Google Cloud and Nvidia are working with Munich-based AI startup Microagi to speed the development of AI robots designed to understand and interact with their surroundings.

Microagi, founded in 2025, builds hardware-agnostic AI models customized for specific robotic tasks in commercial and industrial settings. Its Atlas platform fine-tunes models using customers’ operational data alongside Microagi’s proprietary data platform.

The company said its technology is already in use by robotics companies, including Unitree and UBTECH.

Through the partnership, Microagi will receive access to Nvidia RTX PRO 6000 Blackwell Server Edition GPUs via Google Cloud’s G4 virtual machines. It will also use Nvidia GB300 NVL72 rack-scale systems through A4X Max instances.

The computing infrastructure is intended to support the training and deployment of models that can process multimodal data, such as information from cameras, sensors and other robot inputs. That capability is central to embodied AI systems, where software must interpret physical environments and translate model outputs into actions.

Microagi said the collaboration will help it build customizable software packages for enterprise robotics. For example, companies in hospitality or industrial sectors could deploy robots equipped with AI models trained for specific operational roles, rather than relying on a general-purpose system.

“Building the next generation of embodied AI requires intensive compute, a comprehensive stack of AI technologies, and deep engineering expertise,” Bercan Kilic, founder of Microagi, said in a release.

“Google Cloud stood out as the partner capable of supporting this massive model inference pipeline,” Kilic added. “Together with Nvidia's hardware, it provides the foundation as we scale our business offerings.”

Microagi will also use Google Cloud’s AI stack, including Gemini models and the Gemini Enterprise Agent Platform, to process multimodal information and expand its applications for enterprise customers globally.

The companies said their engineering teams are collaborating to optimize Microagi’s model training and inference pipelines, with the aim of helping the startup bring new robotics products to customers more quickly. For enterprise customers, the development highlights how robotics vendors are increasingly pairing specialized AI models with cloud-based accelerated computing to handle the data and deployment requirements of physical-world automation.

“Robotics is becoming one of the most demanding frontiers for AI, requiring massive physical-world datasets, accelerated compute and a full-stack platform to turn models into intelligent machines,” Tobias Halloran, director of EMEAI startups at Nvidia, said in a statement.