Hyperscale Data says production has started on the first 30 OPR-R2 humanoid robots from its wholly owned subsidiary Omnipresent Robotics.
The company expects to begin receiving components and assembling the robots for deployment at its Michigan AI data center campus in the third quarter of 2026. The first batch is part of a larger plan to deploy 143 OPR-R2 humanoid robots across the campus.
That is an unusual twist in the humanoid race. These robots are not being pitched first as home helpers, factory workers, or warehouse pickers. Hyperscale Data wants to put them inside an AI data center environment, where they can support research, training, testing, facility operations, and real-world data collection.
The first 30 robots will be assigned to Omnipresent Robotics’ Model Training Laboratory. According to the company, they will work alongside AI infrastructure personnel and data center employees while assisting with data collection, model training, simulation validation, facility operations, and development of next-generation embodied AI systems.
Hyperscale Data is trying to turn its Michigan campus into a physical AI training ground.
The company says the lab is being designed to support foundation models and frontier AI systems, including large language models, vision-language-action models, robotics foundation models, and other forms of physical AI. The idea is to combine humanoid robots, high-performance AI computing, simulation environments, and real-world data in one place.
That combination is becoming a common theme in robotics. AI companies need physical data. Robot companies need better models. Data centers have the computing power to train those systems, but they usually do not produce the kind of real-world robot interaction data needed for embodied AI. Hyperscale Data is trying to connect those pieces under one roof.
The planned Robotics Research, Testing and Innovation Center at the Michigan campus is expected to cover 100,000 square feet. The OPR-R2 robots are expected to operate from that center and support development of autonomous workflows, advanced robotics systems, and commercial applications.
The company also says the lab is expected to use NVIDIA-based infrastructure for simulation, training, inference, and robotics workloads. That gives the project another link to the broader physical AI ecosystem, where NVIDIA tools and hardware have become a major part of robot training pipelines.
For now, this is still the beginning of production, not a full deployment. The robots still have to be assembled, delivered, tested, and put to work in the environment Hyperscale Data is describing.
But the direction is clear. The company is betting that future AI systems will not only be trained on text, images, and video. They will also need robots moving through real spaces, collecting data, testing actions, and learning how the physical world behaves.
Join the Discussion