TLDR
- Nvidia has forged a collaboration with medical AI company Abridge to create a specialized AI model for clinical interactions between physicians and patients.
- The technology will leverage Nvidia’s Nemotron open model infrastructure and operate solely on Abridge’s platform.
- Applications include automated medical documentation and assistance with clinical decision-making processes.
- Abridge, currently valued at $5.3 billion, specializes in AI-powered transcription and summarization of medical consultations.
- Deployment is scheduled for later in 2025; Nvidia already holds an investment stake in the healthcare startup.
Nvidia (NVDA) has announced a strategic collaboration with medical technology startup Abridge to create a specialized artificial intelligence system designed for clinical interactions. The Wall Street Journal broke the news on June 11.
Shares of NVDA were changing hands near $137 when the announcement surfaced.
The upcoming AI system will be developed using Nvidia’s Nemotron collection of open-source models. The technology will function exclusively within Abridge’s ecosystem to enhance medical record-keeping and support clinical judgment.
Abridge operates from Pittsburgh and focuses on capturing and converting physician-patient dialogues into text. The company also creates automated medical documentation, patient encounter summaries, and verifies medical billing codes.
Why Open Models?
Davis Liang, Abridge’s director of applied science, explained that economic considerations played a significant role in selecting Nvidia’s open-source architecture. Compact, specialized open models offer better cost efficiency than proprietary alternatives and enable deployment on Abridge’s infrastructure.
Abridge plans to refine the Nemotron models using its anonymized clinical datasets. CEO Shiv Rao emphasized that off-the-shelf models lack the sophistication required for medical applications—clinical reasoning must undergo rigorous training and validation against practical scenarios.
“Generic models are powerful, but clinical intelligence—it still has to be trained, it has to be shaped, and it has to be evaluated against real-world conditions,” Rao said.
The specialized model should reach deployment phase before year-end. It will function as one component within Abridge’s multi-model architecture.
Nvidia’s Push Into Healthcare
Kimberly Powell, who leads Nvidia’s healthcare division as vice president, described the Abridge collaboration as representative of how the company’s open models can serve various healthcare and life sciences sectors—spanning pharmaceutical research, medical equipment development, and telehealth solutions.
Nvidia previously invested in Abridge, which secured $300 million in capital last year, achieving a $5.3 billion market valuation. Since its 2018 inception, the startup has expanded its ambient listening technology throughout prominent healthcare networks.
Dr. Joon Lee, who heads Emory Healthcare in Georgia, reported implementing Abridge’s technology across more than 3,000 medical professionals throughout his organization. He anticipates the Nvidia-enhanced model will significantly advance the platform’s capabilities.
The announcement follows similar industry developments. Microsoft recently revealed its collaboration with Mayo Clinic to develop a healthcare AI system utilizing Mayo’s clinical information. OpenAI and Anthropic have also launched healthcare-focused products.
Abridge hosted a presentation in New York City on Thursday to unveil additional platform improvements.
