The Strategic Partnership Between Hugging Face and Google Cloud: A New Era for AI Deployment

In a significant move poised to revolutionize AI development, Hugging Face has announced a deeper partnership with Google Cloud. This collaboration aims to empower companies to build and customize their own artificial intelligence applications using open models. As Jeff Boudier of Hugging Face aptly stated, "Google has made some of the most impactful contributions to open AI," envisioning a future where all businesses will have the tools to create tailored AI solutions.

A Partnership for Google Cloud Customers

Today, Google Cloud users can leverage the power of Hugging Face’s extensive library of open models across various services. Notably, within Vertex AI, these models are accessible for quick deployment through the Model Garden feature, streamlining the process for customers. For those seeking greater control over their AI implementations, Google Kubernetes Engine (GKE) provides a robust model library alongside pre-configured environments from Hugging Face. This allows for efficient AI inference workloads through Cloud Run GPUs, ensuring that open model deployments can be managed effectively on a serverless platform.

The seamless experience created through this partnership fully utilizes the unique capabilities of both Hugging Face and Google Cloud, offering unparalleled flexibility to customers.

The Fast Lane for Open Models

The adoption of Hugging Face models among Google Cloud customers has surged by an impressive 10x over the past three years, translating into staggering numbers: tens of petabytes of model downloads each month and billions of requests. To enhance the experience for these users, the partnership is set to introduce a CDN Gateway.

This gateway will utilize Hugging Face’s optimized storage and data transfer technologies, coupled with Google Cloud’s advanced storage capabilities. By caching models and datasets directly on Google Cloud, download times will be dramatically reduced, leading to greater efficiency in using these resources. The result? Quicker time-to-first-token and improved model governance for customers, regardless of whether they utilize Vertex, GKE, Cloud Run, or custom deployments in Compute Engine.

A Win-Win for Hugging Face Customers

For users of Hugging Face, the partnership brings exciting advancements, starting with Inference Endpoints. These endpoints allow for a straightforward transition from model to deployment. With improved integrations offered by Google Cloud, users can expect more instance types and most importantly, a drop in costs.

By streamlining the process of deploying models on Google Cloud platforms, Hugging Face aims to make it easier than ever for AI Builders—currently numbering around 10 million—to seamlessly transition from developing models to deploying them in real-world applications. Importantly, taking a private model securely hosted on Hugging Face will soon be as effortless as working with public models.

Harnessing the Power of TPUs

Google’s Tensor Processing Units (TPUs), now in their seventh generation, represent another pivotal aspect of this partnership. By enhancing access to advanced TPUs, Hugging Face users can expect improved performance and a richer software environment. The goal is clear: to ensure that building AI with open models is as accessible using TPUs as it currently is with GPUs, all facilitated by direct support in Hugging Face libraries.

This collaboration is also poised to bolster the security of Hugging Face’s vast repository of models, thanks to Google’s industry-leading security technology. With backing from Google Threat Intelligence and Mandiant, the partnership will deepen the safety of models, datasets, and Spaces used daily on the Hugging Face Hub.

Building an Open Future in AI Together

The overarching vision for this partnership is a future in which businesses can craft their own AIs using open models within a secure framework. The collaboration between Hugging Face and Google Cloud is set to accelerate this conceptual landscape, whether through Vertex AI Model Garden, GKE, Cloud Run, or Hugging Face Inference Endpoints.

This partnership represents a crucial step toward democratizing AI, ensuring that customization and accessibility are at the forefront. As the landscape of AI continues to evolve, the implications of this collaboration promise to reshape how organizations engage with open models to create cutting-edge AI solutions.

This strategic partnership between Hugging Face and Google Cloud is paving the way for intuitive, scalable AI applications at unprecedented speeds. Stay tuned for updates and enhancements that arise from this innovative collaboration!

Inspired by: Source