The Real Question in Driverless Logistics Is Operations, Not Demos

Founders often ask us: the video already looks good enough, so why do customers still not dare to switch?

The view up front. A demo is selected; operations are repeated. A driverless logistics vehicle is embodied intelligence on wheels, and it sits on the physical AI, embodied intelligence and robotics line we have followed for a long time. But companies on this line cannot get around three sets of numbers: right of way, reliability, and the asset. Why? Let us take them one at a time.

First: With Right of Way Unsettled, However Good the Product, It Is Only Borrowing the Road

Right of way, put plainly, is whether the vehicle may legally occupy a public lane. A licence is mostly symbolic admission; when a sensitive period comes, it stops the moment it is told to stop. But parcel delivery needs steady flow, running every one of the three hundred and sixty-five days. A demonstration inside a campus is not commercialisation. Only on open roads are you really open for business. So we care about one thing: whether there is a local partner willing to work on right of way alongside you. And without a partner? All that is left is a show flat.

Second: Reliability Comes Before Cost

Logistics customers really only recognise two words, cost and reliability. And reliability comes first. What they save is not three hundred days of money but three hundred and sixty-five — the missing sixty-five days are enough to give back everything saved before. Uptime is the share of days in a year the vehicle can actually turn out. Without reliability, the customer will not dare cut people or vehicles. And if they do not cut people or vehicles, cost reduction is empty talk.

The third set is the asset. Whether a vehicle lasts eight years or only two gives you two different sets of arithmetic (figures illustrative). Spread over eight years, the monthly figure is a predictable rent. Spread over two, it is close to three times that, as if every three months you carry an extra month of vehicle cost. Without certainty about lifespan, no party will put it on a balance sheet. So who holds it? Nobody.

In other words, this kind of robot has no novelty value, only working value. It occupies a public road resource, so it has to pay that occupation back in the volume of work it does. So we do not look at how many units were sold. We look at how many orders each one runs a day, and how many out of a hundred are sitting idle. A sale is one-off revenue. Running is what turns into repeat orders.

We want to put the other side at its strongest: with enough money and fast enough iteration, time solves both right of way and reliability. That holds, but it is not the only thing that holds. Burning cash buys momentary interest; in the end the customer looks at the numbers. As things stand, the market will return to rational buying and to the arithmetic, and the repeat rate will decide the final share.

We have seen a harsher comparison. Between a beautiful demo recording and genuinely carrying passengers with no driver on the same stretch of road, there were five full years. Five years is the whole distance from a company's seed round to volume production. So the question to ask is not "has it ever been done" but "what does the failure-rate curve look like".

This is a slow industry. A demo can pick its stretch of road, its weather and how many tries it gets. Operations cannot pick. We walk this road with a company focused on driverless delivery on open roads — and we do not walk fast.

(For the full statement of our directions, see "Core Areas of Focus" in the Archive. The above does not constitute investment advice.)