Google unveils 8th-gen TPUs, agent platform, and Workspace AI layer at Cloud Next '26

2026-04-22 · Source: The Decoder · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Robotics & Autonomous Systems · Depth: Advanced, long

Summary

Google introduced its eighth-generation TPUs, a new agent platform, and an AI layer for Workspace at Cloud Next '26, branding the suite as "Agentic Enterprise." The new TPUs are split into two variants: TPU 8t for training and TPU 8i for inference, moving away from a single-chip performance focus towards massive scale. Google's Virgo Network can link up to one million TPUs across multiple data centers, achieving a 97% "goodput" rate. The Gemini Enterprise Agent Platform, built on Vertex AI, simplifies autonomous AI agent creation and operation, offering long-term memory, sandboxed environments, and security features like cryptographic identities and anomaly detection. Workspace Intelligence centrally connects data across Google apps like Gmail, Docs, and Drive, enabling AI models to understand cross-application relationships and automate tasks.

Key takeaway

For CTOs and VPs of Engineering evaluating AI infrastructure, Google's split TPU architecture and "Agentic Enterprise" platform offer a compelling alternative to traditional GPU-centric approaches. Your teams should consider the scalability benefits of Google's million-TPU clusters and the integrated agent development tools to accelerate the deployment of autonomous AI systems, especially for multi-step enterprise workflows. This shift could significantly impact your compute cost-efficiency and development velocity for next-generation AI applications.

Key insights

Google's "Agentic Enterprise" strategy focuses on specialized hardware, scalable infrastructure, and integrated software for autonomous AI agents.

Principles

• Specialize hardware for training vs. inference.
• Prioritize scale over single-chip performance.
• Integrate AI across enterprise applications.

Method

The Gemini Enterprise Agent Platform provides tools for mapping multi-agent workflows, creating agents via natural language, and managing them with long-term memory and sandboxed execution.

In practice

• Utilize TPU 8t for large-scale model pre-training.
• Deploy TPU 8i for low-latency AI inference and reinforcement learning.
• Leverage Workspace Intelligence for cross-app AI automation.

Topics

• TPU 8t
• TPU 8i
• Gemini Enterprise Agent Platform
• Workspace Intelligence
• Agentic Enterprise

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Architect, Machine Learning Engineer, AI Scientist

Related on AIssential

• Google unveils two new TPUs designed for the "agentic era" — Google's new dual-chip TPU architecture optimizes AI training and inference for the emerging "agentic era."
• Google's TPU 8 Is A Direct Attack On NVIDIA - And It Rewrites AI Infrastructure Forever — Hyperscalers are shifting to specialized AI chips and vertically integrated stacks for the agentic era.
• We're launching two specialized TPUs for the agentic era. — Google's new TPUs are purpose-built for agentic AI, enhancing both inference and training capabilities.
• An Interview with Google Cloud CEO Thomas Kurian About the Agentic Moment — Google Cloud is advancing enterprise AI through agent-based automation, integrated infrastructure, and a multi-cloud data strategy.
• Google doesn't pay the Nvidia tax. Its new TPUs explain why. — Google's new specialized TPUs and vertical integration aim to reduce AI compute costs and enhance performance.
• Our eighth generation TPUs: two chips for the agentic era — Google's new TPU 8t and 8i chips specialize in AI training and inference for the agentic era.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Decoder.