Diligent says Moxi 2.0’s AI represents one of the largest datasets of human-robot interaction. | Source: Diligent Robotics
Diligent Robotics, which deploys mobile manipulation robots in hospitals, today unveiled plans for Moxi 2.0, the latest generation of its platform. The company said the launch builds on three years of real-world data, collected from over 1.25 million deliveries in hospital environments.
“Interactive environments are the hardest to simulate, but they’re where robots can deliver the most value,” said Andrea Thomaz, co-founder and CEO of Diligent Robotics. “We’ve collected millions of examples of Moxi operating in dynamic human environments, now built into Moxi 2.0, the first system to truly reflect that lived experience and everything we have learned.”
Representing one of the largest deployed fleets of NVIDIA-powered mobile manipulators in healthcare, Moxi currently operates in over 25 hospitals across the U.S. The robot helps nurses and pharmacy staffers with routine tasks such as delivering medications and lab samples.
While Moxi 2.0 hasn’t been deployed in hospitals yet, Diligent plans to send out the first upgraded robots to hospitals in the first half of 2026. The company is also planning to expand into the senior living sector with its participation in the AgeTech Collaborative from the AARP accelerator program.
“The hardware is going through rigorous testing as we approach manufacturing. In parallel, we continue to iterate the build models to run inside the robot once the hardware is ready,” Vivian Chu, co-founder and chief innovation officer at Diligent Robotics, told The Robot Report. “While many of the hardware elements are complete, we are just starting to implement the new software inside the robot. Once we are through testing, we will begin sending into the hospitals in the first half of the year.”
Founded in 2017, Diligent Robotics has created the Moxi mobile manipulator to improve healthcare workflow efficiency. The Austin, Texas-based company has deployed the robot in more than 25 hospitals across the U.S. These robots help nurses with routine tasks to free them up for patient care and prevent burnout.
Diligent Robotics lists Moxi 2.0 upgrades
On the left, the current Moxi robot, and on the right, the new Moxi 2.0. | Source: Diligent Robotics
With Moxi 2.0, Diligent Robotics is introducing upgraded intelligence and hardware, designed to handle the complexity of crowded, dynamic indoor human environments. The new system pairs next-generation AI compute with the company’s proprietary AI stack, enabling Moxi to reason, predict, and adapt for a wider range of capabilities in real-time, including pre-emptively navigating around beds and wheelchairs.
New features in the robot include:
- Next-gen AI compute: Powered by NVIDIA Thor, Moxi 2.0 is equipped with increased compute, which Diligent said allows it to deploy significantly more powerful and capable AI models and faster inference speeds.
- Next-gen hardware platform: The new design is optimized for manufacturability and hardened to support the fleet’s quickly growing scale.
- Robot foundation model: Diligent is developing an “end-to-end” action model optimized for enhanced dense navigation behaviors and high-precision, complex manipulation. Using its growing dataset of deployment data—expected to grow to petabytes next year with Moxi 2.0—the company said it will release more powerful foundation models to power the next generation of robotic applications.
- Physical user interaction points: Deployment has provided Diligent detailed user feedback on how to service and physically interact with its platform. Moxi 2.0 has improved handles and servicing panels for easier user navigation.
“Moxi 2.0 is built to do more than just deliver items. In senior living communities, we want Moxi to be part of the daily rhythm of life,” said Chu, who spoke at RoboBusiness 2025. “The goal is for Moxi to connect with residents in small, meaningful ways, sharing greetings in the hallway, remembering names, and eventually having real conversations.”
“Years of working in hospitals have taught us that people naturally treat Moxi like part of the team,” she noted. “That same spark of interaction can bring joy and comfort to residents while helping staff stay focused on care. Moxi 2.0 is about combining useful help with genuine human connection.”
Moxi addresses robotics’ chicken-and-egg data problem
Most robotics platforms face a fundamental challenge: The models need deployment to improve, but the models must improve before wide deployment.
Diligent Robotics said it has solved this problem by scaling its MVP humanoid platform early, allowing it to collect massive volumes of high-value data from the real world. A growing fleet feeds the data “flywheel” by capturing millions of edge cases from dynamic environments, fueling faster learning of models to handle them, according to the company.
Today, Moxi already excels at mobile manipulation when faced with human interactions in shared spaces, as well as variations and predictive inference in busy environments, claimed Diligent. Moxi 2.0 will continue building on this foundation, it said
Moxi 2.0’s platform improvements are designed to support hospital rollouts of more than 15 units per site as more advanced capabilities come online. As a result, Diligent said it expects to double its hospital footprint annually and deploy thousands of robots by 2030.
In late 2024, the company added a layer of development oversight with its AI Advisory Board. Since then, Diligent has added more experts to its team.
“[The AI Advisory Board] has helped us iterate and provide feedback on the architecture of Moxi’s AI models,” Chu said. “They also help continue to keep the team up to date on the latest state-of-the-art developments so that we can always be looking at what is coming and stay ahead of technology breakthroughs.”
Diligent builds on new NVIDIA products
The NVIDIA IGX Thor. | Source: NVIDIA
Diligent Robotics is an early adopter of NVIDIA‘s IGX Thor. It is a NVIDIA Blackwell-powered, industrial-grade, enterprise-ready product. NVIDIA designed the new platform for industrial, robotics, and medical edge AI applications and unveiled it at GTC DC today.
“Moxi 2.0 has 10x the compute of Moxi 1.0, thanks to the rapid advancement of the NVIDIA Jetson line,” Thomaz said. “We’re excited about how the new NVIDIA IGX Thor platform pushes the boundaries of AI performance at the edge, unlocking possibilities for future generations of intelligent, human-centered robots.”
The platform features real-time sensor processing, AI reasoning, functional safety, and long-term enterprise support. With eight times the compute power of its predecessor, IGX Thor enables developers to build intelligent systems that perceive, reason, and act faster, safer, and smarter, said NVIDIA. Other early adopters include EndoQuest Robotics, Hitachi Rail, Joby Aviation, Maven, and the SETI Institute.
The platform features two types of NVIDIA Blackwell graphics processing units — an integrated GPU (iGPU) and a discrete GPU (dGPU). These deliver 5,581 FP4 teraflops of AI compute with 400 GbE connectivity.
Compared with NVIDIA IGX Orin, IGX Thor provides up to eight times higher AI compute on iGPUs, two and a half times higher AI compute on dGPUs, and two times better connectivity to seamlessly run large language models and vision language models at the edge.
The post Moxi 2.0 mobile manipulator is built for AI, says Diligent Robotics appeared first on The Robot Report.