Munich Startup microagi Taps Google Cloud and Nvidia Blackwell to Train Europe's Industrial Robots

Fresh off closing what it describes as the largest seed round in German history, Munich-based microagi has locked in a compute partnership with Google Cloud and Nvidia to scale its robotics data platform, Atlas. The deal, announced roughly a week after the funding news, gives the young startup access to Nvidia's newest silicon running through Google's cloud infrastructure, a combination microagi says is already changing the economics of training robots for factory floors.

Atlas is built to handle the full pipeline industrial customers need: pulling structured data out of factory environments, training robotics models on that data, and pushing the resulting models into live production deployments. Rather than tying itself to a single robot manufacturer, microagi has positioned Atlas as hardware-agnostic, and its models are already running on hardware from Unitree and UBTECH, two of the more prominent names in commercial and humanoid robotics.

The technical backbone of the partnership centers on Nvidia's Blackwell generation. Atlas trains and serves its models on GB300 NVL72 rack-scale systems, A4X Max instances, RTX PRO 6000 Blackwell GPUs, and G4 virtual machines, all provisioned through Google Cloud rather than on-premises infrastructure. On the software side, microagi will lean on Google's Gemini Enterprise Agent Platform and its multimodal models to process the video and sensor data that industrial robotics depends on, an area where handling high-volume visual input efficiently is often the bottleneck.

Founder and CEO Bercan Kilic, who started the company in 2025, frames the arrangement as a way to compress the time between collecting factory data and deploying working robotics behavior at scale. Nvidia's Tobias Halloran, director of EMEAI startups, is also cited in connection with the deal, reflecting the chipmaker's broader push to seed European robotics ventures with access to its top-tier hardware rather than leaving that infrastructure concentrated among a handful of US hyperscaler-backed labs.

One concrete metric microagi has pointed to is efficiency: the company says it has roughly doubled the amount of useful computational work it extracts per unit of energy since the collaboration with Google Cloud and Nvidia began. For a category where training costs and power draw are persistent constraints, that kind of gain, if it holds up as Atlas scales to more customers, would matter more than raw compute headlines.

The timing is notable on its own. Coming so soon after its seed round, the infrastructure deal signals microagi is trying to move quickly from funding announcement to demonstrable technical capacity, a sequencing that suggests the startup, and its investors, see the compute partnership as validation that the capital is being put to use immediately rather than banked for later.