Something subtle but significant has happened to the streetscape of major American cities over the past year. What once seemed like a novelty — a small autonomous wheeled robot trundling along the sidewalk, ferrying takeout from a restaurant to a nearby apartment — has quietly become a routine fixture of urban life. In cities like Chicago, sightings of these machines have shifted from rare curiosities to daily occurrences. The transformation is not just visual; it reflects a deeper shift in how we think about mobility, labor, and the deployment of artificial intelligence in everyday spaces.
A Step-Function Year for Deployment
The growth numbers in this segment of the robotics industry are striking. One leading operator scaled its fleet from roughly 100 robots to 2,000 over the course of a single year, spreading them across 20 cities. The financial result has been a roughly 600% surge in year-over-year revenue, with growth running ahead of internal plans. Even more notable than the scale of deployment is the composition of revenue: about 45% now comes from software rather than from the physical act of delivery itself. The robots are no longer just wheeled couriers; they are nodes in a data and platform ecosystem that is being monetized in multiple ways at once.
This shift has come alongside something investors rarely see during periods of aggressive expansion — improving gross margins. Typically, the costs of building hardware, deploying it across new geographies, and absorbing acquisitions weigh heavily on profitability. In this case, the operational improvements were largely planned in advance, and acquisitions were chosen with margin contribution explicitly in mind.
From Sidewalks to Hospital Corridors
The most interesting strategic move is the expansion beyond the streetscape and into the hospital. Through an acquisition centered on a hospital-focused robotic platform, the company now operates delivery robots inside 25 hospitals. The use case is compelling once you understand the daily reality of nursing work: studies and operational data suggest that nurses can spend roughly a third of their working time simply walking — fetching medications, retrieving supplies, and shuttling small items from one corner of a facility to another. Every minute spent on these errands is a minute not spent at a patient's bedside.
Robots that handle the movement of goods inside a hospital allow nursing staff to redirect their hours toward direct patient care. Given the well-documented nursing shortage, this is more than a productivity story; it is a structural complement to a stressed workforce. And the economics differ from street-level delivery in useful ways. The monetization model inside a hospital tends to be more favorable, which is part of why the hospital expansion has been a contributor to better margins rather than a drag on them.
Last-Mile Delivery as the First Killer Application
The broader thesis driving this build-out is that last-mile delivery is the first killer application for robots designed to navigate complex human environments — not the only one. It is the proving ground. The same underlying capabilities that allow a small autonomous machine to dodge pedestrians, negotiate driveways, and complete a curbside handoff can be redirected toward many other domains. Hospitals are one example. Many more form factors and applications are likely to follow over the coming years, including some that will be built on top of the underlying platform by third-party developers and operators.
The framing here is deliberately analogous to the early smartphone era. We are at something like the iPhone 1 or iPhone 2 stage — there are perhaps fifteen more "generations" of capability still ahead. Capabilities, partnerships, and monetization mechanisms will all keep improving, and the curve will play out over years, not quarters.
Multiple Levers, Not Just More Robots
After last year's dramatic fleet expansion, the next phase looks different. Rather than another 20× growth in physical units, the focus is now on revenue per robot. That means pulling several monetization levers in parallel: traditional delivery fees on the consumer side, advertising surfaces on the robot itself, and software-and-data revenue streams powered by the underlying platform. A robot that earns from multiple revenue streams simultaneously can support higher unit economics without requiring proportionally more capital expenditure on hardware.
This pivot — from "deploy more" to "monetize better" — is a healthy sign of maturity. It signals that the founding bet, that the robots could survive in real-world conditions at scale, has been validated. Across the existing fleet, the cumulative distance covered every single day exceeds the span between New York and Los Angeles. That volume of operational data, accumulated in real environments, simply did not exist a year ago, and it is reshaping what is possible.
Crossing Borders and Winning Cities
International expansion is now beginning. A pilot program received a positive vote in Vancouver, which will mark the first deployment outside the United States. Negotiations with the provincial government of British Columbia are underway to set up the operating framework. Vancouver brings new climate considerations — winter conditions, road treatments, weather hardening — that will further stress-test the platform.
Meanwhile, the regulatory and political conversation in the United States has shifted in interesting ways. In one municipal hearing, a representative from a city that already hosts these robots called in to share its experience with another city that was still considering authorization. Word-of-mouth between municipalities is now becoming a meaningful adoption mechanism. Each successful pilot makes the next one easier to approve, because the unknowns shrink.
A Society Quietly Adjusting
Perhaps the most underappreciated story is the cultural one. A few years ago, the idea of a small robot rolling down the sidewalk to drop off a meal seemed like science fiction. Today, residents of multiple cities barely register them. Acceptance has come not through a dramatic public campaign but through repeated, mundane encounters in which the robots simply do what they are supposed to do without incident. Familiarity has done the work that argument could not.
That gradual normalization matters. It lowers the political cost of expansion, builds tolerance for further automation in shared public spaces, and creates the social conditions under which more sophisticated robotic systems can eventually be introduced. The sidewalk delivery robot is not just a product; it is a wedge that opens the door for a broader category of physical AI that will share our streets, our hospitals, and eventually many other environments.
The Long Haul
This is, by any honest accounting, a long-cycle business. Building robots that operate reliably in unstructured environments is hard. Convincing cities to permit them is hard. Building monetization layers that justify the capital cost is hard. None of these challenges resolve in a single quarter or even a single year. But the trajectory is becoming visible: more cities, more form factors, more partnerships, more revenue streams per unit, and a slowly thickening platform that others can build upon.
The robots on the sidewalk today are early hardware running early software. The interesting question is not whether they will succeed at delivering food — they already are — but what else they will deliver, and where, in the iterations still to come.