Google Cloud has signed a significant deal with Fei-Fei Li’s AI startup, World Labs, to power the compute needs required for training their cutting-edge AI models. Despite Li’s past role as Google Cloud’s AI chief, the agreement resulted from Google’s advanced infrastructure and ability to scale AI workloads.

World Labs, which recently raised $230 million at a valuation exceeding $1 billion, will utilize Google’s GPU servers to develop “spatial intelligence” AI models that process geospatial and video data.

This deal places World Labs alongside elite AI startups like OpenAI and Anthropic, which rely heavily on cloud providers for training their compute-intensive AI models. Google’s AI chips and robust infrastructure played a major role in securing the partnership. However, World Labs opted to train its models on Nvidia’s GPUs instead of Google’s proprietary tensor processing units (TPUs).

James Lee, general manager of startups and AI at Google Cloud, emphasized that the decision focused on World Labs’ technical requirements and scalability goals rather than personal ties to Li. “Fei-Fei is a friend of GCP, but the decision centered around their AI roadmap,” Lee said.

As cloud providers compete to meet the surging demand for AI model training, this collaboration underscores the fierce competition to secure AI-driven startups.

Though not an exclusive deal, Google Cloud currently hosts most of World Labs’ workloads and plans to continue providing key infrastructure as the startup scales.

With the AI landscape rapidly evolving, partnerships like this one are shaping the future of data processing and AI innovation.