Grafana Labs has passed $600 million in annual recurring revenue, up at least 50 percent from $400 million in September, as companies move from AI experimentation to production and confront the cost and reliability of running models at scale. The 12-year-old observability company also crossed 10,000 customers globally, a milestone CEO Raj Dutt says reflects a market no longer asking whether AI works but how much it costs, whether it is governed, and whether it can be trusted.
AI agents create new monitoring surface
Traditional observability looked for outages and latency. AI agents introduce a different class of failure: recommending a competitor, leaking data, or behaving in ways no test suite anticipated. Dutt said that shift has expanded the monitoring surface and, with it, the bill. Grafana’s answer is Adaptive Telemetry, a set of tools that reduce the volume of data customers pay to ingest. Dutt acknowledged the approach has cost the company close to $100 million in potential revenue but argued the revenue that remains is higher quality because it aligns with actual usage.
Product velocity and early traction
In July Grafana launched six AI features. Hundreds of customers are already using Agent Observability to track speed and reliability of autonomous agents. More than 18,000 organizations, including Alter Domus and Deutsche Telekom, have adopted Grafana Assistant, an AI tool that helps engineers troubleshoot their own stacks. Dutt called it the fastest-growing product in the company’s history and said the majority of its users are paying customers.
Competitive context
Datadog reported this month that its own customers are increasingly deploying AI and using its platform to monitor and secure those workloads, according to cofounder and CEO Olivier Pomel. The parallel narratives suggest the observability layer is becoming the primary toll booth for enterprise AI adoption, whoever instruments the stack captures the recurring revenue.
What to watch
The $100 million revenue haircut from Adaptive Telemetry is a deliberate trade-off; the test is whether the resulting net retention outperforms peers who bill every gigabyte. Also watch whether Grafana Assistant’s paid conversion holds as the free tier saturates, and whether the six July features deepen lock-in or merely match Datadog feature for feature. The next ARR update will show if the deployment wave is still accelerating or if the easy budgets have been spent.