Agility Robotics announced in July 2026 that Digit units operating in Amazon fulfilment centres have now accumulated 400,000 operational hours across eight sites, up from the initial two-site pilot announced in 2024. The uptime figure — 98.7 percent across that operational window — is the key data point for anyone evaluating humanoid robotics in logistics.
Digit's task in Amazon facilities is tote handling: picking up plastic tote containers from conveyor systems and transferring them to induction stations, or retrieving empty totes from storage areas for refilling. This is a physically demanding, repetitive task that causes high injury rates among human warehouse workers. The robot's bipedal form factor is not required for the task — a conventional fixed arm could do it — but the humanoid design allows Digit to navigate the same aisles and work alongside the same conveyor systems designed for human workers, without facility modification.
The economic model that is emerging: Digit is offered as Robots-as-a-Service at an undisclosed hourly rate, structured to compete with the fully-loaded cost of a warehouse worker including benefits, turnover, and injury costs. Industry analysts tracking the sector estimate the break-even deployment duration at approximately 18 months at a rate competitive with US warehouse labour markets.
Amazon's scale creates a data advantage that compounds. Each Digit unit feeds operational telemetry back to Agility. Across 400,000 hours, that is more operational data than any other humanoid deployment in existence. Edge cases — unusual tote configurations, maintenance interventions, environmental variations — that appear rarely in a single facility occur daily across eight. The training and software improvement cycle that results is structurally faster than what smaller pilot deployments can achieve.
The limitation to be direct about: tote handling is an unusually robot-friendly task. Standardised objects, predictable environments, and a single defined motion profile. The step to unstructured picking — handling arbitrary products without standardised containers — is a qualitatively harder problem that neither Digit nor any other deployed system has solved at production quality levels.