Roche is supercharging drug discovery with a major AI expansion, deploying more than 3,500 NVIDIA Blackwell GPUs worldwide to transform how new medicines are developed. These systems reduce the need for manual testing by simulating vast numbers of chemical interactions to identify viable compounds more efficiently.
The move marks a shift from isolated AI pilot projects to organisation-wide adoption, with AI becoming a default tool for both researchers and manufacturing teams, embedding accelerated computing at the core of Roche’s R&D approach.
New backbone for innovation
The deployment will span Roche’s public cloud infrastructure and physical data centres across the US and Europe. This dual setup enables high-speed access to virtual R&D environments for global teams, reducing delays for scientists running complex simulations across time zones. As a result, Roche aims to scale R&D productivity, next-generation diagnostics and manufacturing efficiencies simultaneously.
“We’re excited to innovate at the intersection of science and technology to accelerate drug and diagnostic solutions development,” says Wafaa Mamilli, Chief Digital and Technology Officer, Roche. “With high-quality data and smarter AI, we will be able to leverage those insights both in pharma as well as in our diagnostic divisions.”
Benefits for patients
The AI factories will use NVIDIA’s BioNeMo platform to train biological foundation models on Roche’s proprietary data. This enables scientists to prioritise high-potential molecules more quickly by predicting which are most likely to bind to disease targets. Mamilli emphasises the human impact: “When we talk about collapsing time, we’re really talking about the patients and their families who are waiting.”
A similar approach is already delivering results at Roche-owned Genentech. Its ‘Lab-in-the-Loop’ strategy – where AI models predict which molecular designs are most likely to succeed – has reduced reliance on manual trial-and-error. With this automated feedback loop, drug candidates can be refined in silico before clinical trials. Nearly 90% of Genentech’s small-molecule programmes now integrate AI, including one oncology candidate designed 25% faster than traditional methods.
Digital twins and diagnostic frontiers
Beyond discovery, Roche will use NVIDIA Omniverse to create digital twins of manufacturing facilities, including its new GLP-1 site in North Carolina. In diagnostics, accelerated computing will support digital pathology, helping detect subtle disease patterns across large imaging datasets to improve accuracy worldwide.
