Azure's New AI Gateway: All Your Models, One Key, Zero SLA — What Could Go Wrong?

Microsoft has released a dedicated AI Gateway tier of Azure API Management in public preview, organized around models, MCP servers, and tools rather than traditional APIs. The gateway fronts multiple providers including OpenAI, Anthropic, Mistral, AWS Bedrock, and Google Vertex AI, with policy cards replacing XML for token limits, quotas, content safety, and model fallback. It federates backends from remote MCP servers, OpenAPI specs, and over a thousand built-in SaaS connectors, splitting central platform control from application-team self-service. Key open questions remain around agent lifecycle governance, key blast radius, coexistence with existing APIM tiers, and pricing, which is unset during preview.