During Next ’26, Google Cloud is introducing a unified AI stack for autonomous agents, built on its own hardware and applications for security, data, and office work.
At its annual Next conference in Las Vegas, Google Cloud is loudly proclaiming that the era of the ‘Agentic Enterprise’ has begun. This somewhat vague buzzword manifests as a multi-layered AI stack, based on hardware and cloud software, upon which organizations build their own autonomous agents that process information, reason, and perform complex tasks. According to Google’s cloud division, this marks a shift from standalone chatbots to integrated, scalable systems.
The announcements include new tools for building, scaling, and managing AI agents, as well as a new generation of Google’s own TPUs. Google is also introducing new agents for data management, cybersecurity, and collaboration tools.
Gemini Enterprise: building, scaling, and managing
A key layer in the AI stack is the Gemini Enterprise Agent Platform, also available as an app. Through this cloud-based platform, AI agents can be developed, coordinated, and managed. The platform combines Vertex AI with new features such as Agent Studio and Agent-to-Agent Orchestration, allowing agents to not only communicate with each other but also execute complex business processes. Gemini Enterprise provides access to the latest Gemini models, as well as those from Anthropic.
Agents can remain active for up to seven days without losing state. Additionally, the platform offers enhanced capabilities for optimizing agents, such as automated stress tests and cost management. Google Cloud states that these tools guarantee the reliability and security of AI systems.
Proprietary hardware
In addition to software, Google Cloud is introducing the eighth generation of TPUs, including the TPU 8t for training and the TPU 8i for inference. Google promises cost efficiency and low latency. This hardware is combined with storage and networking via Lustre and Virgo Networking. Google also offers access to Nvidia chips through its cloud.
Agents for security, data, and collaboration
In the area of security, Google Cloud is launching Agentic Defense, a platform that detects, prevents, and responds to threats using AI. Google is leveraging technology from its recently acquired subsidiary Wiz for autonomous protection from code to cloud to runtime, suitable for multi-cloud and hybrid environments.
New agents for threat detection and remediation are designed to help organizations proactively manage cyber risks. Agents can be placed in both the ‘red team’ and the ‘blue team’ to identify vulnerabilities or secure them.
Agentic Data Cloud provides an AI-native architecture that allows organizations to make data in a multi-cloud environment available to agents without having to move data between platforms. In the Knowledge Catalog, AI agents can be fed with corporate data and context. In Agentic Taskforce, Google Cloud links the Gemini Enterprise platform to Workspace applications.
Google Cloud concludes that the transformation to an ‘Agentic Enterprise’ is not a distant dream, but an immediate step for organizations looking to scale their AI strategy. According to CEO Thomas Kurian, the combination of hardware, software, and security provides the foundation for a new generation of productive AI systems that can help organizations grow.