Artificial intelligence in its physical form is starting to enter the hospital floor. A growing number of health systems and technology vendors are testing machines that combine AI decision-making with physical action. This category is increasingly called “physical AI” to be distinguished from, say, generative AI tools that only generate textual recommendations.
Physical AI systems perceive their surroundings, make decisions, and act in the physical world. In hospitals, that could mean transporting medications, assisting surgeons, monitoring patients, guiding rehabilitation, or coordinating logistics. The technology also extends well beyond the robotic-assisted surgery systems already common in operating rooms.
NVIDIA has moved to position itself at the center of this shift. The company recently introduced a healthcare robotics platform built from four components:
- Open-H, a surgical video dataset;
- Cosmos-H, a synthetic data generator;
- GR00T-H, a vision-language-action model designed for clinical tasks;
- and Rheo, a blueprint for building hospital digital twins.
The platform is meant to give hospitals and robotics developers a shared foundation for training and testing clinical AI systems before deployment.
Executives from Northwell Health, NVIDIA, and Expper Technologies discussed the technology’s near-term path. Their consensus is that the priority for hospitals is not choosing which robot to buy, but identifying which physical tasks create the most risk, delay, or workforce burden, and then measuring whether new technology actually improves them. They expect the next phase of adoption to unfold through narrowly scoped systems handling specific hospital functions, rather than general-purpose humanoid caregivers.
Robotic-assisted surgery remains the most established use case. Industry estimates put robotic-assisted procedures at roughly 60 percent of surgeries performed at major hospitals, with systems such as Intuitive Surgical’s da Vinci 5 leading adoption. A recent meta-analysis of 25 peer-reviewed studies published between 2024 and 2025 found AI-assisted robotic surgery reduced operative time by 25% and intraoperative complications by 30% compared with manual techniques, while improving targeting accuracy by 40%. The same analysis reported average recovery times fell 15% and healthcare costs dropped 10% compared with conventional procedures.
Beyond the operating room, adoption is spreading into long-term and memory care. At the Eskaton retirement and assisted living community in Sacramento County, California, researchers from the Betty Irene Moore School of Nursing at UC Davis have started a 12-month study examining how a companion robot named Abi affects residents with memory loss and dementia. Nursing professor Roschelle Fritz is leading the study, described as the first of its kind to track a companion robot’s role in memory care over an extended period. Family members at the facility have reported the robot appears to support social engagement among residents.
Investment in the wider AI-health sector has continued climbing. Digital health startups raised roughly $4 billion in venture funding during the first quarter of 2026, an increase of about $1 billion over the same period last year and the strongest first quarter since the pandemic-era peak, according to funding trackers. Rock Health, which has historically tracked AI-specific deals separately, retired that category this quarter, saying AI has become standard across digital health rather than a distinguishing feature.
Hospitals and vendors still face open questions around governance, safety validation, and liability before physical AI healthcare systems move from pilot programs to routine infrastructure. Healthcare leaders say those frameworks, rather than the robots themselves, will determine how quickly the technology scales.
