TDK Ventures 创始人：灵巧性是物理 AI 的真正瓶颈

Nicolas Sauvage, founder and president of TDK Ventures, with ANYbotics’ quadruped robot. | Source: TDK Ventures

A robot picks up an object, moves it, and puts it down successfully. Is it dexterous? Perhaps. But now ask it to do that successfully hundreds of times, while objects move slightly, its grip changes, components wear, and unexpected failures occur. That is a much harder test of dexterity, and a much better measure of whether a robot is dexterous enough for the real world.

We are making rapid progress in what robots can manipulate. However, useful dexterity isn’t simply the ability to complete an individual action once. It is the ability to sense, grasp, move, adjust and recover reliably enough to complete an entire workflow without constant human assistance.

A robot that succeeds 99% of the time sounds ready for work, but a job rarely consists of a single action. Consider a workflow that requires 100 physical actions in a sequence. At a 99% success rate for each action, the probability that all 100 succeed on the first attempt is just 36.6%. Increase the success rate to 99.9%, and the probability of completing the workflow rises to 90.5%. At 99.99%, it reaches 99.0%.

This is, of course, a simplified example that assumes equal and independent success rates, but it illustrates an important point. A robot with a 99% success rate is not necessarily 99% automated.

Small reliability gaps become much bigger when repeated across an entire workflow. This means that a robot that looks almost flawless in a demonstration can still be far from ready for autonomous deployment. The difference between two nines and four nines may ultimately decide whether a robot is ready to work alone.

The next major challenge in robotic dexterity is not simply teaching robots to do more, but making the entire system reliable enough to turn those capabilities into useful work.

Dexterity is a system, not a hand

Robotic dexterity is often discussed as a hardware race: more fingers, more joints and more freedom of movement. While those capabilities obviously matter, the customer does not buy degrees of freedom. The customer buys useful work.

Completing that work requires a closed loop across vision, position, contact, force, pressure, grasp, movement, and adjustment. Power, safety, flexibility, and recovery also matter.

Not every workflow needs touch, as in some cases, vision and movement data may provide enough information. But touch can also be useful in telling a robot that an object is slipping, the contact is unstable, too much pressure is being applied, or a component is properly in place.

The objective therefore isn’t to pile every sense onto every robot, but to give the system enough of the right information to act, and recover, reliably.

TDK Ventures helps early-stage materials, energy, cleantech, industry 5.0, mobility, and health tech startups scale. | Source: TDK Ventures

More nines come from learning

Recovery is a huge part of dexterity. If a robot begins to lose its grip on an object, the useful measure isn’t if the original grasp was perfect; it’s whether the system can detect the slip early enough to correct it before a small error causes the entire task to fail.

A failure discovered during deployment might expose a problem in software, sensing, or control, or it might reveal that the hardware itself needs redesigning. Changing the hand changes movement, sensing and calibration, so data collected with one physical design may not transfer directly to another. Real-world deployment is therefore a critical part of the development process.

Deployment reveals slips, poor grasps, wear, and failed recoveries. Those experiences become data that can improve software, simulation, sensing, control, and hardware. Each cycle can make the next deployment more reliable and less expensive.

A technological breakthrough may begin the journey, but it is the learning system around it that multiplies its impact. Robotic dexterity should be viewed in much the same way. Additional nines emerge from repeatedly deploying, learning and improving.

Editor’s note: Physical AI, enabling technologies, and humanoids are among the session track topics at RoboBusiness 2026, which will be on Oct. 20 and 21 in Santa Clara, Calif. Register now to attend.

Register now and help us celebrate 20 years of RoboBusiness!

The right body depends on the work

Reliability also requires us to resist the temptation to equate more human-like robots with better robots. The starting point should be the job. From there, work backwards to determine the environment, physical form, and system required to perform it.

Dexterity should follow the same logic. The most human-like hand may not be the best hand for a particular workflow. The goal is not maximum dexterity. It is the minimum sufficient dexterity needed to complete valuable work reliably, and we are already starting to see this principle being successfully deployed.

Agility‘s Digit is a bipedal robot designed for logistics and industrial work. Digit can connect directly to specialized tools suited to the job, rather than requiring a five-fingered hand or being made to look as human-like as possible. It has the form and dexterity the work actually requires.

ANYbotics applies the same principle to a different environment. Its four-legged robots can navigate stairs and uneven surfaces at industrial sites, while inspecting hazardous areas reduces the need for human exposure. In one deployment, its technology has reportedly carried out more than 33,000 inspections across 450 inspection points.

Starship Technologies takes a different approach. Its delivery robots need to travel along pavements, so a small, wheeled form is better suited to the job than legs. That purpose-built approach has helped Starship to report more than 10 million autonomous deliveries.

These robots look and operate very differently because they perform very different jobs. Yet they demonstrate the same principle: The right physical form is the one that fits the work and can perform it repeatedly in the real environment.

Our investments in Agility, ANYbotics, and Starship have changed the questions I ask as an investor, particularly during our ongoing deep exploration into dexterity. I am less interested in which robot looks most human than in which complete system can reach the reliability its customer actually requires.

Sauvage with Agility Robotics’ Digit humanoid robot. | Source: TDK Ventures

Measure the job, not the demonstration

This changes how we should measure progress in physical AI. The success rate of an individual action shown in a demonstration tells us very little about whether a robot is actually ready to work autonomously in the real world.

I want to know what percentage of complete workflows finish without human assistance. When something goes wrong, how often can the robot recover without help? How does performance change when the object, lighting, position, wear or surrounding environment changes?

Then there are the less glamorous questions that become extremely important at scale. What breaks first under repeated use? How quickly can it be diagnosed, repaired and returned to work? What are the speed, uptime, and cost of each successfully completed job?

These are ultimately better measures of dexterity than what a robot can demonstrate once. A robot does not need the most human-like hand, the greatest number of joints or the widest possible range of skills to create value. It needs the right combination of physical form, sensing, control, and recovery to complete a valuable job reliably in the environment where it will actually work.

That is where I believe the next phase of progress in robotic dexterity will be won. The industry has made extraordinary advances in what robots are capable of doing. Now the challenge is turning those capabilities into systems that customers can trust to perform useful work repeatedly, recover when something goes wrong, and improve through deployment.

About the author

Nicolas Sauvage is the founder and president of TDK Ventures, the global corporate venture capital arm of TDK Corp. He launched the firm in 2019 around a simple principle: entrepreneurs come first.

Today, TDK Ventures manages $500 million across four funds and backs early-stage startups globally, across AI and compute, energy, industrial and robotics, mobility, and advanced materials.

The post Why dexterity is physical AI’s real bottleneck appeared first on The Robot Report.