Friday, March 20, 2026

Roche expands AI infrastructure with NVIDIA Blackwell GPUs

Roche has announced an expansion of its global AI compute infrastructure by launching a large-scale AI factory built on the latest NVIDIA accelerated computing and AI stack. The company is deploying 2,176 NVIDIA Blackwell GPUs on premises across the United States and Europe. Including additional cloud capacity, Roche says its combined on-premise and cloud footprint now exceeds 3,500 GPUs, which it describes as the largest announced hybrid-cloud GPU capacity in the pharmaceutical industry.

Collaboration with NVIDIA started in 2023

The rollout represents the next step in a strategic collaboration with NVIDIA that began in 2023. Roche intends to embed AI more broadly across its value chain, from discovery and development to manufacturing and commercialisation, with the aim of accelerating research and enabling data-driven decisions at scale.

Use cases across R&D, manufacturing and diagnostics

In R&D, Roche points to the NVIDIA BioNeMo platform as an enabler for its “Lab-in-the-Loop” approach, linking biology and chemistry experiments with AI models to support hypothesis testing and faster iteration. In manufacturing, Roche plans to use digital twins of production lines powered by NVIDIA Omniverse libraries to improve process optimisation and factory design.

For diagnostics, Roche highlights accelerated computing and NVIDIA Parabricks to generate insights from very large datasets. In digital pathology, the compute platform is intended to help analyse high volumes of images to identify subtle disease patterns. In digital health, Roche is using NVIDIA NeMo Guardrails to support safer and more reliable healthcare-grade conversational AI.

AI as part of Roche’s digital transformation

Roche positions the AI factory as a cornerstone of its broader digital transformation, with AI designed to complement human expertise and to support teams globally in addressing complex disease challenges.

Company information is available at roche.com.