A Research Lab That Needed a Business Model
Boston Dynamics was founded in 1992 as a spin-off from MIT's Leg Laboratory, with a focus on understanding and replicating animal motion in mechanical systems. For its first two decades, the company operated primarily on government contracts — most significantly from DARPA and the US Army. BigDog, the quadruped robot built to carry equipment over rough terrain, was the most visible product of that era: impressive as an engineering demonstration, difficult to commercialise, and never deployed operationally by the military at scale.
Google acquired Boston Dynamics in 2013, Softbank in 2017, and Hyundai in 2021. Each ownership transition brought questions about commercial viability that the company's research reputation couldn't easily answer. SPOT — the four-legged inspection robot — was the first product to generate consistent revenue, finding buyers in utilities, construction, mining, and public safety. But Spot, at roughly $75,000 per unit, served a relatively narrow market. Boston Dynamics needed a second product with a larger addressable use case. Warehouses offered one.
The Right Problem to Solve
Truck and container unloading is one of the most physically demanding jobs in logistics. Workers — referred to in the industry as "unloaders" — move boxes from shipping containers onto conveyor belts, typically in tight, hot spaces, handling hundreds of packages per hour at weights of up to 30 kg or more. Injury rates are high, turnover is high, and the labour pool for the job is structurally constrained. Automation had long been proposed but rarely achieved: the challenge is that packages arrive in no particular order, in varied sizes and orientations, stacked in ways that don't conform to a predictable pattern.
What made the problem tractable by the late 2010s was not a new mechanical breakthrough but better computer vision. Machine learning systems had become capable of identifying package types, estimating positions, and planning grasping strategies in real time — tasks that had previously required either very constrained environments or human eyes. Boston Dynamics began working on Stretch around 2018, in collaboration with DHL, to understand whether a robot arm mounted on a mobile base could handle unloading at the throughput rates that would make it economically viable.
The Kinema Acquisition
In April 2019, Boston Dynamics acquired Kinema Systems, a Bay Area startup focused on vision-based robotic picking. Kinema's Pick system used 3D sensors and neural networks to identify packages on a conveyor or in a container and direct a robot arm to grasp them — a perception and planning problem that was directly relevant to what Stretch would need to do. The acquisition gave Boston Dynamics a complete vision stack rather than having to build one from scratch, and Kinema's team became core contributors to Stretch's development.
This acquisition was characteristic of how Stretch was built: not as an exercise in theoretical robotics but as an engineering integration project. The arm, the mobile base, the vision system, and the control software each drew on existing technology adapted to a specific set of requirements. The goal was a robot that worked reliably in a real warehouse, not one that expanded the frontier of what robotics could theoretically do.
Designed Around One Task
When Boston Dynamics unveiled Stretch in March 2021, the design philosophy was explicit: this robot does one thing. The 7-DoF arm, the suction gripper, the compact omnidirectional base, the counterweight system — every element was chosen to optimise the truck-unloading use case rather than to maximise versatility. Boston Dynamics' own explanation was that a robot built to do one job well would be more useful and more commercially viable than a general-purpose system that did many things adequately.
That choice had trade-offs. Stretch cannot navigate stairs, carry people, or adapt to tasks outside its design envelope the way a more general platform might. But it also meant that the engineering team could make decisions — the specific reach envelope, the suction gripper, the payload limit — that wouldn't make sense in a general robot. The result was a system rated to 800 cases per hour under optimal conditions, in a package that could fit wherever a standard pallet can fit.
The DHL Partnership and Commercial Launch
DHL Supply Chain announced a $15 million investment in Boston Dynamics in January 2022 and became the first commercial customer for Stretch. The scale of the commitment was significant: DHL is one of the largest logistics companies in the world, operating thousands of warehouses across more than 220 countries. An endorsement from DHL served as validation for the broader logistics industry in a way that a smaller pilot customer could not.
The first commercial deployment came in Spring 2023, when DHL began operating Stretch in live warehouses for truck unloading. By 2024, DHL had committed to deploying more than 1,000 units by 2030 — a target that, if met, would make DHL one of the most significant deployments of industrial mobile manipulation in logistics history. Otto Group separately announced plans for more than 20 facilities, providing a second major anchor customer in the European market.
The 2025 Upgrade and Where Stretch Stands
In January 2025, Boston Dynamics released an upgraded version of Stretch with improved throughput and a reported 40% reduction in handling errors compared to earlier deployments. The upgrade reflected a pattern common in commercial robotics: the initial product proves the concept, live deployment reveals the friction points, and the next generation addresses them. Error reduction at scale matters enormously in logistics, where a mistake at the unloading stage can cause knock-on disruptions through the downstream fulfilment process.
Stretch occupies a specific and deliberately narrow position in the robotics market. It is not trying to compete with Spot on mobility or with Atlas on general capability. It is a commercial product built around a task that exists in thousands of warehouses globally, with throughput requirements that make the economics of automation plausible. Whether Stretch maintains that position as other companies develop competing warehouse robotics solutions will depend on how quickly the cost, reliability, and breadth of packages it can handle continue to improve.