Vertex AI

Vertex AI launched in 2021 as Google Cloud's unified ML platform: a managed service where teams could pick a model, train it, deploy it, and run inference without managing the underlying infrastructure. It became the primary way enterprise developers accessed Gemini models via API, separate from the consumer Gemini app. At Cloud Next 2026, Google announced a structural shift: Vertex AI's features are being consolidated into the Gemini Enterprise Agent Platform, a new product built explicitly for multi-agent, multi-model production workloads.

The core Vertex AI capabilities (Model Garden with access to Gemini, Llama, Claude, Mistral, and other models; Custom Training; AutoML; Model Registry; Endpoints; and Pipelines) persist in the new structure. What changed is the framing: instead of 'deploy one model to one endpoint,' the Gemini Enterprise Agent Platform is designed for teams running hundreds or thousands of concurrent AI agents. It adds built-in agent state management, an agent registry, inter-agent communication protocols, and enhanced security and compliance tooling.

For builders on Google Cloud, the practical implication is that new projects should target the Gemini Enterprise Agent Platform APIs rather than older Vertex AI SDK modules (a migration deadline of June 24, 2026 applied to deprecated SDK components). Vertex AI remains the access point for fine-tuning Gemini models on proprietary data, running model evaluations, and serving custom model checkpoints. It also supports Claude models from Anthropic and open-weight models alongside Google's own Gemini family.