Nvidia Takes Medical Transcription Open-Source With Abridge

Nvidia and Abridge are building an artificial intelligence model that turns doctor-patient conversations into clinical notes, The Wall Street Journal reported Thursday. The model runs on Nvidia's Nemotron open models and will be deployed exclusively inside Abridge's transcription app. A launch is set for later this year, according to executives cited in the report.

The partnership shows Nvidia's open models can be used for drug discovery and medical devices, the report said. In January 2025, Nvidia announced a series of healthcare partnerships. The company said it was working with partners on AI agents that reduce administrative burdens in clinical trials and AI models for drug discovery and digital pathology. In March, Nvidia released an open platform built for healthcare robotics – a stack of datasets and simulation tools designed to train AI on surgical environments and deploy them in real clinical settings. The platform covers two categories: AI that watches surgery and surfaces information to clinicians in real time, and AI that handles coordination work between procedures.

The Abridge model is purpose-built for medical language, a shift from the general-purpose AI models that power most transcription tools today. Open-source design lets Abridge – and any health system that adopts it – fine-tune the model on proprietary patient data without sending that data to a third-party API provider. That matters because medical transcription carries high stakes. A missed symptom or a misattributed medication can change a diagnosis.

Nvidia faces competition from OpenAI and Anthropic in healthcare AI. OpenAI introduced ChatGPT for Clinicians in April, designed for documentation and medical research. Anthropic in May teamed with the Gates Foundation on a $200 million health-and-education AI project, granting free access to its Claude model. The PYMNTS Intelligence report “Healthcare Firms Going Long on GenAI Investment” found that about 90% of large healthcare firms expect a positive return on investment from generative AI in product innovation and fraud prevention. That kind of demand keeps Nvidia's data-center business humming even as other segments cool.

Nvidia shares rose 1% to $202.45 on the session after the WSJ report crossed. The stock carries an Alpha Score of 66 out of 100, a moderate rating that reflects its dominant AI chip position balanced against valuation multiples that have compressed as rivals like AMD and custom-chip startups gain ground. For investors tracking the healthcare AI theme, the Abridge deal is a concrete data point: it shows Nvidia's open-model strategy can win deployments in regulated, high-stakes verticals – not just in developer sandboxes.

The model is scheduled for release by the end of 2025. Before it can be used in clinical workflows in the United States, it will need FDA clearance, a process that can take months. Abridge already has partnerships with several large health systems, including the Mayo Clinic and Sutter Health. If those systems adopt the new model, the open-source approach will have a real production test.