Las Vegas-based Hyperscale Data Inc. has announced it is accelerating the enhancement of its Michigan operations in Dowagiac (southwest of Kalamazoo) into a combined AI data center and robotics hub.
The announcement follows Hyperscale’s recently executed agreement with AGIBOT, a developer of intelligent robotics technology.
Hyperscale is advancing the reconfiguration of key sections of its existing building on its 34.5-acre campus to support AI infrastructure, robotics deployment, and large-scale data generation.
The company currently operates approximately 30 megawatts (MW) of power capacity at the site, and it believes there is potential to expand to over 300 MW over time.
Hyperscale Data is enhancing its Michigan operations to combine high-performance computing infrastructure with robotics capabilities. Sections of the existing building are being re-imagined to support robotics assembly, testing, and deployment alongside AI model training and validation.
As part of the initiative, the company is initially dedicating more than 100,000 square feet within its existing 617,000 square foot facility to AI and robotics operations. Within this footprint, the company plans to:
- Develop robotics assembly and testing capabilities;
- Build real-world environments for data collection and system validation; and
- Integrate compute infrastructure with robotics-driven data generation.
Data generated through the activities is expected to be commercialized and utilized within U.S. markets, supporting domestic AI development and deployment.
“We have seen this movie before. America invents something big. Then the most important aspect is lost and recaptured elsewhere,” says Milton “Todd” Ault III, executive chairman of Hyperscale Data. “The American factory is shut down. The supply chain disappears. The practical know-how goes with it. At Hyperscale Data, we are not interested in repeating that story with respect to robotics.
“We are building in Michigan because Michigan still has industrial muscle memory. We are developing a campus where robots can be trained to perform real jobs, tested until the weak points show themselves, assembled here, and sent back into the field with software that actually fits the work. That will create engineers, technicians, operators, manufacturing jobs, and a place where people learn by doing instead of talking.”
Hyperscale believes the next phase of AI will increasingly depend on real-world data and physical system training, in addition to traditional digital datasets.
The Michigan campus is being developed to support:
- Machine-generated data from robotics operating in real environments;
- Human (egocentric) data capture for contextual learning;
- Testing and validation of robotic systems; and
- Training workflows for advanced AI models.
Hyperscale Data expects to hire more than 500 employees over the next three years to support its Michigan operations. Anticipated roles include robotics engineers, AI data specialists, infrastructure personnel, and operations staff.
The company expects to pursue opportunities to support frontier AI developers — those advancing large-scale models and next-generation robotics systems — and high-performance computing platforms as demand for these capabilities evolves.
“Some customers want the robot and the software development kit,” says Ault. “Others want a model trained on our compute for their floor, their tools, and the way their team actually works. Some want the whole system deployed and supported. All of that is the same business to us, because each job enhances our ability to improve the next project and make it more productive.
“The data matter, too. We plan to sell embodied AI data, evaluation data, and scenario data to frontier labs. Better models are going to need contact with the physical world. A vision-language-action model cannot learn torque, friction, occlusion, or recovery from a corpus of words and pictures. So, we intend to collect the kind of data that comes from real work, real failures, and repeated attempts to get the machine to do the job right. Additionally, we intend to pursue every effort to ensure that all of that data never leave the U.S.
“What we are building is a loop: Supply chain, computing, training, evaluation, assembly, deployment, data creation, then back into training again. Miss one part, and you end up dependent on somebody else for the part that decides whether the system actually works.
“We want that loop in America. We want a serious piece of it in Michigan. We want to leave behind capability, jobs, and machines that can hold up in the real world.”
Through its wholly owned subsidiary Sentinum Inc., Hyperscale Data owns and operates a data center in Michigan at which it mines digital assets and offers colocation and hosting services for the emerging AI ecosystems and other industries.
Hyperscale Data’s other wholly owned subsidiary, Ault Capital Group Inc., is a diversified holding company pursuing growth by acquiring undervalued businesses and disruptive technologies with a global impact.
For more information, visit hyperscaledata.com.
 into a combined AI data center and robotics hub.){kind=link}