Humanoid robots get the reveal videos and the headlines. Warehouse robotics gets neither, and it's arguably the more commercially significant robotics story of the past several years. Agility Robotics' Digit (a bipedal robot purpose-built for logistics tasks) is, as of 2025, the only bipedal robot currently generating revenue from paying commercial customers, and the numbers behind that claim are worth examining directly.
What 18 months of real deployment actually produced
After 18 months of testing at Amazon's Sumner, Washington fulfillment facility, Digit robots achieved a 98% task success rate while operating at an estimated $10-12 per hour, compared to roughly $30 per hour for equivalent human labor. Digit has moved over 100,000 totes in commercial deployment, a scale that reflects genuine, sustained production use rather than a pilot still finding its footing.
Beyond Amazon: a genuinely multi-customer deployment
Agility announced in February 2025 that it would deploy multiple Digit units at Toyota's manufacturing plant in Canada, handling tote loading and unloading, joining earlier deployments at Amazon, Schaeffler Group, and logistics provider GXO. That multi-customer spread matters: a single large customer's pilot is one thing, but Digit moving into production roles across four independent operators is a much stronger signal that the underlying capability generalizes beyond one company's specific facility layout and workflow.
What changed at ProMat 2025
At the ProMat 2025 industry trade show, Agility introduced upgrades aimed specifically at scaled industrial use, including extending battery life to around four hours per charge, directly addressing one of the most practical constraints on continuous warehouse deployment. Digit has also demonstrated genuine multi-tasking beyond its original narrow role: not just picking items on and off an autonomous mobile robot to a conveyor, but stacking totes onto a different floor location entirely, real evidence of generalization rather than a single hard-coded routine.
Why warehouses were the right environment first
Our explainer on embodied AI covers why unstructured environments remain the hard case for physical AI. A well-run modern warehouse is close to the opposite: mapped floor plans, standardized shelving, controlled lighting, no unpredictable pedestrian traffic, exactly the structured conditions current-generation robotics AI handles well, which explains why Digit's real commercial traction has run well ahead of the more general-purpose humanoid robotics deployments covered in our companion piece on Figure AI's BMW pilot.
The AI layer that actually makes this work
The harder problem underneath Digit's numbers isn't the physical hardware. It's the fleet coordination and routing layer, continuously solving which robot should fetch which item along which path while avoiding collisions with dozens of other robots and human workers in the same space. That's closer to a large-scale logistics optimization problem than to a pure walking-and-grasping problem, and it's a domain where AI-driven optimization has matured faster than more general physical-world perception and manipulation.
Why this deployment success is real but narrow
It's tempting to read Digit's numbers as evidence that general physical-world AI has arrived. The more accurate read is narrower and, in its own way, more impressive: this specific combination of a structured environment, a bounded task set, and human-robot collaboration playing to each party's strengths has produced Agility's genuine revenue-generating deployment, and that combination doesn't automatically transfer to less structured settings, which is exactly why general-purpose humanoid deployment elsewhere remains earlier-stage even as Digit's warehouse numbers keep climbing.