Hyperscale Data, Inc. (NYSE American: GPUS) has announced that its wholly owned subsidiary, Omnipresent Robotics, has begun installing OPR-R2 robots at the company's Michigan AI data center campus. The first unit was assembled on-site on July 16, 2026, marking the launch of a large-scale visual data collection and physical AI training program.
A Fleet of Data-Generating Robots
Omnipresent Robotics has placed orders for 143 OPR-R2 robots, with up to 100 units currently planned for installation at the Michigan facility. The remaining units are expected to be allocated to partner pilot programs, with more details to be announced in the coming months.
The company is developing an approximately 100,000-square-foot dedicated robotics facility within its Michigan data center, designed to support robot deployment, physical task execution, visual training data collection and processing, and AI development. The expansion could support more than 500 jobs over the next three years.
Addressing Physical AI's Data Bottleneck
The program targets one of the principal challenges facing physical and embodied AI: the availability of high-quality, real-world data demonstrating how machines perceive, navigate, and interact with physical environments.
Rather than being deployed primarily to evaluate the robots themselves, the OPR-R2 fleet serves as a real-world data-generation platform. Robots perform practical tasks in operating environments while capturing the visual, spatial, movement and interaction data needed to train AI systems to better understand and navigate the physical world.
US-Based Data Pipeline
All data generated through the program will be processed, warehoused, and maintained in the United States, with no data intended to leave the country—a significant consideration for enterprise and government customers with data sovereignty requirements.
A Continuous Development Cycle
Hyperscale Data envisions a continuous physical AI development cycle: robots generate real-world data, the data is processed and validated, AI models are trained and improved, and those improvements are incorporated back into future robotic deployments.
"The arrival of the first OPR-R2 robot in Michigan represents the beginning of an important real-world data-generation program. These robots will perform physical tasks and generate the visual, spatial and interaction data required to train increasingly advanced AI models."
