Google adds AI & partner tools to air-gapped cloud

Google has added new artificial intelligence, infrastructure and software partner features to its air-gapped Google Distributed Cloud platform for organisations that keep sensitive data in sovereign or disconnected environments.

The update includes enterprise search with Gemini, broader model support, new NVIDIA-based systems, networking and security changes, and software from Teradata, Solo.io and Mattermost.

The additions are aimed at customers in defence, intelligence, critical infrastructure and other regulated sectors where public cloud use may be limited by data sovereignty or security rules.

One of the main changes is the preview launch of enterprise search in Gemini on Google Distributed Cloud. The service gives users a conversational interface to query and summarise data held in air-gapped environments, including text, audio and video.

Google said the search tool uses retrieval-augmented generation, with indexing and interactions kept within the sovereign perimeter. This is intended to reduce the need for custom development in environments where analysts often work across compartmentalised information sources.

Model options

Gemini 3.5 Flash is now generally available on Google Distributed Cloud air-gapped. Google described the model as suitable for low-latency reasoning and multimodal work in disconnected environments.

Support has also been extended to open models including Gemma and Mistral, giving customers a choice of models for different workloads on the same platform.

Google said it is among the first large cloud providers to make a frontier model such as Gemini available in air-gapped deployments. The move reflects wider efforts by technology groups to bring generative artificial intelligence tools into environments that cannot connect to standard public cloud services.

Infrastructure changes

NVIDIA HGX B300 systems based on the NVIDIA Blackwell Ultra architecture are now generally available on the platform. They are intended to provide local compute for serving Gemini, open models and related agent-based workloads at production scale.

Virtual machine networking now supports up to 32 Gbps per virtual machine, up from 25 Gbps. Google said the change is meant to improve data movement to graphics processors and reduce bottlenecks in inference pipelines.

The platform now also includes fully managed Cloud DNS, allowing teams to manage DNS records through Kubernetes application programming interfaces and command-line tools already in use, according to Google. Hardware-offloaded IPsec encryption for node-to-node traffic is available at up to 100 Gbps.

Google is also expanding security operations support through integration with Elastic Security. The integration covers threat detection and response functions, including security information and event management and endpoint detection and response, in air-gapped deployments.

Partner ecosystem

The changes also expand the software available through the Google Distributed Cloud partner marketplace. Teradata, Solo.io and Mattermost have been newly added or expanded for sovereign deployments.

Teradata is bringing its data analytics and warehouse software to the platform for use at the distributed edge. Google said this will let customers run analytical pipelines near local artificial intelligence inference without moving data beyond the secure perimeter.

"Enterprises in highly regulated and air-gapped environments are eager to harness the transformative power of generative AI, but they cannot compromise on data privacy or analytical rigor. By bringing Teradata to Google Distributed Cloud, we are enabling our joint customers to run complex analytical workloads and AI inferencing right where their data resides. Together with Google Cloud, we are breaking down barriers between sovereign cloud infrastructure and enterprise-grade intelligence," said Sumeet Arora, Chief Product Officer at Teradata.

Solo.io is adding API management, service mesh and tools including kagent and agentgateway to these environments. Google said the software is intended to help customers connect and govern microservices and artificial intelligence agents across isolated clusters.

Mattermost is also being added for collaboration and operational workflows in defence and public sector settings. According to Google, the software includes artificial intelligence-assisted summaries and translation using models running locally on Google Distributed Cloud.

The announcement highlights how cloud providers are adapting artificial intelligence products for customers that cannot move sensitive information into shared environments. Rather than asking those organisations to build isolated artificial intelligence systems from scratch, vendors are increasingly offering pre-integrated software, model access and hardware within sovereign infrastructure.

For Google, the update expands Google Distributed Cloud beyond core infrastructure into a packaged environment for search, model deployment, analytics, security operations and team collaboration within an air-gapped perimeter.