Today, Hugging Face is announcing a deeper strategic partnership with Google Cloud to empower organizations to build their own AI using open models. This collaboration aims to make Google Cloud the leading platform for open AI development by combining Hugging Face’s vast ecosystem of models with Google Cloud’s powerful infrastructure and tools. Jeff Boudier, a leader at Hugging Face, said, “Google has made some of the most impactful contributions to open AI, from the original Transformer architecture to the Gemma models. I believe in a future where every company can build and customize its own AI. With this partnership, we’re making it easier than ever to do so on Google Cloud.” Ryan J. Salva, Senior Director of Product Management at Google Cloud, added, “Hugging Face has been a driving force in enabling companies worldwide to access, use, and customize over 2 million open models. We’ve contributed more than 1,000 models to the community, and together, we will make Google Cloud the best place to build with open models.” For Google Cloud customers, this partnership enhances the experience of working with open models across multiple services. In Vertex AI, popular models are available in Model Garden and can be deployed with just a few clicks. Customers seeking more control can use GKE AI/ML or pre-configured environments from Hugging Face. AI inference workloads can also run seamlessly on Cloud Run with GPUs, enabling serverless deployments. A key innovation in this collaboration is the development of a new CDN Gateway for Hugging Face repositories. Built on Hugging Face Xet storage and transfer technologies and Google Cloud’s advanced storage and networking, this gateway will cache models and datasets directly on Google Cloud. This will drastically reduce download times, improve reliability, and strengthen the model supply chain. Whether using Vertex, GKE, Cloud Run, or custom setups on Compute Engine, customers will benefit from faster startup times and simpler model governance. For Hugging Face users, the partnership brings new capabilities to Hugging Face Inference Endpoints, the fastest way to deploy models. Customers will gain access to new and more cost-effective instance types, with price reductions expected. The goal is to make deploying models on Google Cloud as simple as using public models, even for private models hosted in enterprise organizations. The partnership also brings full support for Google’s TPUs—its custom AI accelerators now in their seventh generation. Hugging Face is working to make TPUs as easy to use as GPUs through native integration in its libraries, unlocking high performance for open model development. Security is another focus. By leveraging Google’s Threat Intelligence and Mandiant’s expertise, the collaboration will help secure the millions of models, datasets, and Spaces on the Hugging Face Hub, protecting users daily. Together, Hugging Face and Google Cloud are building a future where every organization can build, customize, and host AI on their own terms—secure, scalable, and open. The partnership will accelerate this vision across Vertex AI, GKE, Cloud Run, and Hugging Face Inference Endpoints. If you have ideas for what should be built or improved through this collaboration, share them in the comments.