NVIDIA launches BioNeMo Agent Toolkit for agentic life sciences

NVIDIA introduced the BioNeMo Agent Toolkit, a package of domain-specific tools and skills designed to build agentic applications for the life sciences. Rather than a single model, the toolkit targets multi-step, autonomous workflows — the agentic pattern dominating this week's announcements — applied to drug discovery, protein and molecular work, and related research pipelines.

The launch came at the ISC High Performance conference in Hamburg, where NVIDIA also unveiled software accelerating AI for chemistry, materials discovery, and even dark matter research, underscoring a strategy of building vertical, science-focused stacks on top of its hardware. The company reiterated its infrastructure dominance, noting its technology powers 81% of TOP500 supercomputers and 90% of systems newly added to the list.

The HPC framing is strategic: NVIDIA is positioning agentic AI as the next layer of scientific computing, where its GPUs already underpin the vast majority of the world's fastest machines. By shipping domain toolkits, it deepens lock-in beyond raw silicon into research workflows.

Competitively, BioNeMo extends NVIDIA's life-sciences franchise against cloud-native rivals and specialized biotech-AI players, and complements its broader agentic blueprints (such as the AI-Q Blueprint on Oracle Cloud). The concrete new facts are the BioNeMo Agent Toolkit launch, the ISC software announcements, and the 81%/90% TOP500 figures. What to watch: real adoption by pharma and research labs, and whether agentic life-sciences tools deliver validated scientific results rather than demos.