Joining forces with NVIDIA: how surgical robot makers are challenging da Vinci in the AI race

Recently, the surgical robotics industry has witnessed a surge of partnerships with NVIDIA, with over eight companies in the field announcing collaborations with the tech giant.

For instance, the endoluminal robotics company EndoQuest Robotics plans to integrate NVIDIA's IGX Thor platform into its next-generation surgical robot system. Similarly, UK-based CMR Surgical has announced its status as one of the first global adopters of the NVIDIA IGX Thor platform. The IGX Thor is NVIDIA's newly launched top-tier physical AI and robotics platform. It delivers 5,581 TFLOPS at FP4 precision for AI computing and supports 400GbE network connectivity, representing a comprehensive leap in both computational energy efficiency and perceptual capabilities compared to its predecessor.

In addition, surgical robotics companies including Johnson & Johnson, XCath, Asensus, Moon Surgical, Virtual Incision, Neptune Surgical, and Stereotaxis have all established collaborations with NVIDIA.

An analysis of the collaboration content reveals that these surgical robotics companies are partnering with NVIDIA primarily to drive intelligent innovation and transformation in robotic automation. In both areas, Chinese companies have already established significant groundwork.

Regarding intelligence, surgical robots launched by Chinese companies, such as MedBot, Edge Medical, Tinavi, and Ronovo Surgical, are already equipped with intelligent features including image reconstruction, intelligent navigation, surgical planning, and decision support.

In terms of robotic surgical automation, Chinese innovators have progressed further than many international counterparts. For instance, Great Robotics recently showcased its "NewDawn" intelligent endoscopic minimally invasive robotic surgical solution at the 17th COA (Chinese Orthopaedic Association) Academic Congress and conducted its first public fully autonomous demonstration. Currently, its surgical robot is advancing towards high-level (Level 4+) automation, with the expectation that higher-level autonomous robots will be able to perform more core surgical steps independently, reducing reliance on surgeon experience and eliminating human error.

Simultaneously, Operate Robot pioneered the FARS (Fully Autonomous Robotic Surgery) embodied intelligent robotic platform. Previously, using its first-generation interventional surgical robot equipped with the FARS agent, Operate Robot successfully conducted an automated interventional surgery animal trial at an animal research center. The entire procedure, including guidewire manipulation, catheter delivery, and stent deployment, was autonomously performed by the "FARS" agent. The surgery achieved submillimeter operational precision by relying on multimodal imaging (DSA-CT fusion), aiming to challenge the ultimate goal of "autonomous driving" in interventional intelligence.

Within the vertical field of robot-assisted vascular intervention in Chinese academia, Professor Qi Peng's team from Tongji University, in collaboration with the clinical team of Academician Ge Junbo from Zhongshan Hospital, Fudan University, and together with leading research forces such as Siemens Healthineers and The Chinese University of Hong Kong, is jointly exploring an embodied intelligence-driven autonomous operation pathway for vascular interventional surgical robots based on simulation learning.

The team has already presented preliminary technological demonstrations at top-tier international robotics conferences, ICRA and IROS: by training a "digital twin" control model in a virtual environment and transferring it to a physical robotic system, they have achieved high-precision navigation and instrument manipulation tasks in an experimental setting that simulates human vascular structures. This marks the first validation of the 'simulation-to-reality' pathway in high-risk vascular intervention. It provides a crucial foundation for subsequent system optimization, algorithm robustness, and real-world adaptation. These latest achievements, supported by virtual simulation training and computational power, are being further optimized through their NVIDIA collaboration, integrating simulation systems, AI training frameworks, and real-time execution engines to advance intelligent surgical operations underpinned by flexible organ simulation and real-time computing.

In light of the collective partnerships between surgical robotics companies, research institutions, and NVIDIA, we must consider: Can these NVIDIA-supported surgical robots compete with the market-dominating da Vinci system? Can Chinese domestic surgical robots, which already hold a leading position in intelligence and automation, secure a significant market share in the future? What challenges remain to be overcome in realizing the future of intelligent, automated surgical robots?

In the past, the intelligence of surgical robots primarily relied on algorithms. However, with the exponential growth of data, the increasing complexity of intelligent functions, and rising clinical demands, surgical robotics companies can no longer meet these needs through algorithms alone, turning instead to the critical support of advanced computing power.

As a globally leading provider of AI computing solutions, NVIDIA holds a dominant position in computational power solutions. This is a key reason why an increasing number of surgical robotics companies are partnering with them.

Currently, the demand for computing power in surgical robots is primarily reflected in three areas: real-time environmental perception, data processing, and the fusion and analysis of multimodal data. These needs can be analyzed by examining the preoperative, intraoperative, and postoperative phases.

First, in the preoperative phase, functions such as data acquisition, 3D reconstruction of imaging data, automated medical image segmentation, lesion volume calculation, and surgical pathway planning all require substantial computing power. These have become nearly standard, essential features for modern surgical robots.

Taking Great Robotics' self-developed NewDawn surgical robot as an example, its integrated "NewDawn 3D" is the world's first all-electric, 8-axis, integrated intraoperative mobile imaging system. It achieves 0.16mm ultra-high-definition scanning and low-dose real-time registration. Its AI planning module can automatically generate patient-specific surgical plans based on individual anatomy. Functions like 3D reconstruction, data transmission, and intelligent surgical planning all depend on robust computing support.

Second, in the intraoperative phase, functions including real-time navigation, intraoperative decision support, robotic arm control, and multimodal data fusion demand powerful, real-time computing. Compared to the preoperative phase, intraoperative intelligence has significantly higher requirements for low latency. For instance, while assisting in surgery, a robot must simultaneously perform multiple tasks such as surgical field image segmentation, instrument trajectory tracking, and patient vital sign analysis. Any delay could impact surgical performance and even compromise safety.

Furthermore, during surgery, the robot must fuse and process multimodal data in real-time, including instrument tracking data, force feedback data, X-ray data, endoscopic video, system response data, and other imaging data. This ensures the precision and safety of the robotic system.

Third, in the postoperative phase, functions like postoperative assessment, quantification of operative standards, and complication analysis require computing power. For example, Intuitive Surgical plans to leverage its newly launched da Vinci 5 surgical robot to quantitatively deconstruct surgeries into key steps. By correlating the analysis of surgery type, objectively quantified workflow, and actual patient clinical outcomes, they aim to identify critical objective metrics and establish foundational standards for robotic surgery.

Intuitive Surgical also aims to combine these objective metrics with historical case data to predict outcomes for current patients. This would enable the system to provide more valuable intraoperative suggestions, allowing surgeons to integrate this feedback with the specific patient's condition to select optimal surgical approaches. All such data analytics functionalities rely on substantial computing power.

In summary, the intelligence of surgical robots is inherently built upon a foundation of computing power, with various intelligent functions requiring different levels of support. Basic perception functions like sensor data acquisition, transmission, and processing require medium-level computing power but have high real-time demands. Auxiliary decision-making functions such as surgical planning and navigation require higher-level computing power, concentrated within very short timeframes, necessitating a balance between cost and computational security for companies. Intelligent decision-making functions like automated surgical procedures demand the highest level of computing power.

Moreover, the level of computing power directly impacts robotic performance. Powerful computing can increase response speed, reduce latency across all stages for smoother operation, enhance control precision and reliability for high-accuracy maneuvers and safety, and facilitate the exploration of automated surgery.

Of course, greater computing power comes with higher costs. Companies must avoid indiscriminately increasing computational resources and instead find the optimal balance between clinical need, product performance, and cost.

While numerous surgical robotics companies are partnering with NVIDIA, the depth and direction of these collaborations vary significantly. These differentiated strategies precisely reflect each company's unique innovative focus.

First, Innovations in Intelligence. Surgical robotics companies EndoQuest Robotics and CMR Surgical both plan to integrate NVIDIA's IGX Thor platform into their systems.

Reportedly, NVIDIA's IGX Thor platform is designed for "Physical AI" and robotic applications, offering powerful computing capabilities and high-speed interconnectivity. It supports low-latency multimodal perception and real-time control. Compared to its predecessor, IGX Orin, the new platform delivers an 8x increase in AI computing power on the integrated GPU, a 2.5x increase on the discrete GPU, and doubled connectivity performance, enabling the deployment of multiple large AI models directly on the surgical robot.

Consequently, EndoQuest Robotics will leverage NVIDIA's computing power and technology to achieve low-latency sensor processing, 3D visualization, precise time-synchronized motion control, and secure cloud connectivity. The company also plans to use NVIDIA's technology platform to integrate and deploy more AI functionalities onto its surgical robot.

CMR Surgical aims to enhance the intelligent performance of its Versius surgical robot using the 250-600 TOPS computing power provided by the IGX Thor platform. The Versius system is the world's first multi-port soft tissue general surgical robot to receive FDA clearance, capable of performing complex procedures in areas such as the lungs, thymus, and esophagus.

Equipped with the IGX Thor platform, the Versius robot demonstrates a 3x improvement in real-time AI-powered blood vessel identification, with latency controlled as low as 10 milliseconds. Supported by this platform, Versius maintains a positioning accuracy of 0.3 mm and has, for the first time, achieved touchless vein placement and automated catheter insertion.

Furthermore, other surgical robotics companies like Moon Surgical, Neptune Medical, and Asensus Surgical are also advancing and innovating intelligent features through NVIDIA's other computing platforms or solutions. For instance, Moon Surgical's AI application ScoPilot, built on NVIDIA's Holoscan real-time sensing platform, is now operational on its Maestro surgical robot. This intelligent feature grants surgeons the capability to "control three instruments with two mechanical arms," providing a stable, continuously optimized, and safe surgical field of view, significantly boosting operating room efficiency.

Second, Innovations in Automation. Robotic surgical automation represents another critical frontier for innovators. For example, Virtual Incision plans to use NVIDIA Isaac for Healthcare (a developer framework for AI-powered medical and healthcare robots) to develop its next-generation surgical robot—aiming to enhance robotic task autonomy by generating synthetic surgical data.

In fact, not only international companies but also several Chinese innovators have achieved notable progress in surgical robot automation. In the field of hair transplantation robotics, the second-generation HAIRO® hair transplantation robot developed by Puncture Robotic has made breakthroughs in intelligent planning, hair flow recognition, and automated extraction and hole creation. While foreign robots often focus on partial assistance like follicular recognition, HAIRO® advances toward "automated execution + intelligent decision-making."

Even within the laparoscopic surgical robot market, long dominated by Intuitive Surgical, Chinese innovators are exploring automation. In August 2025, Cornerstone Robotics, in collaboration with a multidisciplinary research team from The Chinese University of Hong Kong, accomplished the world's first validation of autonomous surgery in a clinical setting. Their findings were published in the top-tier robotics journal Science Robotics. The surgery was performed on a live animal model (a pig), relying solely on the surgical robot's endoscopic visual feedback. Driven by algorithms, the Cornerstone Sentire laparoscopic surgical robot—already approved for use—autonomously performed three endoscopic surgical tasks: gauze grasping, vessel clipping, and soft tissue retraction. Results showed that, without human intervention, the Sentire robot could accurately and efficiently complete autonomous operations, with success rates of 83%, 77%, and 67% for the three tasks, respectively.

Additionally, previously mentioned Chinese innovators like Operate Robot and Great Robotics are also making strides in robotic automation. Automation is a crucial direction for the future of surgical robotics, but its realization requires robots to possess higher capabilities in real-time perception and intelligent decision-making, which in turn demands greater computing power. Given this, NVIDIA is expected to play an increasingly pivotal role in future innovations. Companies like Operate Robot and Great Robotics are also actively monitoring various solutions offered by NVIDIA.

It is worth noting that compared to many challengers, Intuitive Surgical's latest da Vinci 5 surgical robot incorporates various intelligent features, boasting 10,000 times the computing power of its predecessor. This significant increase supports the integration of multiple new systems and intelligent functions. However, Intuitive Surgical has made fewer moves in the realm of automated surgery, which may present an opportunity for latecomers to gain a competitive edge.

Finally, NVIDIA's Deep Integration into Surgical Robot R&D. In October 2025, Johnson & Johnson announced a collaboration with NVIDIA to incorporate the NVIDIA Isaac for Healthcare platform into its surgical robot system development workflow.

The Isaac platform is a comprehensive technology stack for robotics development, encompassing high-fidelity simulation, digital twins, and AI model training and inference capabilities. Its core technologies include the Omniverse high-fidelity simulation platform, the Cosmos world foundation model, and multimodal capabilities. Through this platform, Johnson & Johnson can leverage near-photorealistic virtual environments during the development phase for the design, verification, and optimization of surgical robots. This allows for the early development and testing of functionalities such as force feedback response, intraoperative navigation algorithms, spatial path optimization, and exception handling in a virtual setting.

Traditionally, developing a surgical robot required extensive laboratory and animal experiments, incurring substantial costs. The Isaac platform enables the testing and optimization of surgical pathways in highly realistic virtual environments, reducing the need for costly physical trial and error. This signifies that future surgical robot development will not rely solely on physical hardware but can be significantly advanced through pre-testing in virtual environments. This approach is expected to drastically shorten development cycles, lower R&D costs, and accelerate commercialization. The platform can also be used for training, helping surgeons adapt to and master robotic systems earlier.

Previously, NVIDIA collaborated with researchers from institutions like the University of Toronto, University of California, Berkeley, ETH Zurich, and Georgia Tech to introduce the ORBIT-Surgical simulation framework for surgical robot training. Designed to enhance surgical team skills and reduce surgeons' cognitive load, ORBIT-Surgical includes over a dozen benchmark tasks for surgical training, such as single-handed gauze pickup, shunt tube insertion into a vessel, needle positioning, needle handoff between arms, and threading a needle through a ring.

In 2024, the ORBIT-Surgical research team demonstrated at the ICRA robotics conference how a digital twin trained in simulation could be transferred to a physical robot in a lab setting. This demonstration validated that researchers can train and optimize surgical robots in virtual environments before transferring the capabilities to physical systems, a method that can significantly reduce development costs and shorten cycles.

In summary, technologies like simulation platforms and digital twins offer substantial advantages for surgical robot R&D. In this context, NVIDIA is poised to occupy a central role in the future innovation of surgical robotics, especially as the field transitions from a phase of hardware innovation to a new era driven by software and computing power, accelerating progress toward intelligence and automation.

However, while the destination is clear, the path is winding. Intelligence and automation are universally acknowledged as future directions, but achieving them is fraught with challenges. For instance, surgical robots require vast amounts of accurate data, yet they face the "data silo" problem. The clinical application of robotic automation also encounters regulatory and ethical hurdles requiring support from oversight bodies. Furthermore, the industrial ecosystem needs improvements in areas like the supply chain, industry chain, and reimbursement systems. These difficulties must be addressed one by one by innovative companies.