Watch Fewer Launch Events, Go Look at the Line

In 2022 we wrote a line: robots today are like the internet in the nineties — not a short gust, but a cycle of twenty or thirty years. Four years on, the line still stands. What needs changing is where we stand.

The view up front. Robotics is not a single field. It is where manufacturing meets intelligence. Among our five directions, physical AI, embodied intelligence and robotics sits second. The watershed here is not the launch event. It is the production line.

Why? Because the two ask different questions. A demo answers one: can it be done. The line answers three: is it stable under harsh conditions, do the numbers work, and does the data feed back into the model — the data flywheel. A machine that cannot answer those is an exhibit.

What Stalls on the Line Is Really Just Two Curves

So what trips a machine up first once it reaches the shop floor? Two things: the share it gets right first time, and takt time. The first climbs on data volume. The second is pushed up by engineering. Every other story rests on the assumption that these two curves keep rising. An edited video is not evidence. We only look at the ramp curve.

The second ruler is the arithmetic. Building cleaning, welding, mowing, material handling — the use cases keep splitting finer, but the reason anyone pays is remarkably consistent: replacing the total cost of a job category, wages, management and risk counted together. A manufacturing client's arithmetic is plain: the equipment has to pay for itself in about a year before a buyer moves. Where the payback period does not balance, however stunning the product, it is still an exhibit. The denominator does not lie.

One more thing to be clear about: embodied machines replace people, not the previous generation of industrial robots. An industrial arm at a fixed station already has strong takt and precision, and measuring a humanoid against it is the wrong question. The real gap is where the work is unstructured, multi-task, and still done by someone bent over. Ask the wrong question and no answer, however elegant, helps.

Volume-Production Capability Is the Systematically Underrated Item

Third, we are fairly sure of this one. Most investors have never spent time on a production line. Admitting that volume production matters and being able to assess whether a team can achieve it are two different things. So the deadliest variable gets flattened out of the price. Flattened is not the same as gone. A blind spot in the market is a window for whoever understands.

This is not a guarantee. It is a judgement that has to be proven again every year. Batch delivery capability is hard to verify in a meeting room. Process, yield, supply chain — you measure them on your feet, on site, visit after visit. Judgements will be wrong. When they are, we say so.

So what do we watch next? Three things: how many lines were added this year, how long a single line runs without stopping, and whether the returning data really goes into the model. All three are more honest than a launch event. They are also unphotogenic and never trend. So for judging a robotics company, our advice is still the same: watch fewer launch events, go look at the line.

(For the full statement of our directions, see item 02 under "Core Areas of Focus" in the Archive. The above is a general view and does not constitute investment advice.)