Google is replacing Vertex AI with a new platform for building enterprise AI agents

Google launched the Gemini Enterprise Agent Platform, replacing Vertex AI with a unified enterprise AI stack for model selection, agent building, governance, and security. The platform includes access to 200+ models and new tools such as Agent Studio, Agent Development Kit, Agent Runtime, and Agent Gateway, with several large customers already using it in production. The launch strengthens Google's competitive position versus Amazon Bedrock AgentCore and Microsoft Foundry as Cloud revenue reached $17.7 billion in Q4 2025 with margins above 30%.

Analysis

This is less a product announcement than a distribution and control grab: Google is trying to turn AI agents into a cloud-native workflow layer where it can monetize model usage, orchestration, and governance simultaneously. The second-order effect is that enterprise buyers may increasingly standardize on a single vendor for agent identity, policy enforcement, and runtime, which raises switching costs far beyond model quality alone. That is structurally bullish for GOOGL because it deepens cloud attach rates and makes Workspace a sales funnel into the AI stack.

The competitive pressure lands most directly on AMZN and MSFT because the race is shifting from “best model access” to “best operating system for agents.” If Google can make long-lived agents, cross-platform communication, and auditability feel turnkey, it can steal share from multi-cloud buyers who care more about governance than raw inference benchmarks. The real risk for AWS and Azure is not immediate displacement, but that they get commoditized into compute pipes while Google captures the higher-margin control plane.

For CMCSA and PYPL, the upside is more operational than narrative: both are credible proof points that agentic workflows are moving from pilots to production, which should compress implementation cycles for adjacent enterprises over the next 2-4 quarters. The broader cybersecurity angle is important: once agents can hold state for days and transact across systems, identity and policy enforcement become budget line items, not optional features. That supports a faster spending cycle in security-adjacent cloud tooling even if overall IT budgets stay flat.

The contrarian view is that this may be too early to matter financially for the hyperscalers: enterprise agent adoption is still gated by liability, data access, and workflow redesign, so revenue lift could lag the headline product cycle by 6-12 months. Also, Google’s “full-stack” pitch is only durable if customers believe it will remain open enough to avoid lock-in; otherwise procurement teams will default back to multi-cloud hedging. Near term, the market may overestimate the immediate win rate for Google and underestimate the long-run margin expansion from governance-heavy AI infrastructure.

Key Decisions for Investors

• Overweight GOOGL vs MSFT and AMZN over the next 3-6 months; the risk/reward favors the company that controls both the agent control plane and distribution layer, with upside from cloud attach and workspace monetization versus limited incremental upside for the current leaders.
• Put on a long GOOGL / short AMZN pair trade into any post-launch weakness; target a 1-3 month horizon, betting that enterprise buyers will prioritize governance and integrated runtime over breadth of cloud ecosystem.
• Buy CMCSA on a 6-12 month horizon as a proof-point beneficiary; use weakness to build exposure because customer-service automation can translate into measurable opex savings faster than broad enterprise AI adoption.
• Initiate a smaller long PYPL position or call spread into the next 2 quarters; if production agent payments scale, PYPL can gain credibility as a transaction rail for autonomous commerce, but position size should reflect execution and regulatory risk.
• Avoid chasing MSFT/AWS on this headline; if anything, consider hedging their enterprise AI premium until evidence shows they can match Google’s control-plane integration without diluting ecosystem openness.