Roche has announced a major expansion of its global artificial intelligence (AI) infrastructure, unveiling a large-scale “AI factory” powered by next-generation accelerated computing from NVIDIA. The initiative represents one of the most significant investments in AI capabilities by a pharmaceutical company to date, aimed at accelerating drug discovery, diagnostics development, and overall digital transformation.
The newly deployed infrastructure includes 2,176 high-performance GPUs installed across facilities in the United States and Europe. With this addition, Roche’s combined on-premise and cloud-based computing capacity now exceeds 3,500 Blackwell GPUs, marking the largest publicly disclosed GPU footprint in the pharmaceutical industry. The expansion builds on a strategic collaboration between Roche and NVIDIA that began in 2023.
The AI factory is designed as a high-performance supercomputing platform embedded across Roche’s entire value chain—from early-stage research and development to manufacturing and commercial operations. Company executives say the goal is to integrate AI deeply into scientific workflows, enabling faster insights, improved efficiency, and ultimately quicker delivery of therapies and diagnostics to patients.
Wafaa Mamilli, Roche’s Chief Digital and Technology Officer, highlighted the importance of speed in healthcare innovation, noting that reducing development timelines can have a direct impact on patient outcomes. She described the AI factory as a combination of advanced computing power and scientific expertise, capable of transforming how medicines and diagnostic tools are created and delivered.
In research and development, Roche is leveraging NVIDIA’s BioNeMo platform to enhance its “Lab-in-the-Loop” approach, which integrates experimental biology and chemistry with AI-driven modeling. This allows scientists to test hypotheses at scale and accelerate discovery processes that would otherwise take significantly longer.
The infrastructure is also expected to drive improvements in manufacturing. Using digital twin technology powered by NVIDIA Omniverse, Roche can create virtual replicas of production systems to optimize processes, reduce inefficiencies, and improve scalability. In diagnostics, tools such as NVIDIA Parabricks enable rapid analysis of large datasets, while digital pathology applications can identify subtle disease patterns across vast numbers of medical images.
Additionally, Roche is incorporating AI into digital health solutions, using NVIDIA NeMo Guardrails to ensure the safety and reliability of healthcare-focused conversational AI systems.
Aviv Regev, Head of Genentech Research and Early Development, emphasized that the expanded AI capabilities will allow researchers to build more advanced predictive models and accelerate the transition from scientific discovery to clinical application.
The move underscores Roche’s commitment to leveraging cutting-edge technology to remain at the forefront of AI-driven healthcare innovation, potentially shortening the timeline for delivering life-saving treatments to patients worldwide.