Two Unitree G1 humanoids, priced at about $13,500 apiece, reportedly assisted in removing a gallbladder from a live pig, per UC San Diego's research — in one test a humanoid worked alongside a human surgeon, and in another two humanoids collaborated. A human surgeon ran them by teleoperation the entire time.
Around the same time, per reports, Johnson & Johnson's Ottava — the world's first table-integrated soft-tissue surgical robot — won FDA de novo authorization for ten general surgery procedures.
The news cycle wrote itself a pincer. Premium integration on one flank, ultra-cheap hardware on the other, both, per the popular narrative, aimed at the roughly 70-to-80% market share Intuitive Surgical's da Vinci has reportedly held for two decades. Squeezed from both ends.
It's a seductive symmetry. It's also a category error.
Ottava and the pig-surgery humanoids are not converging on the same market from opposite price points. They are not playing the same game, and they are not threatening the same thing. To see why, you have to look at what each side actually bought.
Start with what J&J won, because the coverage mostly missed it. The interesting phrase in the FDA announcement isn't "table-integrated." It's "de novo."
A de novo authorization does not merely clear a device. It creates a new device classification. Ottava is now the regulatory predicate for the category "table-integrated soft-tissue robotic surgical system."
That is the whole ballgame. Any future competitor that wants to sell a table-integrated soft-tissue robot in the U.S. will need to demonstrate substantial equivalence to Ottava's design under the 510(k) pathway.
Read that again. Substantial equivalence to Ottava. J&J's architectural choices — where the arms stow, how the table and arms synchronize, how setup automates — become the bar that every rival has to clear or explain away.
J&J didn't just get a competitor slot next to da Vinci. It got to write the rulebook for the category it invented. For years after launch, the evidentiary standard for the whole device class is a description of J&J's engineering.
That is a moat. It is the opposite of a squeeze. A squeeze compresses your margins from both sides; a de novo classification hands you the deed to the ground your future competitors have to build on.
A squeeze compresses your margins from both sides; a de novo classification hands you the deed to the ground your future competitors have to build on.
The FORTE trial that backed the authorization was, per J&J, modest in scale — a prospective multicenter study across six U.S. hospitals, 30 patients undergoing Roux-en-Y gastric bypass, all primary endpoints met at 30 days, no conversions to non-robotic surgery. Small, clean, sufficient. And now it's precedent.
None of which means Ottava dethrones da Vinci. The Street is openly skeptical that it moves share at all.
J.P. Morgan's Robbie Marcus called the smaller footprint and bedside attachment "interesting selling points" but expected Intuitive to see "very modest/immaterial" loss over time. "While OTTAVA does address certain real physician pain points," he wrote, "we think it's unlikely to materially dent Intuitive's share."
His reasoning is instructive: doc checks still point to Intuitive as the clear favorite "given its physician relationships and entrenched integration into the hospital workflow and academic programs." That is architectural incumbency — surgeons trained on da Vinci, hospitals built around it, residencies teaching it.
So the real contest is incumbency versus incumbency. A company that, per a J&J executive, has 140 years in surgery, wielding a regulatory moat and challenging a two-decade global installed base of more than 10,000 da Vinci units, per Intuitive. That's a heavyweight fight over workflow lock-in and switching costs. It has nothing to do with cheap humanoids.
Now the other flank. What did the UC San Diego team actually buy for $13,500?
A proof of concept, and an honest one. The Nature paper is described as the first preclinical demonstration of a general-purpose humanoid completing a surgical procedure on a living organism. That's a real milestone. It is also very far from an operating room.
The robots did not operate independently. Teleoperation translated the surgeon's hand movements into the robots' movements. In one test a humanoid worked alongside a human surgeon; in another, two humanoids collaborated. In every case a human was driving.
And the driving was rough. Per the team's own account, the robots reportedly needed several recalibrations during surgery, making procedures longer than with specialized systems. They also concede they're still working to reduce latency for remote operations.
Hold that recalibration detail, because it's the honest metric. The question that decides whether humanoid surgery is real is not the sticker price — it's whether anyone ever publishes a falling intervention count. How many mid-procedure recalibrations this year versus last? Until that number bends toward zero, you have a demo, not a system.
The $13,500 figure is real. It just measures the wrong thing. The cost of the chassis is not the cost of clinical readiness.
Consider what the humanoids used to hold instruments: standard surgical tools, via adapters. That's clever for a research build and a liability for a product. Adapters are latency, play, and failure points between a general-purpose gripper and a purpose-built end effector — exactly the kind of slop that dedicated systems engineered out over 25 years.
The UCSD team's own framing is telling. Senior author Michael Yip talks about "remote communities where staffing is challenging" and "austere environments like search and rescue" — battlefields, rural clinics, space. The pitch is versatility and deployability under surgeon shortage, not beating da Vinci on a hospital procurement spreadsheet.
That's a different market entirely, and it's years of recalibration away from being anything but a proof of concept. It does not squeeze Ottava. It doesn't touch it.
So where does disruption actually come from? Not from the humanoid form factor storming the OR. From below the surgical suite.
Look at ENvue Medical's ENvue Drive, a robotic-assisted tool for placing feeding tubes. Over a million feeding tubes are inserted in U.S. patients each year, most placed blindly through the nose — and, per ENvue's CEO, 3% to 5% land in the lungs, with consequences up to lung collapse.
That's the opportunity: high-volume, routine, error-prone bedside procedures, not marquee surgeries. ENvue's system runs on an already 510(k)-cleared electromagnetic navigation platform, with an in-house sensor reportedly refreshing about 40 times a second and a robotic arm that inserts under clinician supervision.
CEO Doron Besser makes the strategic point directly. Healthcare robotics started with sophisticated OR procedures — orthopedics, spine — but the simpler bedside procedures are far more common, and as robotics get cheaper to build, automating them becomes practical. Feeding tubes first, then vascular access, then peripherally inserted central catheters, millions a year.
That is where cheap automation expands access. Not a humanoid picking up a scalpel — a narrow, boring, high-volume device that makes a routine procedure safer and lets a nurse do it with confidence.
So retire the pincer. There is no squeeze from both ends.
On one end, a bigger incumbent challenged an incumbent's architectural lock-in and walked away with a regulatory classification that makes its own design the standard. On the other, a teleoperated research demo that ran slow and needed babysitting on a pig, aimed at battlefields and rural clinics, not hospital procurement.
The price gap between a $13,500 humanoid and a million-dollar da Vinci is real. It just isn't the axis anything is being decided on. Clinical readiness, tool integration, and latency are the barriers — and the recalibration count is the number to watch.
Cost expands access from below, through unglamorous bedside robots like a feeding-tube inserter. Moats get built above, through de novo authorizations. The humanoids, for now, are in neither business.