Blackwell

Blackwell is NVIDIA's GPU microarchitecture generation, succeeding the Hopper generation, designed to accelerate artificial intelligence training and inference workloads at data-center scale as well as professional workstation applications.

The architecture represents a significant generational leap in performance, with flagship chips such as the GB300 delivering substantially higher throughput than their Hopper predecessors — particularly for so-called agentic AI workloads, where AI systems autonomously plan and execute multi-step tasks. Blackwell hardware has topped AgentPerf, a benchmark specifically designed to evaluate agentic AI performance, signaling that the architecture was built with emerging AI application paradigms in mind.

Blackwell's commercial impact has been substantial. NVIDIA's quarterly revenue reached record levels driven by what analysts have described as "Blackwell dominance" in the data-center market. The architecture spans product lines from enterprise data-center accelerators to high-end workstation GPUs such as the RTX Pro 6000 Blackwell, reflecting NVIDIA's strategy of deploying the same generation across cloud, enterprise, and professional segments. Its combination of raw performance gains and broad market adoption has cemented Blackwell as a defining platform in the current AI infrastructure buildout.