Google goes big at Cloud Next 2026

If 2025 was the year of experimenting with AI, 2026 is the year of putting it to work. Google is shifting its entire cloud and infrastructure strategy to support Agents: autonomous digital task forces that can manage everything from security triaging to complex code migrations.

The Gemini Enterprise Agent Platform

The headliner of the show is the Gemini Enterprise Agent Platform. Think of this as the mission control for the agentic enterprise. As organizations move from building a single assistant to managing thousands of specialized agents, they need a way to govern them. The platform (an evolution of Vertex AI) includes:

• Agent Registry: A central library to index and discover internal agents and skills.
• Agent Gateway: An “air traffic control” tower for administrators to enforce security policies and monitor agent activities.
• Long-running Agents: The ability for agents to run autonomously for days to solve complex problems, managed through a new unified Inbox.

8th Generation TPUs: Dual-Chip Strategy

To power these millions of agents, Google unveiled its 8th generation Tensor Processing Units (TPUs). For the first time, Google is splitting the TPU lineup into two specialized chips:

• TPU 8t (Training): Optimized for massive scale, a single superpod can scale to 9,600 chips. It delivers 3x the processing power of the previous Ironwood generation.
• TPU 8i (Inference): This is the workhorse chip designed to run trained models. It focuses on massive throughput and low latency, allowing Google to run millions of agents simultaneously and cost-effectively.

Google is ‘Customer Zero’ for AI coding

One of the most staggering stats from Sundar Pichai’s keynote involves how Google itself is using these tools. Pichai revealed that 75% of all new code at Google is now AI-generated and then approved by engineers.

This isn’t just basic autocomplete; Google’s engineers are now using agentic workflows to orchestrate autonomous task forces. For example, a recent complex code migration was completed six times faster than what was possible just one year ago. Even the Gemini app for macOS we recently covered went from a Swift prototype to a native release in just a few days thanks to this platform.

Cybersecurity with Wiz

Following the $32 billion acquisition of Wiz, Google is integrating Wiz’s Cloud and AI Security Platform directly into its cybersecurity stack. The new “AI Application Protection Platform” (AI-APP) provides autonomous protection across multi-cloud and hybrid environments, using AI to detect, find, and most importantly, fix software flaws automatically.

Altogether, you can really sense the momentum leap from Google in the AI landscape. There’s never really been any sign of slowdown, but the hardware and the AI running on it are both taking massive steps forward with these announcements, and as it has been for a few years now, it’s hard to determine where AI will take us next.