At Google Cloud Next 2026, the vendor unveils joint product and platform updates — including an embedded security layer

The company announced a slate of joint product and platform updates at Google Cloud Next 2026 and was named a Google Cloud Partner of the Year for the fifth time. Organizers credited the vendor with helping customers move generative AI projects from proofs of concept into production by integrating search, observability and security capabilities. One concrete rollout is Elastic Security as an embedded security layer for Google Distributed Cloud (GDC) air‑gapped environments. The deployment is aimed at organizations that must run workloads fully disconnected from the public internet, including government agencies, national security organizations, financial institutions and telecommunications providers.

On the models front, the vendor surfaced Jina Embeddings v3 as a self‑deployable model in the Gemini Enterprise Agent Platform’s Model Garden. The Gemini Enterprise Agent Platform — described as the evolution of Vertex AI-exposes access to more than 200 models, and the Jina model was the first of its kind available there, enabling teams to run high‑performance retrieval models inside their own cloud environments.

For infrastructure, the company introduced a CPU‑optimized Arm hardware profile for Elastic Cloud Hosted powered by Google Axion, Google’s custom Arm‑based processors. Elastic and Google Cloud said Axion can deliver up to about 25% better price‑performance for Elastic Cloud workloads and provided guidance mapping Axion SKUs to observability, security and search workload requirements. A joint session with Google Cloud explored how modern AI architectures pair specialized, open models with infrastructure choices. Presenters highlighted Cloud Run’s serverless GPU architecture as a scalable, cost‑effective runtime for deploying custom models and argued that running models closer to data helps produce permission‑aware, relevant answers across unstructured and semi‑structured sources.

Taken together, the announcements give builders three practical levers: an embedded security option for air‑gapped deployments, self‑deployable retrieval models in a curated Model Garden, and an Arm CPU profile tuned for Elastic workloads that can lower price‑performance. The vendor says these updates are intended to shorten the path from prototype to production for agentic, permission‑aware AI applications.