Awakend.ai tracks where AI value is moving: off trophy models and into the machines and balance sheets that run them.
This week, robots got their first sense of touch inside a simulator, Waymo announced it is bringing driverless taxis to Tokyo, and AI data centers began moving away from fans toward liquid cooling as a structural requirement.
The longer look, last, is surgery. For the first time, an AI system watched a live neurosurgery and named the anatomy as the operation happened. It did not operate. It watched, and it told the surgeon what it saw.
Today's Highlights (3 min read)
Robots can now train on touch data before they touch anything real
Waymo is bringing fully driverless taxis to Tokyo in 2027
Data centers might move from fans to liquid cooling as standard
UCL tested AI as a real-time surgical assistant for the first time
AI in the operating room is on track to 3x by 2030
THE AI UPDATE
Robots Are Finally Getting a Sense of Touch
Synaptics announced on September 15 that its touch sensing module now works inside NVIDIA Isaac Sim and Holoscan. It lets developers test how a robot grips and holds objects before any physical hardware is involved. Until now, robots could only learn to grip by picking things up in the real world and failing. The module detects pressure and grip across 60 points and runs ten times faster than foam sensors, feeding into the same data pipeline as vision and motion.
The gap it closes is in simulation. Most robotic grip failures happen because the system never learned what pressure feels like until it was already holding something real. Developers can now train on touch data before a robot touches anything. For the first time, the simulator has all three senses.
Waymo Is Bringing Its Robotaxis to Tokyo
Waymo announced on September 16 that it is bringing its driverless robotaxi service to Tokyo in 2027, partnering with taxi app Go and Nihon Kotsu, a century-old Japanese taxi operator. It will be Japan's first fully driverless commercial taxi service. Around 100 vehicles will operate across the city, bookable through both the Waymo and Go apps.
The significance isn't the partnership, but where the cars will drive. Tokyo is one of the most complex road environments in the world, dense, fast, and unforgiving. Driverless vehicles passing that test would move the conversation about autonomous transport forward in a way that San Francisco alone never could.
The Next Generation of AI Data Centers Might Not Have Fans
AI servers are generating heat that fans were never designed to handle. CoolIT published findings on September 23 showing that beyond 250 kilowatts per rack, air cooling simply cannot remove heat fast enough. Their liquid cooling system addresses this by running fluid directly to the chip, absorbing the heat, and carrying it out through a closed loop instead of blowing air over it with fans. The technology is already running in seven of the world's top ten supercomputers.
For data centers, this is an infrastructure question with a deadline. AI workloads are not getting lighter. Every new generation of models runs on denser, hotter hardware. The facilities built to run them are going to need a different approach to cooling, and the direction is becoming clear.
THE ANALYSIS
Could AI Support Surgeons During Live Surgery?
More than 15,000 Pituitary Tumour are diagnosed in the United States every year. Most stay small and cause no harm, but when they grow large enough, they press deeper into the base of the skull and can take your sight. Surgery can stop it. But removing the tumour without disturbing what surrounds it is one of the most demanding procedures in medicine. Going a millimetre wrong can mean blindness, stroke, or death.
For decades, surgeons have relied on their own training and their eyes alone to navigate this critical anatomy. Medical AI might now be able to change that and a team at University College London (UCL) is testing whether it can.
The Millimeter Problem
Pituitary surgery is performed through the nose. The surgeon advances a thin endoscope through the nasal cavity, past the sinuses, and into the base of the skull, where a marble-sized gland sits surrounded by structures that should never be touched. The arteries and optic nerves are millimetres away. An error at that scale can mean vision loss, stroke, or vascular injury.
What the surgeon sees is only what the endoscope's camera shows: a live, moving image of tissue in a space that offers almost no margin. Identifying what is what, in real time, is the central visual challenge of the procedure.
A Second Pair of Eyes
Most medical AI never enters the operating room. It reads the MRIs, flags the pattern in the pathology slide, and by the time surgery begins, it has left the room. But the UCL system works differently and assists the surgeon as the operation happens.
To build it, the UCL Hawkes Institute team went back through hundreds of recorded pituitary surgeries. Expert surgeons labelled the anatomy frame by frame, marking the nerves, the arteries, the gland as they appeared on screen. The AI trained on that library until it could recognise the same structures on its own, in new patients, in real time.
In the Operating Room
On August 26, 2026, the system entered a live operating room at the National Hospital for Neurology and Neurosurgery in London for the first time.
The patient was Rhys Hibbert, 48, whose 11mm tumour had been pressing against his optic nerves and steadily taking his peripheral vision. As Professor Hani Marcus operated, the system watched the same camera feed and named what it saw on a monitor beside the surgical field: the carotid artery, the nerve controlling vision, the structures that could not be touched.
It did not control any instrument or make any decisions, just named the anatomy. In a space where a millimetre of error can cost a patient their sight, that information changes everything.
Professor Marcus described it as "an expert second pair of eyes that never tires, never loses concentration, and has seen more operations than any human surgeon alive."
The tumour was removed successfully. The trial enrolled eight patients in total, and the system deployed successfully in six operations, with no AI-related complications reported.
What This Means: Pituitary surgery is one of the most technically demanding procedures in medicine, with almost no room for error. For years, the assumption was that AI could support surgeons beforehand with scans and data but not during.
This trial tested that assumption directly. The trial showed that AI can assist a surgeon during a live physical procedure, not only in a lab or on historical data. But it does not yet prove whether that holds across hospitals, surgical teams, and patient types at scale. The next step the researchers have outlined is a larger case series and multicenter validation before this enters routine clinical practice.
THE MARKET
AI in the Operating Room Is on Track to 3x by 2030
Growth
The AI-based surgical market is worth $1.17 billion in 2026 and is on track to reach $3.29 billion by 2030, nearly three times larger in four years, growing at 29.6% annually. Neurology is the fastest-growing application segment, driven by demand for precision guidance in exactly the kind of high-stakes procedures UCL is working in.
Pricing
The contrast is sharp. A full robotic surgical system such as the da Vinci costs between $1.5 million and $2.5 million per unit, plus more than $2,000 per procedure in disposable costs alone. Meanwhile, the UCL approach runs on local GPU hardware already installed in the hospital, adding a perception layer to the endoscope the surgeon is already using. No new machine, no separate hardware purchase.
Adoption
Caresyntax, a surgical data platform, was running across more than 3,500 operating rooms and processing 3 million procedures a year as of August 2024. Intuitive Surgical has more than 11,000 da Vinci systems installed globally. Real-time AI guidance for live anatomy recognition is not yet cleared for routine clinical use anywhere. The UCL trial is the first publicly reported first-in-human deployment of its kind.
Takeaway:
The operating room is already one of the most data-rich environments in medicine. Every procedure generates live video, sensor data, and instrument telemetry that goes almost entirely unanalyzed. The infrastructure is in place. UCL is testing whether AI can read that data during surgery, not after. That is a harder problem than analyzing a scan, and the market it unlocks, if it works, is not a niche.
The Timeline Guide:
Each timeline and graph represents the realistic stage of the covered technology, plotted from concept to scale, capturing where it stands today and when broad deployment is likely.
What did you think of today's issue?
If you think we could do better, where can we improve?
Hit reply with anything else on your mind; we actually read every note.
See you in the next upload
Note: Images in this newsletter are often AI generated for illustrative purposes only. Market forecasts are hypothetical, inherently uncertain, represent our best-guess estimates only, and should not be considered investment advice.
