Google Cloud links Datadog observability

Google Cloud and Datadog have widened their AI monitoring partnership to cover Google’s newer agent-building tools, not just models and infrastructure. (cloud.google.com 1) (cloud.google.com 2) The newest step came on January 23, 2026, when Google Cloud said Datadog LLM Observability now automatically instruments systems built with Google’s Agent Development Kit, or ADK. (cloud.google.com) That setup lets teams trace planner decisions, tool calls, token usage, latency, costs, and errors in agent workflows without changing application code, according to Google Cloud and Datadog. (cloud.google.com) The companies had already announced a broader push on June 10, 2025, when Datadog added monitoring for agents deployed through Vertex AI Agent Engine in its AI Agents Console. (cloud.google.com) (datadoghq.com) In plain terms, observability is the record of what an AI system actually did: which prompt it received, which tool it picked, which other agent it handed work to, and where it slowed down or failed. Google’s documentation says instrumenting generative AI applications is the only way to understand the reasoning used by autonomous agents because that reasoning is not deterministic. (docs.cloud.google.com) That is a different problem from classic cloud monitoring, which mostly tracks servers, databases, and application response times. Google says AI agents can drift, hallucinate, and regress silently, creating failure modes that do not look like traditional software bugs. (cloud.google.com) Google is building its own native tooling here too. Its Gemini Enterprise Agent Platform documentation, published this month in preview, describes topology views, traces, logs, and metrics for deployed agents and Model Context Protocol servers using OpenTelemetry data. (docs.cloud.google.com) That means the Datadog tie-up is landing as Google is also standardizing how agent telemetry gets collected inside its own cloud stack. Google’s observability docs recommend OpenTelemetry as the common instrumentation layer for ADK and LangGraph-based applications. (docs.cloud.google.com 1) (docs.cloud.google.com 2) Datadog has been extending the partnership beyond the application layer as well, including Gemini and Vertex AI auto-instrumentation, Cloud TPU monitoring, GPU fleet monitoring, and expanded BigQuery cost monitoring. Google said Datadog revenue through Google Cloud Marketplace had more than doubled over the prior two years as of the June 2025 announcement. (cloud.google.com) The result is a fuller picture of how AI systems behave from prompt to infrastructure bill. For Google Cloud and Datadog, the pitch is that agent software now needs the same kind of tracing that distributed apps got a decade ago. (cloud.google.com) (datadoghq.com)