Roche is significantly expanding its artificial intelligence infrastructure with the deployment of more than 3,500 NVIDIA Blackwell GPUs across facilities in the U.S. and Europe, creating what the company states is the largest announced GPU footprint for a pharmaceutical firm. This investment will power advancements ranging from biological foundation models to drug discovery and the creation of digital twins for manufacturing processes, embedding AI throughout Roche’s operations. The expanded capacity is already accelerating research; nearly 90 percent of Genentech’s eligible small-molecule programs now integrate AI, with one oncology program seeing molecule design speed increase by 25 percent. “We’re excited to innovate at the intersection of science and technology to accelerate drug and diagnostic solutions development,” said Wafaa Mamilli, chief digital and technology officer at 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.”
Roche Deploys 3,500+ NVIDIA GPUs for Accelerated Computing
With an investment exceeding 3,500 high-performance NVIDIA Blackwell GPUs, Roche is establishing a substantial AI computing infrastructure spanning both hybrid cloud and on-premises environments across the United States and Europe. This represents the largest announced GPU footprint currently available to a pharmaceutical company. The deployment isn’t merely about increasing computing power, but fundamentally reshaping how Roche approaches pharmaceutical and diagnostics development, embedding accelerated computing into every stage from initial discovery to final delivery of healthcare solutions. This shift moves artificial intelligence beyond isolated pilot projects and establishes it as an enterprise-wide capability, accelerating scientific ambitions across the organization. The company’s strategy centers on connecting experiments, data, and AI in an iterative loop, dubbed Lab-in-the-Loop, to address complex scientific challenges, with nearly 90% of Genentech’s eligible small-molecule programs already integrating AI.
One oncology program benefited from a 25% acceleration in designing a degrader molecule, while another saw the creation of a backup drug candidate in seven months, a process that previously took over two years. Beyond drug discovery, Roche is leveraging NVIDIA Omniverse libraries to construct digital twins of its manufacturing facilities, simulating and optimizing complex systems before physical implementation, as demonstrated by the new GLP-1 manufacturing facility in North Carolina. Mamilli confirmed that the company is using NVIDIA Omniverse to build high-fidelity digital twins of its manufacturing sites, such as the facility in North Carolina. The company also intends to apply AI to regulatory documentation, quality assurance, and production scheduling, aiming for efficiency gains throughout its global supply chain.
Genentech Integrates AI into 90% of Small-Molecule Drug Programs
Genentech is rapidly integrating artificial intelligence into its core research and development processes, a strategy reflecting the broader industry trend of leveraging computational power to accelerate drug discovery. While early AI applications in pharmaceuticals often remained isolated experiments, Genentech now applies AI to nearly 90% of its eligible small-molecule programs, signaling a fundamental shift in operational methodology. This expansion isn’t merely about automating existing tasks; it’s about fundamentally altering how scientists approach complex biological challenges. A key component of this transformation is Genentech’s “Lab-in-the-Loop” strategy, connecting experimental data, AI analysis, and iterative refinement to address particularly difficult research questions. The benefits are already demonstrable; in one oncology program, AI-assisted design accelerated the creation of a degrader molecule by 25 percent. This acceleration is fueled by a substantial investment in computing infrastructure, exceeding 3,500 high-performance GPUs, and the implementation of platforms like NVIDIA BioNeMo. This allows Roche to explore a significantly larger range of biological and chemical possibilities, enhancing the potential for identifying novel therapeutic targets and accelerating the delivery of innovative healthcare solutions.
We’re excited to innovate at the intersection of science and technology to accelerate drug and diagnostic solutions development.
Wafaa Mamilli, chief digital and technology officer at Roche
NVIDIA Omniverse Powers Digital Twins for Manufacturing Optimization
These virtual replicas aren’t merely visual representations; they are dynamic simulations intended to optimize complex systems before physical implementation, a strategy that could significantly reduce both time and cost. The integration of digital twins extends beyond initial facility design, encompassing optimization of production scheduling, quality assurance protocols, and even regulatory documentation. Roche anticipates that even incremental improvements in these areas will have a substantial impact on its global supply chain. The company’s investment in over 3,500 NVIDIA Blackwell GPUs, distributed across both cloud and on-premises infrastructure in the U.S. and Europe, underpins this ambition. This isn’t simply about automating existing processes, but about fundamentally reshaping how Roche approaches manufacturing, allowing for proactive identification and resolution of potential bottlenecks. “When we talk about collapsing time, we’re really talking about the patients and their families who are waiting,” Mamilli explained, emphasizing the patient-centric motivation driving this technological advancement. By simulating and refining manufacturing processes in a virtual environment, Roche aims to accelerate the delivery of critical pharmaceuticals and diagnostics to those who need them most.
We are using NVIDIA Omniverse to build high-fidelity digital twins of our manufacturing sites such as the facility in North Carolina.
