Serve adds Grubhub, and sidewalk delivery quietly becomes a shared utility

Serve Robotics has added Grubhub to a partner list that already included Uber Eats and DoorDash, with its cooler-sized robots starting to carry Grubhub marketplace orders in Chicago, Los Angeles and Alexandria, Virginia, under a deal with Grubhub parent Wonder ^[1][2][3]. The consequence is structural rather than technical: once all three large US ordering apps can summon the same robots, sidewalk autonomy stops being something one platform owns and becomes a shared utility none of them control.

Serve says it now has more than 2,000 robots deployed across the US, serving over 4,000 restaurants and reaching around three million people, a fleet built out since it promised Uber Eats up to 2,000 bots in 2023 ^[4]. It also turned on Washington, DC and San Jose with DoorDash, its seventh and eighth markets, which it says lifts combined reach to roughly eight million people across a footprint that also includes Atlanta, Dallas and Miami ^[5]. The two reach numbers in the same announcement differ by a factor of about 2.7, which suggests "reach" is doing more than one job in the release ^[6].

The utility posture has a plain operational logic behind it. A fleet with three demand sources is easier to keep busy than one tied to a single app, and idle time is what kills hardware businesses. The Grubhub launch starts with more than 100 participating merchants in Chicago and close to 200 in Los Angeles ^[7], which is roughly 300 against the 4,000 restaurants Serve already claims, under a tenth of its existing merchant base ^[8]. That is incremental order flow layered onto streets Serve has already mapped, not a new geographic bet.

The rest of the announcement reads like a company hedging against delivery fees alone. Serve is testing "micro-depots", small low-cost bases intended to let it enter new neighbourhoods without full-scale facilities ^[9]. It has launched an in-robot advertising line, including a hamburger-suited character called Chomp co-created with Grubhub, which treats a robot parked on a busy pavement as a moving billboard ^[10]. And it bought Diligent Robotics, maker of the Moxi hospital robot, in 2026, and is rolling out a faster Moxi 2.0 to hospitals in Chicago and Los Angeles on the theory that the same autonomy stack works in a corridor as well as on a street ^[11].

The Next Web's assessment is that deploying robots is not the same as earning money from them: sidewalk delivery is low-margin, with stubborn hardware, maintenance and mapping costs, and Serve is still spending to prove unit economics rather than banking them ^[12]. The publication also notes that the release is denominated in robots and reach rather than revenue and margin ^[13]. That caveat carries weight in a category that has been about to arrive for most of a decade, from Domino's experiments with Nuro to Boston Dynamics teaching Spot to reach a doorstep ^[14].

What to watch is whether the shared-layer arrangement survives contact with platform incentives. If autonomy becomes a real cost line rather than a pilot, the apps have obvious reasons to demand exclusivity, take a stake, or build their own ^[15]. Watch also whether micro-depots translate into market count growth ^[9], whether the Grubhub merchant list expands past its launch numbers ^[7], and whether Serve starts reporting anything per robot or per market instead of totals ^[13].