Hyperscale Data bets on robotics in Michigan

Company Hyperscale Data, Inc. (NYSE American: GPUS), which operates data centers focused on artificial intelligence, announced a significant expansion into the world of robotics. Its wholly-owned subsidiary Omnipresent Robotics concluded an agreement with Singaporean-Chinese manufacturer AgiBot for the acquisition of up to 143 intelligent robotic units. The contract also includes the right to resell these robots under its own Omnipresent brand and to build a specialized robotics data collection center at the Michigan data center site.

According to the official press release and confirmation from MarketWatch, the new operation will occupy approximately 100,000 square feet (almost 9,300 square meters) and its task will be generating robotics datasets, remote control (teleoperation), and training AI models directly in robotic bodies. The new agreement builds on a previous partnership contract that both companies signed on April 15, 2026.

Who is AgiBot and why its expansion is important

AgiBot is among the fastest-growing Chinese manufacturers of humanoid and service robots. In its home market, it competes with giants such as Unitree or UBTech and has significantly accelerated its international expansion in recent months. In April 2026, the company announced the launch of humanoid robot rentals in 14 countries worldwide and is now penetrating the American market through a strategic partnership with Hyperscale Data.

For Czech and European readers, this development is interesting for several reasons. AgiBot previously announced plans for expansion in Europe, and a similar partnership could emerge on the old continent in the future. Czech companies and startups in the field of robotics are thus receiving a clear signal: Chinese technologies in the area of embodied AI are already mature enough for American companies to integrate them into their critical infrastructure. European companies that have so far hesitated to deploy humanoid robots should watch this trend carefully.

What is embodied AI and why it needs its own data centers

Embodied AI, or "embodied artificial intelligence," refers to AI systems that do not run only on servers in the cloud, but directly in the bodies of robots. These systems must learn physical interaction with the environment — manipulating objects, recognizing obstacles, balancing, or walking on uneven surfaces. The key factor is the quality and volume of training data, which must contain connections between visual perceptions, language instructions, and physical actions.

That is precisely why Hyperscale Data is building a specialized center in Michigan. In it, robots under the supervision of human operators will collect so-called VLA data (Vision-Language-Action) — records of how the robot sees the world, what instructions it receives, and what actions it performs. This data is subsequently used for training and improving autonomous models. The teleoperation bay, i.e., the workstation for remote control, will allow human operators to control robots in real time and thus generate valuable dataset records.

Does this mean an advantage for Americans, or for China?

The agreement between Hyperscale Data and AgiBot is symbiotic from a business perspective. Hyperscale Data gains immediate access to an advanced robotic hardware platform, while AgiBot penetrates the American market without having to build its own sales and service network. For the American side, however, this step is also politically sensitive: it involves the transfer of Chinese robotic technology into critical American infrastructure, at a time when the United States is tightening control over investments in Chinese technologies in the AI area.

From a European perspective, it is interesting that while the USA is actively building embodied AI centers with Chinese hardware, Europe remains behind in this area. No Czech or European data center has yet announced an similarly ambitious project combining robotics, teleoperation, and VLA data collection on an industrial scale. Yet Europe could benefit from developing its own standards for safety and ethics of embodied AI — an area that the EU AI Act is beginning to regulate, but where practical infrastructure is still lacking.

What awaits robots in Michigan

According to available information, the Michigan center will serve several purposes at once. Primarily, it will be for collection and processing of robotics data for training AI models. Furthermore, remote control scenarios will be tested here, which can find application in logistics, industrial manufacturing, and services. And last but not least, the center will function as a training environment for local workers, because operating 143 robots requires qualified operators, technicians, and data analysts.

Hyperscale Data explicitly states that it wants to make Michigan a key embodied AI hub for the United States. If this plan succeeds, it could become a model for similar centers around the world — including Europe. Czech companies operating in the field of industrial automation should therefore watch how the American market responds to the combination of Chinese hardware and its own data centers. An opportunity for a European partnership with AgiBot or similar manufacturers may soon arise.