Google Cloud Next '26 wrapped up in Las Vegas last week with 260 announcements, 32,000 attendees, and three keynotes designed to leave you feeling like you'd witnessed the future of computing. They mostly succeeded. The stage production was impressive. The demos were polished. The messaging was confident.
Most of those 260 announcements are noise. A few of them are genuinely important. And somewhere in between is a product story that Google still hasn't figured out how to tell clearly — which, for a company betting everything on the agentic era, is worth talking about.
Here's the filter.
A2A at 150 Organizations in Production
The headline grabber at Next '26 was the Gemini Enterprise Agent Platform rebrand. But the announcement I kept coming back to was buried in the infrastructure details: Google's Agent-to-Agent (A2A) protocol has grown from 50 launch partners to 150 organizations in production.
That's significant. Not because the number is large — it isn't, by enterprise adoption standards — but because of what it signals. A2A is an open interoperability protocol that lets agents from different vendors communicate without any of them needing to understand each other's internal architecture. A Salesforce agent can hand a task to a Google agent, which queries a ServiceNow agent for IT asset data, and the whole chain executes without any of the three systems sharing a data model or API contract. They just speak A2A.
Native A2A support is now built into the Agent Development Kit, LangGraph, CrewAI, LlamaIndex Agents, Semantic Kernel, and AutoGen. When a protocol reaches this many frameworks in active production use, it's not a bet anymore — it's becoming infrastructure. Pay attention to it.
ADK v1.0 Stable
The Agent Development Kit hit stable v1.0 at Next '26, with releases across Python, Go, Java, and TypeScript. This is less glamorous than a new model announcement, but it matters more for most developers.
A stable 1.0 means the API surface is committed. You can build on it without worrying about a breaking change in the next release cycle. ADK is open-source, model-agnostic, and deployable to any container or Kubernetes environment. It's optimized for Gemini but not locked to it — you can use it with other models, including Claude and OpenAI's offerings.
If you're building agents professionally, this is the framework to anchor your development work around. It's not the flashiest announcement from the week, but it's the one with the longest shelf life.
Antigravity
This was the announcement that got the least mainstream coverage and is, I think, the most interesting development for individual developers.
Google Antigravity is a new agentic IDE — a direct competitor to Cursor and Windsurf, built from the ground up with an agent-first philosophy. Not AI autocomplete bolted onto an existing editor. An environment where agents plan, execute, and verify complex tasks across your editor, terminal, and browser. The Manager View lets you dispatch multiple agents simultaneously on independent tasks — parallel processing for your development workflow.
It's free for individuals in public preview, cross-platform (Mac, Windows, Linux), and model-agnostic — it supports Claude and OpenAI models alongside Gemini. That last point matters. Google is betting that if they build the best agentic development environment, developers will use it regardless of which model they're running. That's a smart play.
If you're already evaluating AI development environments, the cost of adding Antigravity to your list is essentially zero.
Gemini Enterprise Agent Platform
The Vertex AI rebrand is more than a naming exercise. All future Vertex AI services will be delivered exclusively through the Gemini Enterprise Agent Platform — Google's end-to-end system for building, governing, scaling, and optimizing agents. The platform consolidates Agent Studio (a low-code agent builder), ADK, A2A and MCP protocol support, Agent Registry, Agent Observability, and Agent Identity under one roof.
The Agent Gallery extends this with third-party agents from Salesforce, Workday, ServiceNow, Atlassian, Adobe, and others. Google is building the enterprise marketplace for agentic AI, positioned as the platform layer that sits underneath everyone else's agents.
If you're building enterprise AI on Google Cloud, this is where the roadmap lives. Understanding it now — before the migration conversations start — is worth the investment.
The Honest Section: Where Google Makes It Harder
Here's where I'm going to say something that didn't appear in any of the official recaps.
Google spent three days telling developers this is the agentic era. Then they shipped Firebase Studio, Antigravity, and Gemini CLI as three distinct developer-facing products with significant capability overlap and no clear guidance on which one to reach for.
This is Google's chronic product problem — and it's well-documented. The messaging app graveyard (Hangouts, Allo, Duo, Chat) is the canonical example, but it shows up across the portfolio. Google builds excellent individual products and then struggles to explain which one you should actually use.
The difference now is the stakes. If developers are confused about which messaging app to use, they pick iMessage and life goes on. If developers are confused about which agentic development platform to invest their time in, they pick Cursor or Claude Code and Google loses them from the ecosystem for years.
A developer sitting in a Next '26 session could walk out genuinely uncertain: should I build my next agentic application in Firebase Studio? In Antigravity? Directly with the ADK? These aren't rhetorical questions — they're practical decisions with real implications for where you put your time. The conference didn't answer them clearly. Neither do the product pages.
The $750 million partner fund Google announced is impressive, and it also tells you something. When your product story requires hundreds of partners and forward-deployed Google engineers to explain it to customers, the product story has a problem. Google has the right pieces. The integration between those pieces — and the narrative about which piece serves which purpose — is still a work in progress.
What Should You Actually Do?
Watch A2A. It's becoming protocol-level infrastructure regardless of which IDE or framework wins the developer tools war. Understanding it now puts you ahead of the enterprise mandates that are coming.
Commit to ADK if you're building agents in production. Stable 1.0 is a real milestone — it means you can ship on it without the ground shifting underneath you.
Try Antigravity. The preview is free. If you're already evaluating AI development environments, add it to your list. It's the most interesting new entrant in that space, and you won't know until you run it on a real project.
On the enterprise platform question: Gemini Enterprise Agent Platform is where Google's investment is going. If you're building on Google Cloud, that's the foundation to understand.
And on the product confusion: Google made serious, substantive moves at Next '26. The underlying capability is real. Just don't let the keynote decide which tool you reach for — it was designed to impress 32,000 people simultaneously, not to guide your technical decisions. Do your own evaluation.
Sources: Google Cloud Next 2026: News and updates, The new Gemini Enterprise: one platform for agent development, Build with Google Antigravity, Google Cloud Commits $750 Million to Accelerate Partners' Agentic AI Development
