Google has announced that Gemini will be available on Google Distributed Cloud (GDC). This gives organizations access to Google’s most advanced AI models in their own data centers. This step allows companies with strict regulations or data restrictions to still use high-quality AI technology without moving their data to external clouds. The public preview will start in the third quarter of 2025.
Google is collaborating with NVIDIA to bring AI to the Google Distributed Cloud. By enabling NVIDIA to supply Blackwell systems, Google can run the Gemini models on-premises. Thus, organizations can have a complete on-premises AI environment.
Benefits for organizations with strict data requirements
Organizations that have to deal with strict regulations, sovereignty, latency requirements or large data volumes have had little access to the latest AI technology until now. They have mainly had to rely on open-source models and tools.
With Gemini on Google Distributed Cloud, organizations no longer have to choose between the best AI technology and the need to keep data on-premises. The air-gapped version of GDC, which Gemini will be available on, has already been approved by the US government for projects with ‘Secret’ and ‘Top Secret’ classifications.
Capabilities of Gemini on Google Distributed Cloud
The Gemini models that Google makes available on GDC offer excellent AI performance. They can analyze contexts of one million tokens. They are multimodal (processing text, image, audio and video) and work in more than 100 languages. Developers can use AI inferencing through the Gemini API without worrying about infrastructure or model management.
Vertex AI: one platform for cloud and on-premises
In addition to Gemini, customers can use the Vertex AI platform at GDC. This platform accelerates the development, implementation, and management of AI applications. It offers pre-trained APIs, generative AI tools, Retrieval Augmented Generation (RAG), and a built-in embeddings API with an AlloyDB vector database.
Google Agentspace: uniform access to on-premises data
Also new is Google Agentspace search, which will be available as a public preview at GDC in the third quarter of 2025. This offers an out-of-the-box solution. It provides access to all company data securely while taking into account the user rights.
Agentspace includes a multimodal search agent, connectors for common enterprise systems, and access controls. These ensure search results are tailored to user rights.
What is Google Distributed Cloud?
Google Distributed Cloud is a fully managed Google cloud solution that runs on-premises. It is available in both connected and air-gapped (fully isolated) variants and scales from a single server to hundreds of racks. GDC takes care of infrastructure management, allowing IT departments to focus on facilitating IT and developers to build applications.
This solution is an important step in the market for on-premises AI infrastructure, where other players such as HPE are already active. In addition to HPE, IBM offers its z17 mainframe to be able to run generative AI models on-premises.
Also read: Google Distributed Cloud offers sovereign cloud for AI in every data center