Google Cloud launched the Gemini Enterprise Agent Platform at Cloud Next 2026 in Las Vegas on April 22, formally retiring Vertex AI as a standalone service and replacing it with a unified system for building, deploying, governing, and optimizing autonomous AI agents. The platform reported 40% quarter-over-quarter growth in paid monthly active users during Q1 2026, with production customers including Bosch, Capcom, GE Appliances, NASA, PepsiCo, and Unilever.

From Model Platform to Agent Infrastructure

The rebrand signals more than a naming change. Vertex AI was designed for the early generative AI era, when the primary challenge was building and deploying models. Gemini Enterprise Agent Platform restructures the entire surface around agent lifecycle management.

“Today, we’re managing a different level of complexity, with agents interacting across multiple systems, and often without security and governance guardrails,” Google Cloud VP of Product Management Michael Gerstenhaber wrote in the launch post. “To move toward a truly autonomous enterprise, one where agents can act with the same independence and reliability as a member of your team, you need a foundation that can sustain that level of trust.”

CEO Sundar Pichai framed the shift in his keynote blog post: “The conversation has gone from ‘Can we build an agent?’ to ‘How do we manage thousands of them?’”

All Vertex AI services and future roadmap developments will now be delivered exclusively through the Agent Platform, according to Google’s announcement.

Platform Architecture

Google organized the platform into four pillars: build, scale, govern, and optimize.

Build. Agent Studio provides a low-code visual interface for business users to drag and drop agent logic. The upgraded Agent Development Kit (ADK) supports graph-based multi-agent frameworks, native ecosystem integrations with BigQuery and Pub/Sub, and batch and event-driven agent execution. The platform provides access to more than 200 AI models through Model Garden, including Gemini 3.1 Pro, Gemma 4, and third-party models like Anthropic’s Claude, according to Google’s blog.

Scale. A re-engineered Agent Runtime supports long-running agents that maintain state for days. Agent Memory Bank dynamically generates and curates long-term memories from conversations, with Memory Profiles enabling low-latency recall of interaction history, as detailed by SiliconANGLE.

Govern. Each agent receives a unique cryptographic identity with auditable action trails mapped to authorization policies. Agent Registry indexes internal agents, tools, and skills for centralized discovery. Agent Gateway acts as fleet-level oversight, enforcing security policies across all deployed agents, per SiliconANGLE’s reporting.

Optimize. Agent Simulation tests agents on synthetic workloads before deployment. Agent Evaluation continuously scores production performance. Agent Observability provides visual reasoning traces for debugging, and Agent Optimizer automatically refines system instructions.

Adoption Metrics and Customer Base

Pichai disclosed 40% quarter-over-quarter growth in paid monthly active users in Q1 2026, according to Google’s keynote blog. Named customer deployments span industries: Bosch (manufacturing automation), GE Appliances (product design), NASA (mission planning), PepsiCo (supply chain optimization), and Unilever (marketing operations), as reported by NewsBytes.

Comcast rebuilt its Xfinity Assistant using ADK, moving “beyond simple scripted automation to conversational generative intelligence,” according to CTO Rick Rioboli in a customer quote published by Google. Burns & McDonnell described using the platform to turn “decades of project data into real-time, actionable intelligence” through agents that combine “deterministic business rules with probabilistic reasoning.”

Companion Announcements

Google bundled the platform launch with two infrastructure announcements. Agentic Data Cloud introduces a cross-cloud lakehouse architecture and knowledge catalog for agent data access spanning Google Cloud, AWS, Azure, and on-premises infrastructure, according to Pichai’s blog. The cross-cloud positioning removes a lock-in concern that has historically limited agent deployment in multi-cloud enterprises.

Separately, Google unveiled Agentic Defense, a security operations platform built with Wiz that targets threats specific to autonomous agents, including prompt injection, model poisoning, and unauthorized action execution.

The Platform Consolidation Race

Google’s move lands one day after Microsoft unveiled Agent 365, its own rebrand from Copilot to autonomous operational agents. The timing is not coincidental. Both companies are betting that agent governance, not model capability, is the primary enterprise bottleneck.

The distinction in Google’s approach is integration depth. By absorbing Vertex AI entirely, Google is forcing customers to adopt agent-centric infrastructure as the default development surface, not as an add-on to existing ML tooling. Whether that integration creates lock-in or genuine operational advantage will depend on how enterprises evaluate the four-pillar architecture against Microsoft’s governance-first approach and Salesforce’s CRM-native Agentforce.

For platform teams evaluating agent infrastructure, the announcement compresses the decision timeline. The two largest cloud providers now offer dedicated agent platforms with governance built in. The question is no longer whether agent management needs its own infrastructure layer, but which vendor’s architecture matches your deployment pattern.