Nvidia and the startup Abridge announced a partnership on June 12, 2026, to develop a healthcare-specific foundation model trained on clinical conversations. The model aims to improve accuracy, reliability, and auditability in clinical workflows, including documentation, evidence grounding, and workflow automation. Built on Nvidia's Nemotron open model family, the new system will utilize Nvidia Blackwell AI infrastructure for pre-, mid-, and post-training processes using de-identified data.

By performing domain adaptation early in the training lifecycle, the companies intend to create a model capable of clinical reasoning from its foundation. Executives noted that this approach allows for the optimization of quality, cost, and efficiency across care settings and specialties. Abridge, which currently supports over 100 million annual clinical conversations across 300 health systems, plans to use this infrastructure to transition from an AI scribe tool to a comprehensive clinical intelligence platform that integrates payer and life sciences workflows.

During a keynote event in New York City on Thursday, Abridge CEO and co-founder Shiv Rao, M.D., emphasized that control over the full AI stack is necessary to manage latency and ensure clinicians can access information immediately during patient interactions. Kimberly Powell, vice president of healthcare at Nvidia, stated that the partnership reflects a shift from generative AI toward systems that can reason and perform work. According to Powell, Nvidia views AI as a full-stack computing challenge encompassing energy, chips, data centers, and foundation models, with Abridge focusing on the vertical application layer.

This collaboration adds to a series of healthcare initiatives for Nvidia, which also maintains investments in Abridge through its NVentures arm. The company has previously formed infrastructure partnerships with organizations including Verily, Innovaccer, Eli Lilly, Roche, Thermo Fisher Scientific, Qiagen, and Droplet Biosciences to integrate its AI technology stack into various life sciences and medtech applications. Powell noted that Nvidia expects healthcare to become one of the largest technology industries, requiring domain-specific intelligence that exceeds the capabilities of generic models.

Nvidia and the startup Abridge announced a partnership on June 12, 2026, to develop a healthcare-specific foundation model trained on clinical conversations. The model aims to improve accuracy, reliability, and auditability in clinical workflows, including documentation, evidence grounding, and workflow automation. Built on Nvidia's Nemotron open model family, the new system will utilize Nvidia Blackwell AI infrastructure for pre-, mid-, and post-training processes using de-identified data.

By performing domain adaptation early in the training lifecycle, the companies intend to create a model capable of clinical reasoning from its foundation. Executives noted that this approach allows for the optimization of quality, cost, and efficiency across care settings and specialties. Abridge, which currently supports over 100 million annual clinical conversations across 300 health systems, plans to use this infrastructure to transition from an AI scribe tool to a comprehensive clinical intelligence platform that integrates payer and life sciences workflows.

During a keynote event in New York City on Thursday, Abridge CEO and co-founder Shiv Rao, M.D., emphasized that control over the full AI stack is necessary to manage latency and ensure clinicians can access information immediately during patient interactions. Kimberly Powell, vice president of healthcare at Nvidia, stated that the partnership reflects a shift from generative AI toward systems that can reason and perform work. According to Powell, Nvidia views AI as a full-stack computing challenge encompassing energy, chips, data centers, and foundation models, with Abridge focusing on the vertical application layer.

This collaboration adds to a series of healthcare initiatives for Nvidia, which also maintains investments in Abridge through its NVentures arm. The company has previously formed infrastructure partnerships with organizations including Verily, Innovaccer, Eli Lilly, Roche, Thermo Fisher Scientific, Qiagen, and Droplet Biosciences to integrate its AI technology stack into various life sciences and medtech applications. Powell noted that Nvidia expects healthcare to become one of the largest technology industries, requiring domain-specific intelligence that exceeds the capabilities of generic models.