Diligent Robotics has begun commercial shipments of Moxi 2.0, its hospital mobile manipulator, to Endeavor Health Edward Hospital, Providence Saint John’s Health Center and Children’s Hospital Los Angeles. The Robot Report covered the rollout on August 17, 2026. Diligent was acquired by Sub Robotics for $29 million in January 2026; Moxi 2.0 was first shown in October 2025.
The hardware step is large. Moxi 2.0 carries roughly ten times the onboard compute of its predecessor and perceives 10 to 15 times faster, on an NVIDIA A2000. It runs up to 18 hours a day, up to nine hours continuously, and charges about 30% faster.
The learning loop is the product claim
Diligent trains on five years of operating data from more than 25 US hospitals. The stack combines NVIDIA Isaac Sim with the Cosmos open world model, using a 3D lidar tokenizer, with training run on AWS SageMaker HyperPod. T-Mobile private 5G handles the connectivity.
That is a conventional list of vendors. What is not conventional is that the data source is a paid deployment rather than a research fleet. Children’s Hospital Los Angeles reports more than 40,000 deliveries and over 16,000 hours of staff time offset since it started, expanded from two robots to three, and saw utilisation rise about 10% in the second quarter.
Key Facts
- Commercial shipments announced August 17, 2026, to Endeavor Health Edward Hospital, Providence Saint John’s Health Center and Children’s Hospital Los Angeles
- Five years of operation across more than 25 US hospitals
- About 10x onboard compute and 10–15x faster perception, on NVIDIA A2000
- Runs up to 18 hours a day, up to 9 hours continuously; charges roughly 30% faster
- Children’s Hospital Los Angeles: 40,000+ deliveries, 16,000+ hours of staff time offset, fleet expanded 2 to 3 robots, utilisation up about 10% in Q2
- Stack: NVIDIA Isaac Sim plus the Cosmos world model with a 3D lidar tokenizer; training on AWS SageMaker HyperPod; T-Mobile private 5G
- Acquired by Sub Robotics for $29 million in January 2026
Why it matters
Almost every embodied AI company now claims a data flywheel. Very few can point to the hours. The distinction that matters is between fleets that generate data as a by-product of paid work and fleets that exist to generate data, because only the first kind survives a budget review.
The counterweight is scope. Moxi does supply delivery in corridors, not dexterous bedside manipulation, and a world model trained on hospital corridors is a world model of hospital corridors. Whether that generalises is unproven — Diligent has not published benchmarks against it. But it is a live commercial system running a modern perception stack, which puts it ahead of most of the category on the only axis that eventually decides these things.
Frequently Asked
What is Moxi 2.0?
Diligent Robotics’ hospital mobile manipulator, now shipping commercially to Endeavor Health Edward Hospital, Providence Saint John’s Health Center and Children’s Hospital Los Angeles. It carries about 10x the onboard compute of the previous generation.
What world model does Diligent use?
NVIDIA’s Cosmos open world model with a 3D lidar tokenizer, alongside Isaac Sim, with training run on AWS SageMaker HyperPod.
What results have hospitals reported?
Children’s Hospital Los Angeles reports more than 40,000 deliveries and over 16,000 hours of staff time offset, expanded its fleet from two robots to three, and saw utilisation rise about 10% in the second quarter.