Key Takeaways:
Google and Nvidia are supplying computing power to German startup Microagi, betting humanoid robots will require massive cloud infrastructure to train and deploy.
Google and Nvidia are partnering with German data-robotics startup Microagi to provide cloud computing and AI models for humanoid factory robots, the companies said Wednesday, as Big Tech races to capture the infrastructure layer of physical AI.
"The global push into robotics is going to require massive amounts of compute," Marianne Janik, Google Cloud's vice president for northern Europe, said in an interview. Google isn't receiving data from Microagi or its customers in return, she added.
The deal gives Microagi, which raised a $55 million seed round, access to Nvidia GPUs and Google's AI models and cloud platforms. The startup collects real-world training footage across more than a dozen countries and consults with manufacturers on deploying robots in their factories. Google DeepMind already partners with robot hardware makers including Apptronik, Boston Dynamics and Agile Robots, but the Microagi deal shows Google's push to court European startups through its cloud business.
The partnership comes as Chinese tech giants like Alibaba expand their European cloud footprint and several local robotics firms have come under Chinese investment. "Of course there's going to be competition," Janik said. "We can see it, we can sense it."
Europe's Humanoid Race Intensifies
Beyond Microagi, a cluster of European humanoid startups is racing to deploy robots in factories with far less capital than US rivals. London-based Humanoid AI, founded in 2024, has signed deployment deals with Schaeffler, Bosch and Siemens while operating with $74 million in funding — compared with the billion-dollar war chests of Figure and Apptronik in the US.
Humanoid AI's chief product officer Sotirios Stasinopoulos, who spent over a decade in China at UBtech, said the company's wheeled HMND 01 robot can lift 15 kilograms and perform core tasks at about 80% of human speed. The company uses Nvidia GPUs for compute and Google's Gemini vision-language model in its four-layer AI architecture, which breaks factory tasks into deterministic workflows.
"The goal is fleets, not units," Stasinopoulos said. "If you only want a robot to do one thing, you might as well get an old-school automation solution."
The Compute Bottleneck
Training and deploying humanoid robots at scale requires enormous computing resources. Nvidia's H100 and Blackwell GPUs are the primary hardware used for robot training, while cloud platforms from Google, Amazon Web Services and Microsoft Azure compete to host inference workloads. TSMC, which manufactures Nvidia's chips, controls 73% of the global chip foundry market and is investing $265 billion in US fabrication plants, according to Counterpoint Research.
For startups like Microagi and Humanoid AI, access to subsidized compute through partnerships with Google and Nvidia can reduce training costs by millions of dollars — a critical advantage when competing against well-funded US and Chinese rivals.
Nvidia shares declined ahead of Alphabet's earnings report, reflecting market caution about Big Tech capital spending. The Microagi partnership is small relative to Nvidia's data center business, but it positions Google and Nvidia to capture the infrastructure layer of physical AI as humanoid robots move from prototypes to factory floors. European manufacturers adopting AI-powered robotics will need cloud compute, and Google's early partnerships with Microagi and Humanoid AI give it an edge over Amazon and Microsoft in the region.
This article is for informational purposes only and does not constitute investment advice.
