Today, we’re excited to announce a new and deeper partnership with Google Cloud that enables companies to build their own AI using open models.
“Google has made some of the most influential contributions to open AI, from OG Transformers to the Gemma model. I believe in a future where every company builds and customizes their own AI, and this new strategic partnership makes it easy to do that on Google Cloud,” said Jeff Boudier of Hug Face.
“Hugging Face is the powerhouse that enables businesses large and small around the world to access, use, and customize over 2 million open models today, and we’re proud to contribute over 1,000 models to the community,” said Ryan J. Salva, senior director of product management at Google Cloud. “Together, we are making Google Cloud the best place to build with an open model.”
Partnerships for Google Cloud customers
Google Cloud customers use Hugging Face’s open model for many of its leading AI services. Vertex AI allows you to deploy the most popular open models within Model Garden with just a few clicks. Customers who want more control over their AI infrastructure can find similar model libraries available in GKE AI/ML or use preconfigured environments managed by Hugging Face. Customers also use Cloud Run GPUs to run AI inference workloads and enable serverless open model deployments.
What we have in common: Together with Google Cloud, we leverage the unique capabilities of each service to create seamless experiences and give customers choice.
Gateway to an open model – fast lane for Google Cloud customers
Google Cloud customers’ usage of Hugging Face has increased 10x over the past three years, which now equates to tens of petabytes and billions of requests worth of model downloads every month.
To ensure Google Cloud customers have the best experience building with Hugging Face models and datasets, we’re working together to create a CDN gateway for Hugging Face repositories built on both Hugging Face Xet-optimized storage and data transfer technology and Google Cloud’s advanced storage and networking capabilities.
This CDN gateway caches Hugging Face models and datasets directly to Google Cloud, significantly reducing download times and increasing the robustness of the model supply chain for Google Cloud customers. Whether you’re using Vertex, GKE, Cloud Run, or building your own stack on VMs on Compute Engine, you can benefit from faster time to first token and simplified model governance.
Partnership for Hug Face Customers
Hugging the facial inference endpoint is the easiest way to go from model to deployment in just a few clicks. Through this deepening partnership, we will provide Hugging Face customers with the unique capabilities and cost performance of Google Cloud, including inference endpoints. Expect more new instances to become available and prices to drop as well.
We will make all the results of our product and engineering collaborations easily available to the 10 million AI builders on Hugging Face. From your model page to deployment to Vertex Model Garden or GKE takes just a few steps. Getting a private model hosted securely in your enterprise organization on Hugging Face should be as easy as using a public model.
Google’s custom AI accelerator chip, the TPU, is now in its seventh generation, with steady improvements in performance and software stack maturity. We want Hugging Face users to fully benefit from current and next-generation TPUs when building AI in an open model. We’re excited to make TPUs as easy to use as GPUs in Hugging Face models thanks to the library’s native support.
Additionally, this new partnership will allow Hugging Face to leverage Google’s industry-leading security technology to make the millions of open models on Hugging Face more secure. This joint effort, powered by Google Threat Intelligence and Mandiant, aims to protect your models, datasets, and spaces as you use Hugging Face Hub every day.
Build an open future of AI together
We want to see a future where every company can use an open model to build their own AI and host it with full control within their own secure infrastructure. We’re excited to make this future a reality with Google Cloud. Whether you’re using Vertex AI Model Garden, Google Kubernetes Engine, Cloud Run, or the Hugging Face Inference endpoint, our close collaboration accelerates this vision.
Is there anything you would like us to create or improve thanks to our partnership with Google? Let us know in the comments!
