Google Cloud has unveiled the Gemini Enterprise Agent Platform, a comprehensive system for building, deploying and managing autonomous AI agents, positioning it as the central product announcement at its Cloud Next conference in Las Vegas.
The platform is the evolution of Vertex AI, Google's existing machine learning service, and all future Vertex AI development will be delivered through it rather than as a standalone product.
It brings together more than 200 models through Google's Model Garden, including its own Gemini 3.1 Pro, Gemini 3.1 Flash Image (also known as Nano Banana 2), the Lyria 3 audio model and Anthropic's Claude Opus 4.7.
"We've officially entered the agentic era," Google Cloud said.
The platform is split across three user tiers: a no-code Agent Designer for business users who can build agents using natural language or a visual interface, a low-code Agent Studio for developers, and a full Agent Development Kit (ADK) for engineering teams building complex multi-agent architectures.
ADK now processes more than six trillion tokens per month across Gemini models.
The Gemini Enterprise app, the consumer-facing layer of the platform, includes a new Agent Inbox for managing and monitoring agent activity, support for long-running agents that can operate autonomously in secure cloud sandboxes for days at a time, and trigger-based workflows for routine tasks.
Paid monthly active users of Gemini Enterprise grew 40% quarter on quarter in the first three months of 2026.
Google also announced an Agentic Data Cloud built around a Knowledge Catalogue that uses Gemini to autonomously tag and connect enterprise data, and a Cross-Cloud Lakehouse standardised on Apache Iceberg that allows agents to query data held in other cloud environments, including AWS, without requiring migration.
On the infrastructure side, Google unveiled eighth-generation TPUs split into two lines: the TPU 8t for training, delivering 121 exaflops per superpod, and the TPU 8i for inference, connecting 1,152 chips per pod with three times more on-chip memory.
The company also announced A5X instances powered by Nvidia's Vera Rubin NVL72 rack-scale systems, its Virgo networking fabric for supercomputer-scale interconnects and Managed Lustre storage capable of moving 10 terabytes per second.
Security enhancements include integration with Wiz's cloud security platform and three new specialised agents for threat hunting, detection engineering and third-party context analysis.
Related reading
- Microsoft warns AI agents risk becoming "double agents" as it unveils security controls at RSAC
- Meta expands AWS Graviton use for agentic AI
- Google launches autonomous research agents for finance, life sciences and market intelligence
Workspace Intelligence, a new unified context layer, embeds Ask Gemini across Docs, Drive, Meet and Gmail.
Early enterprise customers include Home Depot, which is using Gemini to power a phone and in-store shopping assistant, Papa John's for AI-driven ordering, Citadel Securities for quantitative research, Mars for marketing and Unilever, which said it is deploying agents across its operations to serve 3.7 billion consumers.
The recap
- Google Cloud unveils Gemini Enterprise Agent Platform for building agents
- TPU 8i offers 80% better performance per dollar for inference
- Company will offer NVIDIA Vera Rubin NVL72 systems alongside GPUs
