Surgery's Next Robot Is the Room: Part 1

A Tuesday morning, a few years from now

The first patient of the day is on the way to OR 6 for a gallbladder removal. The room knows the plan: the procedure, the surgeon, the instruments, and the case uses a new type of surgical robot. The trays on the back table were matched to the case before the patient arrived, so the scrub tech opens only what this surgeon will use.

The surgeon stands at the table with the instruments they decided to use for this case. Next to them, the robot's arms are set up for this surgeon's preferences and this patient's anatomy, and hold the laparoscope steady. When the surgeon wants their intracorporeal camera angle to follow an instrument, the scope follows it, and both of the surgeon's hands stay on the work.

The robot sees the local surgical field and internal surgical workspace in fine detail. Sensors around the room keep a live 4D picture of everything around the table: what is ready, where people and equipment are, and what phase the case is in, recorded moment by moment to the system of record and other relevant systems. The robot's picture and the room's sit on one timeline and in one shared map. Nobody has to stop and report where things stand, so the team’s attention stays on the patient.

The room follows the case phase by phase, so it can estimate when it will end and updates that estimate the moment something changes. When the surgeon starts to close, the patient’s family gets a message that surgery is nearly done, and recovery knows when to expect them. Transport is already on the way for the next patient, timed so they arrive as the room is ready instead of waiting on a stretcher in the hallway. The cleaning team gets a notification of when to arrive with the severity of cleaning required, not a phone call. The count has been kept as the case went, so the nurse confirms it instead of starting from scratch. The next case's instruments were checked against its plan earlier in the day, and anything still needed is already on its way from central supply, timed to arrive before the patient does. In time, another robot will be the one that brings it.

After the patient leaves, a draft record of the case is waiting, built from what every system saw, with the evidence attached. The team checks it and signs it, and the nurse who would have spent the end of the shift on paperwork goes home on time. The day’s last case starts when it was planned to, and a patient who might once have been bumped to tomorrow has their surgery today.

Each part is needed, none of them takes over the others, but all of them work as a cohesive unit. The room coordinates without directing, collecting what every system knows and sharing it with the rest. The surgical robot still decides how it moves when it’s instructed to, and the clinical team still directs the case. Working together, they keep the whole day on plan: the room senses what is happening, every system knows what comes next, and each one acts on it. In that sense, surgery's next robot is the room: not any one machine in it, but the robotic surgical assistant, the sensors, the software and the computing underneath, working as one with the surgical team.

How this is becoming reality

Three things are arriving at the same time.

The first is pressure. In US hospitals, an operating room minute cost an average of $57.71 in 2022, up 54% since 2014. Today, that is closer to $70 a minute when accounting for inflation, and a minute that runs into overtime costs more. Worldwide, the Lancet Commission on Global Surgery counts a shortage of 1.1 million surgeons, anesthesiologists and obstetricians, and perioperative nurses and surgical techs are even harder to hire, so many hospitals lean on overtime, travel nurses, and heroics to cover the gap.

The second is technology that works at the bedside. Surgical robots now run AI in real time on the robot itself, and a room can be sensed in 3D without anyone having to manually report what happened. That finally lets clinicians and staff focus on the patient instead of on the paperwork and admin around them.

The third is a faster and safer way to build. Physics-based simulation lets robots and workflows be trained and tested before they reach a patient.  Developers can vary room layouts, equipment positions and sensor visibility, training and evaluating how a robot responds before testing those behaviors in the real world.

Most of the trial and error happens in a virtual room instead of a real one. A new robot or a new way of working reaches hospitals sooner, and it arrives already tested against the rare situations a single hospital might wait years to see. NVIDIA and Foxconn report that digital twins of Taiwanese Hospitals cut deployment time by 40%. For a hospital, that means value sooner and fewer surprises once a system is live.

What is missing is the connection between them. The OR is one of the most technology-dense rooms in the hospital, and yet, it remains one of the least connected.

"Ask where a case stands and someone still phones the room, or walks down the hall and opens the door to look. Every system in there sees its own slice, and the team fills in the rest with a whiteboard, memory, and heroics." - Thomas Knox, VitVio

Adding more devices will not fix that on its own. Each new one adds another view of the room that nothing else can use.

Three views of one room

The operating room we describe needs three capabilities: a robot that assists the surgeon at the table, a room that understands itself and orchestrates, and a platform that lets both learn safely and run in real time. They come from different companies that have separate origins but that are working together.

Inside out: the robotic surgical assistant

Surgical robotics started at the bedside. AESOP, the first robot the FDA cleared for surgery in the 1990s, held the laparoscope steady for the surgeon. A laparoscope is the thin camera surgeons insert through a small incision to see inside the body.

Moon Surgical has gone back to the bedside with a different idea of what the technology is for: assisting the whole surgical workflow. Maestro's two arms hold and position the laparoscope and instruments while the surgeon stays sterile at the table, which frees the rest of the team for the patient care only they can provide. The goal is what Moon calls Scalable Surgery: minimally invasive surgery that is predictable and repeatable, so a hospital can treat more patients in the same allotted time. In a 2026 study of gallbladder removals, operative time varied 35% less with Maestro, with the same average time. That is the point. Predictable cases are what let a schedule hold.

Maestro is built to understand the procedure from the inside out, sensing and responding in real time at the table with its computing on board. Inside the patient, it reads the laparoscope's video, the same view the surgeon works from, to recognize instruments and anatomy. At the table, onboard sensors perceive the workspace within the surgeon's reach, and each arm feels the forces it meets. That perception powers ScoPilot, the first FDA cleared AI to move an instrument in a patient, which lets the scope follow a chosen instrument on demand. The latest releases also set the arms up automatically from the surgeon's preferences and the patient's anatomy, and generate the post-operative note with a third party model.

Maestro is now in use on 22 systems across the US and Europe, with more than 4,000 procedures performed

"Holding a camera and tissue for two hours is a poor use of a skilled team member. Maestro takes that over, so your team can get back to the work only they can do. Beyond managing the instruments, it can check the room, build the op note, keep the schedule predictable, adapt to the surgeon, and even help the team learn. Better assistance, and a better use of your team." - David Noonan, Moon Surgical

Outside in: the room that understands itself

Knowing what is happening in an operating room has always depended on people. Nurses enter key moments into the record by hand, from the patient arriving to the first incision, and the status board is only as current as the last entry or phone call. Now a room can be sensed in 3D and understood live, as the case happens.

A robot sees its field of view in fine detail. The room sees everything around it. Room-scale spatial intelligence, which VitVio builds, puts calibrated sensors around the OR and turns them into a live 3D model: people as positions and movement, equipment as objects in space, the case as a sequence of phases, actions, tools, and activities. Sharing the calibration allows new devices with onboard sensors, such as a robot, to be registered dynamically into the same shared space all other devices are using.  That shared understanding also provides a foundation for simulation. A virtual operating room grounded in real layouts and workflows gives developers an environment in which to train robots for the conditions they will encounter.

Every system receives structured context, not raw sensor data: what happened, where, when, why, and how sure the system is.

"The once disparate systems now share one picture of the room, a new device can join simply by sharing its information in a common language. That places it in the same space as everything else, gives it the context that the room already knows, and lets what it knows add back to that picture for everyone else." - Thomas Knox, VitVio

The backbone: the platform that lets both learn and run

Neither view works without computing in the right place. A robot has to respond in milliseconds, so its computing rides on the robot. A room's sensors produce more data than it makes sense to send anywhere, so it is processed on GPUs on servers inside the hospital. Both run on NVIDIA computing today. For example, Maestro uses NVIDIA IGX with Holoscan for onboard computing and real-time sensor processing, bringing AI capabilities directly into the surgical workflow.

Building physical AI requires three separate computing platforms: one to train models, one to simulate and test robot policies in physically-accurate virtual environments, and one to run them in real-time at the point of care. NVIDIA Isaac for Healthcare brings those together for medical robots: physics-based simulation, simulated sensors, pretrained models, and a path to deploy the same software at the edge. Omniverse-based simulation provides virtual environments in which developers can model rooms, robots and sensors; Isaac Lab supports robot-policy training and evaluation; and Cosmos can expand the variation in synthetic training data. Together, these tools help developers move from a robot that works in one setup toward one that can adapt across many. This matters because surgery will never produce the volumes of data that trained self-driving cars. Real data from operating rooms, physically accurate simulation and computing at the bedside have to work together to close that gap.

"Most medtech companies are rebuilding the same foundations: sensing, compute, simulation. A common platform frees them to work on what makes their product different. It lets a robot, a room and a scheduling agent share context and coordinate care." - David Niewolny, NVIDIA

What changes when they work as one

For separate systems to act as one, they need three things in common.

• One picture of the room. Where people, equipment and the robot are, in one coordinate space; what phase the case is in; and how sure each system is.
• One plan for the day. The schedule and what each case needs, dynamically updated as the day changes, so every system works toward the same next step.
• Clear lines of authority. The room informs and documents, the clinical team directs and decides, and the robot assists the surgeon. The hospital sets the rules, and any step that happens on its own is one the hospital has switched on and can switch off.

With those in place, each view fills the other's gaps. The robot covers the surgical field, the instruments and its own setup in fine detail. The room covers where people and equipment are and where the case stands. A predictable case lets the room plan the next one, and a room that knows what comes next lets every system, and person, get ready for it.

"Hospitals plan around the case they can predict, not the fastest one. When the robot makes each case more consistent and the room can see where it stands, that consistency becomes time the hospital can use." - David Noonan, Moon Surgical

We have seen a first small piece of this. In September, a Maestro was wheeled into a simulation operating room in Manchester, England, and for one day its sensor feeds streamed into the room's 3D model, in the same coordinate space. The room gained two viewpoints close to the table, where people often block its sensors. It was a prototype in a simulation center, not a hospital. It showed that the two views can be joined, and it's the first step in a much larger vision both VitVio and Moon Surgical share.

A next step is to use that shared spatial understanding to simplify robotic interactions. The aim is to create the learned behaviors that guide a robot’s actions—in virtual rooms that reflect real operating environments.

None of this needs a breakthrough. The foundations are here.  The hard part is getting systems built by different companies to agree on where things are, and on who decides.

"The hospital of the future depends on intelligence that works together. When robots and AI share an understanding of the operating room, they can do more of the work around surgery, giving clinicians more time to care for patients." -David Niewolny, NVIDIA

What it is this all for?

Surgeons, nurses, anesthesia teams and techs are the scarcest resource a hospital has. The point is to give them back the time they now spend on phone calls, searching, counting and paperwork, and to get more care out of the operating rooms hospitals already have. For patients, it means less time waiting, fewer delays and cancellations, and a team unburdened administrative tasks

That is also the bar the profession has set. AORN's 2026 position statement on AI asks for clear governance, transparency, privacy protection, and a role that complements perioperative nurses.

"Success means giving surgical teams time back so hospitals can care for more patients and more people can get the care they need. " - David Niewolny, NVIDIA

Next: how we get there

An operating room that works as one system will not arrive all at once, and it should not. In Part 2, after NVIDIA GTC Berlin, we set out the road as we see it: how autonomy should come in stages, why every robot and workflow should rehearse in simulation first, how a robot keeps improving after it ships without becoming less safe, and what the people in the room need before they trust it.

No single company will build the operating room that works as one. The pieces are coming from the ecosystem. What we can agree on now is how they share a common space and framework for collaboration.

About the authors

• Thomas Knox is co-founder and CEO of VitVio, which builds 3D spatial intelligence for operating rooms. He previously led product/innovations at Kiwi.com, AiFi and VusionGroup.
• David Noonan, PhD, is CTO of Moon Surgical, maker of the Maestro system. He previously led systems, algorithms and robotics at Auris Health and holds more than 50 US patents.
• David Niewolny is Senior Director and Global Head of Business Development for Healthcare and Medical at NVIDIA, where he works with medical technology companies building on NVIDIA platforms.