Hospital robotics adoption slows as evidence and costs come under scrutiny

On 1 October, The Robot Report published its round-up of the ten most-read robotics stories of September 2026. The clinical entries sat alongside warehouse and humanoid news rather than leading it. That same day, CNBC reported that Blue Cross Blue Shield Association had estimated hospitals' use of AI-assisted medical coding contributed to close to $1 billion ($942 million) in additional costs for its health plans between 2023 and 2025.

The two stories have little in common on the surface. Underneath, they describe the same problem: hospitals are bolting new automation onto an old billing and delivery machine, and the machine is starting to push back. The surgical evidence is the sharper test. On 1 October, News-Medical reported that robotic assisted hip and knee replacements show no clear benefit over conventional methods, and Medical Xpress reported that robotic hip and knee surgery fails to cut the overall risk of repeat surgery. MedPage Today, writing on 30 September, called the U.K. studies an "epic fail" for the case that robots improve joint replacement outcomes. Bioengineer.org framed the same work as a landmark U.K. study showing no clear advantage.

These are not fringe claims. They land at a moment when hospitals in several countries are expanding robotic programmes, and when vendors are pushing into procedures beyond the ones that made their names.

What the newest sources actually say

The Robot Report's 1 October round-up is a useful map of the industry's own priorities. It lists Boston Dynamics opening its Robotics Metaplant Application Center at Hyundai Motor Group Metaplant America to train Atlas humanoids for automotive work, Agility Robotics unveiling Digit 5 as it awaits a SPAC merger, and Intrinsic releasing Intrinsic Core, an open-source set of ROS-compatible capabilities. It also notes that shares of Hangzhou-based Unitree Robotics fell nearly 40% in September from their first-day closing price on 19 August, erasing roughly $20 billion in market value from that point.

None of that is hospital surgery.

But the capital discipline visible in the humanoid and industrial segments is the same discipline now being applied to surgical robots, which have longer sales cycles, heavier regulatory burdens and, in many markets, reimbursement tied to procedure volumes rather than technology. On 1 October, Medical Device and Diagnostic industry asked whether compact robotics can finally convert orthopedics' 80% non-robotic surgeon market. That is the commercial version of the clinical question. If the U.K. studies hold, the pitch to that 80% becomes harder: a surgeon who has not bought a robot now has published evidence to justify waiting.

Korea Biomedical Review reported on 30 September that urology could lead physical AI adoption on the path toward autonomous surgery. MassDevice reported on 30 September that Intuitive Surgical received a CE mark for da Vinci 5 cardiac procedures, a concrete regulatory expansion rather than a marketing claim. Kalkine Media reported on 1 October that Intuitive Surgical had won robotic approvals. The same day, Healthcare Asia Magazine reported that advanced tech demand is expected to drive China's robotic surgery growth.

So the picture is not a retreat. It is a split: approvals and expansion in some specialties, doubts and cost scrutiny in others.

Billing, coding and the trust problem

The CNBC report on 1 October deserves separate treatment because it is the clearest example of automation changing hospital economics without changing care. Blue Cross Blue Shield Association estimated that AI-assisted medical coding contributed to close to $1 billion in additional costs for its health plans between 2023 and 2025, with roughly 70%, or $653 million, tied to additional diagnoses that were not accompanied by a change in care, according to Luke Chalker, BCBSA's senior vice president of product and data science, who spoke to CNBC.

Chalker stopped short of attributing the entire increase to AI. "While multiple factors contribute to coding intensity, the findings suggest AI-enabled coding and documentation tools are playing a role," he told CNBC. Christopher Whaley, a health economist at Brown University who studies hospital coding, told CNBC that AI appears to be "accelerating, and in some sense making it easier to capture, the existing and underlying billing incentives that are in the system."

That distinction matters for anyone tracking surgical robots. The same hospitals buying robots are buying coding tools, and the financial return on the two is measured in different places. A robot that improves a surgeon's ergonomics but not the patient's outcome still generates procedure volume. A coding tool that adds a secondary diagnosis generates revenue without touching the operating room. Both can look like productivity from the inside.

WIRED reported on 1 October that measles is forcing hospitals to adapt to a new normal, with U.S. cases topping 3,600, the highest yearly total since 1991, and Pennsylvania recording 943 cases, including 176 hospitalizations and five measles-related deaths as of 30 September. That is a different kind of operational stress, but it shows how quickly hospital capacity assumptions can break. Robotics programmes compete for the same capital and the same staff attention.

Where the infrastructure is heading

NVIDIA used the ROSCon conference in Toronto to release Isaac ROS 5.0, a collection of GPU-accelerated packages built on ROS, according to the company's blog on 22 September. The release adds support for ROS Lyrical and Ubuntu 24.04, and NVIDIA says FoundationPose now provides an agent-ready inference library that lets robots perceive and track object position and orientation up to 5.5x faster. RealSense is optimizing its latest AI-native 3D stereo depth cameras, including the D585 Pro, for Isaac ROS and the NVIDIA Jetson Thor edge AI platform.

On 28 September, The Robot Report wrote that Gecko Robotics is working with NVIDIA's Open Agent Safety Platform, which includes the OpenShell open-source software and the Sentry reference system design. NVIDIA CEO Jensen Huang said in a statement that "safety and security require full-stack engineering." Justin Boitano, vice president of enterprise AI at NVIDIA, said during a press briefing that "for the agent economy to grow, we need a trusted foundation across silicon and software."

Microsoft Research argued on 23 September that running physical AI inference exclusively on onboard GPUs can limit robot performance, battery life and scalability, and that offloading inference to edge or cloud GPUs can offer significant advantages. Across representative mobile manipulation workloads, offloading improved task success rates and enabled larger AI models, according to the blog.

The geopolitics are moving too. SCMP reported on 29 September that only 10 per cent of Hong Kong firms surveyed by Accenture had integrated generative AI across multiple business functions, compared with about 21 per cent of mainland enterprises, according to Robert Hah, Hong Kong office lead at Accenture Greater China. CNBC reported on 27 September that Chinese AI models accounted for 57%-67% of tokens used on OpenRouter in the week of 14 September, up from 6%-13% in February, and rose to 55% on Vercel in August from 11% in January.

For hospitals, that matters less as a chip story than as a procurement story. SCMP reported on 22 September that Chinese chipmaker Hygon launched its Hygon 1000 series processors for robotics and industrial edge devices, built on its x86-compatible C86 architecture, with four cores and eight threads and support for DDR4 and DDR5 memory. If physical AI hardware gets cheaper and more local, the cost case for hospital robots changes again, in both directions.

The European argument

Il Sole 24 Ore published an interview on 1 October with Bruno Siciliano, professor of automation and robotics at the University of Naples Federico II and director of the Prisma Lab, who won the 2024 Pioneer Award from the IEEE Robotics and Automation Society. Siciliano said Europe should develop a model of physical AI based on interaction with humans rather than replacement, and argued that a fully anthropomorphic robot with legs is not always productive in an industrial plant because of cost, stability and movement in confined spaces.

He pointed to BMW's rethink of its use of Figure AI humanoids at its South Carolina plant as a well-known example. Siciliano also said the euRobin network discussed an AI-based robotics act at the European Parliament in Brussels in early September, with a proposed budget of 30 billion euros and a target of 33% of the world market by 2035, modelled on the Chips Act.

IEEE Spectrum profiled Barbara Mazzolai on 22 September, describing her work on bioinspired robotics and sustainability robotics. The Robot Report's 1 October round-up noted that Boston Dynamics veteran Marc Raibert's new startup Dynamic Creatures is focused on robotic characters for entertainment, and that startup Flourish announced plans to sell Flourish One, a $3,555 wheeled humanoid, with early adopters able to teach it household chores in under 30 minutes using a smartphone.

None of these are surgical systems. They are the pipeline that will supply the components, models and expectations that surgical robotics inherits. The question hospitals face is no longer whether robots can perform a procedure. It is whether the evidence, the billing and the maintenance costs justify the next purchase order.