San Jose— “The only way to hold more people at GTC is to grow San Jose!” exclaimed Jensen Huang, CEO of Nvidia, as he took the stage to address over 25,000 in-person (and roughly 300,000 virtual) attendees during his keynote at San Jose’s SAP Center at the annual NVIDIA GTC conference.
Nvidia, the Silicon Valley-based microprocessing giant, delivered a packed program to capture the impact of AI across a range of industries, including life sciences, at what Huang described as the “Super Bowl of AI… where everyone wins.”
Over 700 healthcare companies from more than 40 countries were represented at GTC to discuss today’s pressing AI applications, from the promise of protein design and foundation models to digitizing healthcare records and fully autonomous labs. Among the line-up of speakers were Nobel Laureate Francis Arnold, PhD, Vijay Pande, PhD, general partner at Andreesen Horowitz, Patrick Hsu, PhD, co-founder and a core investigator of the Arc Institute, and more.
Kimberly Powell, vp of healthcare at Nvidia, highlighted that we’ve come a long way in a short period of time in applying large language models and generative AI approaches for the field of drug discovery.
“We’re taking all the right models and packaging them up so people can access them. They’re getting built into software and R&D platforms of the pharmaceutical industry. So far in 2025, we’re seeing rapid adoption of these capabilities, and we know we need to keep pushing the capabilities higher,” Powell told GEN.
Among the wealth of Nvidia announcements was the unveiling of DGX Spark powered by the NVIDIA Grace Blackwell platform, which delivers the utility of an AI supercomputer in a desktop-friendly size ideal for researcher and data scientist workloads.
NVIDIA BioNeMo, which provides science-specific AI frameworks, pre-trained models, and generative AI tools to support drug discovery, also received a round of updates.
Sapio Sciences, a software development company focusing on drug research, announced the integration of NVIDIA BioNeMo into the Sapio Lab Informatics Platform, which brings computational drug discovery directly into Sapio ELN (Electronic Lab Notebook). Within Sapio ELN, researchers can access BioNeMo NIM microservices to identify and optimize drug candidates using molecular modeling, including AlphaFold2 NIM for predicting protein structures, MoIMIM NIM to design and optimize small molecules, and DiffDock NIM, a docking model developed by MIT.
Additionally, Cadence announced the expansion of its multi-year collaboration with Nvidia, focusing on driving accelerated computing and agentic AI. Notably, the integration of NVIDIA BioNeMo NIM microservices with Orion, Cadence’s cloud-native molecular design platform, will accelerate tools for drug discovery by combining AI and cloud GPUs. NIM microservices expand Orion’s capabilities in the areas of de novo protein structure prediction, small molecule generative AI, and foundational AI models for antibody property prediction.
Foundational partnerships
Nvidia’s commercial announcements were also paired with partnership highlights that emphasized the rise of foundation models that generalize across biological tasks.
In February, Nvidia, in collaboration with the Arc Institute, announced therelease of Evo 2, the “largest publicly accessible AI model for biology to date.” The genome foundation model was trained on more than 9.3 trillion nucleotides from the genomes of more than 128,000 species across the tree of life to provide both predictive and generative capabilities, including identifying disease-associated gene variants and synthesizing new genomes computationally.
Evo 2 is available for public use on the NVIDIA BioNeMo platform and as an interactive user-friendly interface calledEvo Designer. In addition, the authors have made its training data, training, inference code, and model weightsopen source.
Nvidia has described an ecosystem approach to partnering to ensure that what’s being built is co-developed with industry experts, such as researchers at Arc, to ensure utility in the field.
“We partner with people who are pushing the envelope. Arc is clearly doing that at the very bleeding edge of what’s in biological understanding. The consequence of the collaboration is a platform for everyone,” Anthony Costa, PhD, director of digital biology at Nvidia, toldGEN.
Hsu, one of the lead researchers of Evo 2, explained what makes these technologies powerful is when there are users providing feedback on areas of research that “we don’t do at Arc.” Example Evo 2 user applications span from basic research to the patient setting, including understanding biosynthetic gene clusters for synthesizing small molecules and analyzing labeled patient data for patient stratification.
“When people get more sophisticated on these models, we learn more about the science and technology to get ready to build the next thing,” weighed in Costa.
Taken together, the theme of GTC remained collaboration and wide utility.
“As a first-time GTC attendee, I’m blown away by the horizontality of the attendees. AI touches every part of society and it’s really remarkable to see the breadth of industries at GTC,” said Hsu.