THE WORLD'S LARGEST fleet of industrial robots — more than one million strong, deployed across 300-plus fulfillment centers — can sort, stack, and shuttle inventory with extraordinary precision. What it cannot do is walk up a flight of stairs and leave a package at your door. Amazon's acquisition of Rivr, a Zurich-based robotics startup formerly known as Swiss-Mile and valued at $110 million as recently as August 2024, is the clearest signal yet that the company views last-mile delivery, not warehouse automation, as the frontier that still matters most.
The deal makes strategic sense on paper. Amazon has spent the better part of a decade automating the inside of its logistics network; it now expects warehouse robotics to generate $12.6 billion in savings between 2025 and 2027, according to internal projections. Morgan Stanley estimates the shift could yield $2 billion to $4 billion in recurring annual cost reductions by 2027 alone. But the last mile — the stretch from delivery van to doorstep — remains stubbornly, expensively human. It accounts for roughly 53% of total shipping costs, averaging about $10 per urban package, and labor represents 50–60% of that figure. For a company that delivered more than nine billion items same-day or next-day globally in 2024, even marginal per-stop savings compound into serious money.
Scouting for answers
Yet anyone paying attention will recall that Amazon has been here before. In 2019, the company launched Scout, a cooler-sized sidewalk robot designed to deliver packages autonomously. By 2022, the program was dead. Scout required human "ambassadors" to accompany it on every run — which rather defeated the point — and customer feedback was tepid. About 400 employees were reassigned. The episode became a cautionary tale: fully autonomous last-mile delivery, for all its theoretical elegance, crashed into the messy reality of curbs, stairs, weather, and suspicious dogs.
But Rivr represents a fundamentally different bet. Where Scout tried to replace the driver, Rivr's four-legged, wheeled robots are designed to assist one. The concept is collaborative: a human courier drives a van loaded with packages while Rivr's robot handles satellite deliveries nearby, shuttling parcels from the vehicle to separate doorsteps. Packages are stored inside the robot's torso and dispensed on arrival. The driver does more stops per route; the robot handles the physical last hundred feet. It is not autonomy as a substitute for labor — it is autonomy as a force multiplier.
Amazon informed its delivery service partners about the acquisition earlier this week, framing it as a test of how robotics might integrate into existing operations. The language was careful, emphasizing "safety outcomes" and "delivery experience." But the subtext is unmistakable: the company that plans to replace up to 600,000 warehouse positions by 2033 is now probing whether it can squeeze more output from its 2,000-plus third-party delivery partners without hiring more drivers.
Paws and effect
Still, the gap between concept and commercial scale is considerable. Rivr's robots have been piloted with JustEatTakeaway.com and logistics startup Veho, but these were small-scale trials, not proof of unit economics at Amazon volumes. The broader delivery robot sector remains nascent, albeit growing; Starship Technologies leads with roughly nine million autonomous deliveries across 270 locations, while Serve Robotics — backed by Uber and Nvidia — has completed just over 100,000. DoorDash launched its own Dot robot in late 2025. None has demonstrated profitability at meaningful scale. Market estimates vary wildly, with analysts projecting anywhere from $800 million to $3.2 billion in global delivery robot revenue by 2030, depending on whose methodology you trust.
Amazon's own recent history underscores the difficulty. Blue Jay, a multi-armed warehouse robot unveiled with fanfare in October 2025, was quietly discontinued months later due to high production costs and manufacturing issues. Zoox, the company's autonomous vehicle subsidiary, remains focused on robotaxis rather than package delivery. And the company's $200 billion in projected 2026 capital expenditure is overwhelmingly earmarked for AI infrastructure, not delivery robotics.
The Rivr acquisition, then, is less a declaration of victory than an admission of where the hard problems remain. Amazon has largely solved the challenge of moving goods through its own facilities at machine speed. The final hundred feet — across sidewalks, up staircases, past garden gates — is where physics, regulation, and consumer tolerance still conspire against automation. Rivr's wheeled quadrupeds are an intriguing answer to that problem, albeit one whose economics are entirely unproven at the scale Amazon requires. If the robots can genuinely boost a single driver's throughput by even 20%, the math gets interesting fast across billions of annual deliveries. If they cannot, Amazon will have bought itself another expensive science project — and the last mile will remain, for now, a human affair. ■
For more, join 75,000 subscribers getting tech's favorite brief here
