Market shifts and the Datadog defection
Datadog revenue reached $1.121 billion in Q2 2026, which showed 36% year-over-year growth. Its metered-dimension pricing model drives enterprise customers toward the Grafana LGTM stack. Organizations migrate to Grafana when Datadog renewal costs exceed the engineering salary required to manage Loki, Grafana, Tempo, and Mimir. Datadog holds 51.82% of the observability market, while Grafana maintains 4.03%. You already know that Datadog uses a proprietary model, but the sudden defection of cost-sensitive users changes the competitive landscape. Engineering teams leave Datadog because they cannot manage the billable dimensions that compound as infrastructure grows. This shift happens because Datadog meters hosts plus a half-dozen separate dimensions, including custom metrics and log ingest. The broader observability market reaches $18 billion by 2029, though CloudToggle estimated overall adoption at just 12.0% of businesses in March 2026.
AI and cost optimization
Grafana Labs uses AI to lower observability bills through Adaptive Telemetry. Adaptive Metrics reduces billable metric series by 80%. Adaptive Logs reduces log ingest by 50% by identifying and pruning low-value patterns. TeleTracking used Adaptive Metrics to cut its billable metrics series by 50%, reducing costs by 33%. The company also acquired TailCtrl to accelerate development of Adaptive Traces, though this feature remains in the research phase.
| Product | Pricing Model | Specific Limit/Rate |
|---|---|---|
| Grafana Cloud Free | Usage-based | 50 GB logs and 50 GB traces |
| Grafana Cloud Pro | Usage-based | $19/month platform fee |
| Adaptive Metrics | Tiered | $6.50 per 1,000 series (Tier 1) |
| Adaptive Logs | Usage-based | $0.50/GB for additional usage |
Grafana Labs also offers a Startup Program for 50 eligible companies with less than $10 million in funding and fewer than 25 employees. Participants receive up to $100,000 in Grafana Cloud credits for 12 months. This program aims to prevent future vendor lock-in. The company also provides k6 Studio, a GUI-based, low-code test authoring tool that generates scripts compatible with both synthetic monitoring and load testing for developers who lack strong programming skills or prefer not to hand-script in JavaScript. The Pro plan includes 13 months of metrics retention and 30 days of log, trace, profile, and k6 performance test retention.
Deep tracing and LLM monitoring
The October launch of Tempo 2.7 changes how engineers interact with distributed tracing. Previously, Tempo only retrieved traces by ID, but new features allow users to dig deep into application data. This functionality provides insights that users previously ignored. Grafana engineers also built tools to track ML experiments. The solution uses a Python library to instrument training code and a Go backend to store data about training runs. This tool charts model metrics like loss versus epoch to help developers.
Grafana also provides GPU monitoring via eBPF, which helps AI developers get information about workloads without manual instrumentation. This capability exists in a branch of the Beyla open source project. Grafana also provides Browser Testing, which runs a real browser to capture WebVital metrics, screenshots, and timings to validate the full frontend user experience. This functionality is generally available. The company also offers Scripted Monitoring, which allows developers to perform black-box synthetic monitoring of services using JavaScript. This feature is in public preview.
How will the convergence of LLM monitoring and distributed tracing redefine the standard observability workflow?
Grafana engineers also explore ways to use open source tools to observe LLMs, including the OpenLIT SDK. This SDK produces OpenTelemetry traces and metrics for the LLM calls including latency, cost, and the number of tokens generated. Grafana also provides Incident Rooms with real-time transcription. A bot uses LLMs to summarize conversations and capture insights from incident bridge calls. This feature is experimental.
