NVIDIA expands healthcare AI push with protein design and radiotherapy
NVIDIA said on Oct. 21, 2026, that it is expanding collaborations in healthcare and life sciences around protein design, autonomous laboratory workflows and adaptive radiotherapy, a move that underscores how the company is using its AI and accelerated-computing stack beyond core chip sales. The announcements, tied to GTC Berlin, included a new open protein-design model with the University of Washington's Institute for Protein Design and further work with Siemens Healthineers on cancer treatment planning, according to NVIDIA's newsroom release.
Evidence and context
NVIDIA and the UW Medicine Institute for Protein Design introduced RFD4-Proteina, described as a new NVIDIA BioNeMo open model that combines programmable protein design with search capabilities. In NVIDIA's account, the model generates targeted protein designs and lets scientists and agents search and refine candidates using molecular inputs, constraints and scoring objectives. The company said the model was tested across four benchmark suites — protein binding, small-molecule binding, enzyme scaffolding and DNA binding — and achieved higher computational design success rates and generation throughput than the baselines evaluated in the research draft.
David Baker, 2024 Nobel Laureate in chemistry and director of the Institute for Protein Design at the University of Washington, said the field is moving from discovering proteins nature has made to designing new ones, and that making the model open should let researchers build on the capability. NVIDIA said RFD4-Proteina is available through partners including Amazon Bio Discovery, Anthropic's Claude Science, Benchling, Databricks, Google Cloud, Hugging Face, Nebius, OpenAI and Tamarind Bio.
The company also said Siemens Healthineers is working with NVIDIA on the next generation of adaptive radiotherapy. NVIDIA said the collaboration is exploring agentic co-planning, patient-specific simulation and real-time computing, with Nemotron open foundation models under consideration for treatment-planning support and CUDA-accelerated dose calculation and Holoscan for simulation and sensor-data integration. Arthur Kaindl, head of Varian, Siemens Healthineers' cancer-care business, said adaptive radiotherapy should adapt treatment to the patient and that the collaboration is intended to give clinicians greater speed and insight.
In the life sciences workflow, NVIDIA said companies including Automata, Benchling, Biolevate, Boltz, Danaher, Hologen and Therna Biosciences are using BioNeMo and related technologies to build scientific agents and tools. The examples ranged from agentic workflow planning and experimental data grounding to autonomous-lab integration, molecular-structure prediction, protein-design coordination and links between imaging data and clinical insights. NVIDIA framed these efforts as attempts to shorten the cycle between computational design and laboratory testing.
The company also highlighted broader hospital-operations work with Avelios, Corti, Dedalus, Dawn Health, Sword, PolyAI, Apian and VitVio, including AI scribes, multilingual patient dialogue systems and digital twins for hospital operations. NVIDIA said these efforts are aimed at reducing administrative demands and expanding access to specialized care, but the release did not provide independently verified clinical outcome data for the collaborations.
Separately, NVIDIA cited figures for the scale of the opportunity, saying global scientific R&D reached $3.8 trillion and annual pharmaceutical budgets are approaching $300 billion. It also said that by 2050, about 35 million people worldwide are projected to be diagnosed with cancer each year. Those figures were presented in the company's announcement as context for the need for faster discovery and more efficient care delivery.
At 12:28 UTC on Oct. 21, 2026, NVIDIA's shares were priced at $216.47, down 0.35% on the day, with a market capitalization of $5.25697 trillion, according to the supplied market data. The stock move was not attributed in the source material to the healthcare announcements.