Roche builds massive AI factory with 3,500 NVIDIA GPUs to speed up drug development

Roche just made a big bet on artificial intelligence. The Swiss pharmaceutical company announced it's expanding its AI infrastructure with 2,176 new high-performance GPUs spread across the United States and Europe. Combined with its existing setup, Roche now operates more than 3,500 NVIDIA Blackwell GPUs - the largest announced GPU footprint in the pharmaceutical industry.

The company calls this its "AI factory" - a supercomputing platform designed to speed up everything from drug discovery to manufacturing. Roche wants to use this computing power to get life-saving medicines and diagnostics to patients faster by embedding AI across its entire operation.

How does it work?

Roche's AI factory uses NVIDIA's latest accelerated computing technology to power different parts of the drug development process. The system works across several key areas:

• Research and development: The NVIDIA BioNeMo platform connects with Roche's "Lab-in-the-Loop" system, where biological and chemistry experiments work alongside AI models to test hypotheses at scale
• Manufacturing: Digital twins powered by NVIDIA Omniverse create virtual copies of production lines, letting engineers optimize processes and factory designs
• Diagnostics: NVIDIA Parabricks software helps analyze massive datasets to find insights
• Digital pathology: The system scans large numbers of medical images to spot subtle disease patterns
• Digital health: NVIDIA NeMo Guardrails ensures conversational AI meets healthcare safety standards

The infrastructure combines on-premises computing power with cloud resources, giving Roche's global workforce access to supercomputing capabilities for tackling complex medical challenges.

Why does it matter?

Drug development typically takes years and costs billions of dollars. Roche believes AI can significantly shorten these timelines. "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.

The company's approach focuses on augmenting human expertise rather than replacing it. Scientists can now build more sophisticated predictive models and test ideas that weren't previously possible. This could lead to:

• Faster identification of promising drug compounds
• More efficient clinical trial design
• Better manufacturing processes
• Improved diagnostic accuracy

Aviv Regev, who heads Genentech Research and Early Development, says this computational power will help "further shorten the path from biological insight to life-saving medicine."

The context

This expansion builds on a collaboration between Roche and NVIDIA that started in 2023. Pharmaceutical companies are increasingly turning to AI to address the high costs and long development times that plague the industry. The average new drug takes over a decade to develop and costs more than $2.6 billion.

Roche, founded in 1896 in Basel, Switzerland, has grown into the world's largest biotechnology company. The company operates in both pharmaceuticals and diagnostics, with subsidiaries including Genentech in the United States.

The AI factory represents part of Roche's broader digital transformation strategy. The company sees AI as essential for its vision of preventing, stopping, and curing diseases. With this infrastructure in place, Roche is positioning itself as a leader in AI-driven drug discovery and development.

Other pharmaceutical companies are also investing heavily in AI, but Roche's GPU footprint appears to be the largest publicly announced so far. This gives the company a significant computational advantage as the race to develop AI-powered medicines intensifies.