Uber exits quietly: Serve Robotics loses two-thirds of annual forecast

On August 11, a mandatory filing with the US Securities and Exchange Commission revealed that Uber had sold its entire stake in the delivery robot manufacturer without warning the company in advance. Five days earlier, Serve had slashed its own revenue forecast for 2026 from 26 to 9 to 10 million dollars because delivery volume through Uber Eats had fallen for the first time in seventeen quarters. In the second quarter, the company generated 3.2 million dollars in revenue while losing 64.1 million.

Serve was seen as a flagship case of sidewalk robotics: founded as a spin-off from Uber in 2021, over 2,000 robots delivered, listed on the Nasdaq. The timing is problematic. The contract with Uber expires in early 2027, and Serve only wants to renew it if conditions change.

The most revealing figure of the quarter isn't in the financial statements, but in the metrics table. On average, 792 robots were actually in operation daily, compared to 812 in the previous quarter. Operating hours per day decreased from 10,295 to 9,809. Of the over 2,000 robots delivered, fewer than half are regularly in use, and the fleet is shrinking in actual deployment while it grows on paper. This is precisely the pattern the entire industry fears: not lacking technology, but lacking utilization.

The backstory goes further than the announcement suggests. Uber bought the delivery service Postmates for 2.65 billion dollars in 2020, from whose research division Serve emerged a year later as an independent company. In 2023, both sides agreed to expand to up to 2,000 robots for Uber Eats. This goal had been the narrative for years with which Serve went public and raised capital. The fact that the parent company itself is now exiting without warning carries more weight than the mere stock sale.

Serve itself points to restructuring already underway. Business with DoorDash grew by almost 50 percent in the same quarter compared to the previous quarter. Advertising on the robots now accounts for nearly half of food delivery revenue, with recurring revenue representing more than half of total revenue. Through the acquisition of Diligent Robotics, the company has also expanded into hospitals, with seven multi-year contracts extended and two new facilities added. Even a laundry chain is now among its customers. Chief Financial Officer Brian Read calls this a deliberate decision to "concentrate fleet and capital on opportunities with the highest returns." Accordingly, Serve also lowered expected operating costs for 2026 from 160 to 170 to 140 to 150 million dollars.

The buffer is there. At the end of June, the balance sheet held 240.4 million dollars in cash and securities, which is sufficient for several quarters at the current pace. But money changes nothing about the actual problem: a robot that drives but receives too few orders is more expensive than no robot at all.

In Switzerland, this equation once worked out and then stalled. Starting in September 2016, the Post tested delivery robots from Starship for four months in Bern, Köniz, and Biberist, completing around 200 routes and 1,000 kilometers, with a load capacity of up to ten kilograms. In 2017, a five-week trial followed with Jelmoli in downtown Zurich. It didn't go any further after that, and anyone walking through a Swiss city center today won't encounter a single one of these devices. The technology was mature enough for the trial back then. The volume never was.

For everyone in this country considering purchasing or leasing service robots, it contains the most uncomfortable question of all, and it has nothing to do with sensors or battery life: Who guarantees the operating hours? A device whose profitability depends on a single major customer carries along that customer's strategic shifts, whether it likes it or not. In the second quarter, Serve lost around twenty dollars for every dollar it earned. That can be justified by growth. Without growth, it becomes a problem.