Amazon Bought the Robot That Climbs Stairs to Your Door

The Last 100 Yards: Why Amazon Needs Rivr

Amazon delivers roughly 20 million packages per day across the United States alone. The company has spent over a decade automating its warehouses, deploying more than 1 million robots (the Kiva-derived fleet acquired in 2012) across fulfillment centers worldwide. But once a package leaves the van, the delivery associate carries it by hand. That final stretch, from curb to doorstep, remains stubbornly manual.

This is where Rivr fits in. The company builds four-legged robots with wheels at each foot, a design that CEO Marko Bjelonic has described as a "dog on roller skates." On flat sidewalks, the wheels spin freely for speed and energy efficiency. When the robot encounters stairs, a curb, or a garden gate, it locks its wheels and transitions to a walking gait, using its articulated legs to step over obstacles. The cargo sits in a locked bin mounted on top of the robot.

Amazon said the robots would work alongside delivery associates, not replace them. The stated goal is to improve safety outcomes for human workers and customer experience by letting the robot handle the physical carry while the associate manages the route. That framing is deliberate: Amazon has faced years of scrutiny over injury rates at its facilities, and framing Rivr as a safety tool rather than a labor replacement is a calculated messaging choice.

Under the Hood: How Rivr's Hybrid Locomotion Works

Rivr's robots trace their technical DNA to ANYmal, the quadruped research platform developed at ETH Zurich's Robotic Systems Lab. The key innovation is the wheel-leg hybrid: each of the robot's four legs terminates in a powered wheel, giving it two distinct movement modes. On pavement, the wheels provide efficient rolling locomotion. On stairs or rough terrain, the wheels lock into position and function as rubber-treaded feet.

AI-generated image

Hybrid wheeled-legged locomotion systems use powered wheels at each joint terminus. Source: Editorial illustration

The transition between rolling and walking happens autonomously. The onboard perception stack, which includes LiDAR, multi-directional cameras, and IMUs, builds a real-time 3D map of the terrain ahead. When it detects an obstacle like a staircase, the control system shifts the robot's gait within a fraction of a second. The entire perception-to-actuation pipeline runs on an onboard GPU, with no cloud dependency for navigation decisions.

The RIVR Two, the latest production model, carries up to 30 kg (66 pounds) in its top-mounted cargo bin and travels at speeds up to 15 km/h (9.3 mph). Battery life runs roughly 3 hours per charge, with a 2-3 hour recharge cycle. The robot covers over 30 km per charge on flat ground, though stair climbing reduces that range. Each leg uses electric actuators with integrated force sensing, allowing the robot to detect and adapt to surface changes in real time.

Who's Behind Rivr

Rivr was founded by Marko Bjelonic and a team of researchers from ETH Zurich, originally operating under the name Swiss-Mile. The company's early work focused on demonstrating that wheeled-legged quadrupeds could handle real urban environments, not just lab floors. Before the acquisition, Rivr had raised $25 million total, including a $22.2 million seed round in 2024 that valued the company at roughly $100 million.

The investor list for that seed round reads like a preview of the acquisition: Amazon's Industrial Innovation Fund and Bezos Expeditions (Jeff Bezos's personal investment vehicle) both participated. HongShan, the China-focused venture firm formerly known as Sequoia China, also invested. Amazon essentially scouted, funded, and then bought the company within a two-year window.

Before the acquisition closed, Rivr had been running a pilot program in Austin, Texas, with Veho, a package delivery company. The pilot tested the robots on actual delivery routes, loading them from vans and sending them to residential doorsteps. Bjelonic had set a target of scaling to 100 robots by 2026, though it's unclear whether that milestone was reached before Amazon's offer landed.

What This Means for Last-Mile Delivery

AI-generated image

Amazon operates hundreds of delivery stations across the U.S., each dispatching thousands of packages daily. Source: Editorial illustration

Amazon's acquisition of Rivr is notable for what it says about the company's robotics strategy. After killing Scout in October 2022, calling it a failure to find a viable path to scale, Amazon has been conspicuously absent from the autonomous delivery space. Competitors like Nuro (road vehicles) and Serve Robotics (sidewalk bots) continued to push forward, while Amazon focused its robotics budget on warehouse automation.

Rivr represents a different approach than Scout. Where Scout was a cooler-sized box on six wheels that could only operate on flat sidewalks, Rivr's robots can climb stairs, open gates, and traverse the kind of terrain that characterizes most American neighborhoods. That's a meaningful distinction: according to U.S. Census data, roughly 36% of American housing units are in multi-story structures where stairs are unavoidable.

The economics are worth examining too. Amazon spends an estimated $90 billion annually on shipping and fulfillment. Last-mile delivery, the journey from the nearest distribution hub to the customer's door, accounts for roughly 53% of that cost according to Capgemini research. Any technology that reduces per-delivery cost in that segment, even by a few percentage points, translates to billions in annual savings at Amazon's scale.

The broader delivery robotics market is watching closely. FedEx tested its Roxo robot before pausing the program. Walmart has partnered with Serve Robotics for limited food delivery. But none of these programs have demonstrated a robot that can handle the full door-to-door challenge. If Amazon can scale Rivr's technology across its delivery network, it would be the first major logistics company to deploy a robot capable of genuine doorstep delivery at volume.

Frequently Asked Questions

What's Coming Next

Amazon's playbook with robotics acquisitions is well established. It bought Kiva Systems for $775 million in 2012, rebranded it as Amazon Robotics, and scaled from a few hundred units to over 1 million warehouse robots. Rivr is a much smaller acquisition, but the trajectory could follow a similar pattern: integrate the technology, prove it in controlled tests, then scale manufacturing once the economics work.

The near-term milestones to watch: Amazon field tests in at least one U.S. metro area (likely Austin, where Rivr already has operational data), public statements on per-delivery cost reduction targets, and any expansion of the Rivr engineering team. Amazon's job postings for robotics roles in its Last Mile division will be a reliable indicator of how seriously the company is investing in scaling this technology. If Rivr's headcount doubles within six months, the signal is clear.

For now, the acquisition is a signal more than a product. But Amazon has a track record of turning small robotics acquisitions into industry-defining infrastructure. The delivery robots on your doorstep may still be a few years away, but the engineering to get them there just got a $2 trillion company behind it.