- Roche has deployed more than 3,500 NVIDIA Blackwell GPUs across hybrid-cloud and on-premises environments in the US and Europe, forming what the company says is the largest announced GPU footprint in the pharmaceutical industry
- The AI factory spans drug discovery, pharmaceutical manufacturing, diagnostics, and digital health, with applications including biological foundation models, digital twins, and AI-driven pathology image analysis
Roche has launched a large-scale AI computing infrastructure powered by more than 3,500 NVIDIA Blackwell GPUs, deployed across hybrid-cloud and on-premises environments in the United States and Europe, the company announced on March 16, 2026, at NVIDIA’s GTC conference.
AI Embedded Across the Value Chain
The infrastructure spans Roche’s full operations — from R&D through manufacturing to diagnostics and digital health — marking a shift from isolated AI pilots to an enterprise-wide capability. The on-premises deployment adds 2,176 GPUs to existing cloud infrastructure, bringing the combined total to more than 3,500 high-performance Blackwell GPUs.
“In healthcare, time is the most critical variable; every day saved means a life-changing medicine or diagnostic reaches a patient sooner,” said Wafaa Mamilli, Roche’s Chief Digital and Technology Officer.
Drug Discovery: Lab-in-the-Loop at Scale
In R&D, the infrastructure supports Genentech’s Lab-in-the-Loop strategy, which connects experiments, data, and AI models in an iterative cycle. According to Roche, nearly 90% of Genentech’s eligible small-molecule programs now integrate AI. Reported outcomes include:
- A degrader molecule for oncology designed 25% faster
- A backup molecule delivered in seven months, compared to more than two years previously
The NVIDIA BioNeMo platform will be used to train and fine-tune biological and molecular foundation models using Roche’s proprietary datasets.
Manufacturing: Digital Twins in North Carolina
In pharmaceutical manufacturing, Roche is deploying NVIDIA Omniverse libraries to build digital twins of production facilities. The technology is already being applied to a new GLP-1 manufacturing site in North Carolina, allowing engineers to simulate and optimize facility design before physical construction is complete. AI development is also underway for regulatory documentation, quality assurance, and production scheduling.
Diagnostics and Digital Health Applications
Roche’s diagnostics division adds a second major application area. NVIDIA Parabricks software supports analysis across large genomic and clinical datasets, while AI models are being developed for digital pathology — scanning high volumes of images to identify disease patterns. For digital health and conversational AI applications, Roche is using NVIDIA NeMo Guardrails to maintain safety and reliability standards in clinical settings.
Infrastructure Scale and Strategic Context
The investment extends a collaboration with NVIDIA that began in 2023. “By providing the massive computational power needed to continue to scale our Lab-in-the-Loop strategy — a space we have pioneered for over five years — our scientists can build more sophisticated predictive frontier models,” said Aviv Regev, Executive Vice President and Head of Genentech Research and Early Development.
Roche says the infrastructure is designed to support the next phase of AI development, including larger biological foundation models, agentic research workflows, and AI-driven lab automation.
