said, calling it NVIDIA's most successful launch ever.
Beyond hardware, NVIDIA introduced NIM (NVIDIA Inference Microservices), a new way to package and deliver software connecting developers with GPUs to deploy custom AI. Huang also detailed plans to bring generative AI to industries like healthcare, manufacturing, and telecommunications.
Extending NVIDIA Omniverse's capabilities, Huang announced Omniverse Cloud APIs to integrate the platform's digital twin and simulation technologies into third-party software. Notable partners include Ansys, Dassault Systèmes, Microsoft, Siemens, and others.
"The future is generative... which is why this is a brand new industry," Huang stated. "The way we compute is fundamentally different. We created a processor for the generative AI era."
Watch NVIDIA CEO Jensen Huang’s GTC keynote:
Are you interested in buying NVIDIA Products?
Chat with a Personal Shopper:
- Viber Community Marketplace Sales Chat (8 AM to 10 PM | Mon to Sun):
- FB Chat (8 AM to 10 PM | Mon to Sun):
