Jul, 134 min to read
Platform
News · Google Cloud Next '26 makes the agent, not the model, the thing you operate
Apr, 244 min to read
PlatformGoogle Cloud Next '26 makes the agent, not the model, the thing you operate
The Gemini Enterprise Agent Platform, a no-code Agent Designer, eighth-generation TPUs, and live agents at Home Depot, Papa John's and Unilever.
The agent platform
The Gemini Enterprise Agent Platform is a workspace to build, govern and scale agents, with Gemini 3.1 Pro, Gemini 3.1 Flash Image (Nano Banana 2), Lyria 3 and Anthropic's Claude Opus 4.7 all available inside it. Agent Studio lets developers and business users build agents in natural language without machine-learning expertise, a no-code Agent Designer creates trigger-based workflows, long-running agents run autonomously in secure cloud sandboxes, and an Agent Inbox centralizes monitoring.
Silicon and data underneath
Eighth-generation TPUs split into 8t for training and 8i for inference, which Google says deliver 80% better performance per dollar, with Managed Lustre moving 10 terabytes per second and NVIDIA Vera Rubin NVL72 systems alongside. On the data side, the Cross-Cloud Lakehouse queries data where it lives — including on AWS — standardized on Apache Iceberg, so agents can reach data without a migration project first.
Where it is already running
Google backed the launch with named deployments rather than hypotheticals: Home Depot is running an AI phone and in-store shopping assistant, Papa John's built an ordering agent that uses contextual memory, Mars and Citadel Securities use it for quantitative research, and Unilever has deployed agents across its organization to serve a stated 3.7 billion consumers.
Governance and monitoring become the deliverable
The center of gravity has moved from integrating a single model to running many agents safely, which makes the Agent Inbox and the secure sandboxes the strategically interesting pieces rather than any one model. It is also telling that Google's own platform ships Claude Opus 4.7 next to Gemini — a signal that serious agent stacks are multi-model by default, and that routing and policy across providers is now part of the job.
Find this story relevant to you?
Contact us to find a unique solution
Need an AI engineering partner that can actually build?
We help businesses integrate AI, build AI-powered products, automate high-value workflows, and modernize the software systems behind them.
Related reading
More analysis around product delivery, operational AI, and the systems work that makes deployment hold up in reality.