Nvidia designed the B200 processor for generative-artificial-intelligence data centers.
The B200 is part of Nvidia’s Blackwell platform, announced in 2024. It is intended to accelerate large language models and other demanding AI workloads, where training and inference require enormous amounts of matrix and tensor computation.
Nvidia’s data-center products combine processors, high-speed interconnects, networking, and software rather than relying on a standalone chip alone. The B200 is associated with the GB200 Grace Blackwell superchip, which combines a Blackwell GPU with a Grace CPU. Product names can be confusing because Blackwell describes the architecture and platform family, while B200 identifies a specific GPU-class processor.