Vendor
Grafana Cloud pricing 2026: an independent read
Grafana Cloud charges per active metric series, not per host. Cardinality is the bill, labels are the lever, and the free tier is the most generous in the market. Here is the math at three real-world scales.
TL;DR
Metrics: $6.50 per 1,000 active series above 10K free. Logs (Loki): $0.45/GB ingest above 50 GB free ($0.05 process + $0.40 write). Traces (Tempo): $0.45/GB above 50 GB free. Visualization users: $8/active user (3 free), plus a $19/mo Pro platform fee. A 100-host team with disciplined Kubernetes labelling sits at $950 to $1.4K/mo. Cardinality explosion can push that to $10K+ on the same hosts.
Direct answer
Grafana Cloud free tier 2026: metrics, logs, traces and limits
| Metrics (active series) | 10,000 series |
| Logs (Loki) | 50 GB ingested / month |
| Traces (Tempo) | 50 GB ingested / month |
| Profiles (Pyroscope) | 50 GB ingested / month |
| Grafana visualization users | 3 active users |
| k6 load testing | 500 virtual user hours |
| Synthetic monitoring | 100K API + 10K browser test executions |
Free-tier metrics, logs, traces and profiles all retain for 14 days. Paid (Pro) usage above these allowances bills $6.50 per 1,000 active series, roughly $0.45 per GB of logs / traces / profiles, $8 per active visualization user, plus a flat $19/month platform fee that extends metrics retention to 13 months and logs / traces to 30 days. Verify on the current Grafana Cloud pricing page before purchasing.
Interactive
Grafana Cloud pricing calculator
Your Environment
Move the dials. Watch costs change.
Pre-filled scenarios
Estimated monthly cost (USD list)
sorted: cheapest first
Grafana Cloud bills per active series, not per host, so the figure above moves most when you change custom metrics per host, not host count. For the full six-vendor tool with per-line-item breakdowns and the exact arithmetic, see the multi-vendor cost calculator.
The pricing model
Why active series, not per host
The active-series model is the most architecturally honest cost meter in the cloud monitoring market. Grafana Cloud charges based on the number of unique time series being actively reported, where a series is one metric name plus one specific combination of label values. This maps directly to the underlying Prometheus storage cost. The vendor pays for the storage and compute proportional to series count, the customer pays a margin on top of that. There is no per-host fiction, no per-GB fiction, just the actual cardinality being collected.
For customers running a clean monitoring stack, this model is genuinely cheap. A 10-host Kubernetes cluster with Prometheus-style metrics produces roughly 1,500 to 5,000 active series, comfortably inside the 10,000-series free tier. A 100-host cluster with disciplined labelling produces 30,000 to 60,000 series, billing $130 to $325 per month at the $6.50 per 1,000 series rate. The same workload on Datadog at $15 to $18 per host per month plus custom metric overages typically lands at $2,500 to $4,500 per month.
For customers who do not control cardinality, the same model is brutally expensive. A single metric with one label that has 100,000 unique values (a request path with high-cardinality user IDs in the URL, a Kubernetes container ID, a Kafka offset) creates 100,000 series. Multiply by ten hosts and you are at one million billable series, $6,435 per month, on a single misconfigured metric. The Datadog model would have absorbed the first hundred custom metrics per host for free and only charged overage above that.
The other meters work the way Prometheus operators expect. Loki indexes log labels rather than full text, which makes log storage roughly five to ten times cheaper than Splunk-style indexing on equivalent volumes. Tempo charges per gigabyte of trace data, with sampling expected at the source via OpenTelemetry collectors. Pyroscope (continuous profiling, integrated since the 2023 acquisition) charges per gigabyte. The bundle is structurally cheaper than per-host vendors at small to mid scale and structurally more expensive at high cardinality without discipline.
Every product line
The Grafana Cloud product matrix
| Product | Free tier | Paid (Pro) | How it bills |
|---|---|---|---|
| Metrics (active series) | 10,000 series | $6.50 per 1,000 series above free | Cardinality drives cost, not host count |
| Logs (Loki) | 50 GB/month | $0.45/GB ingest ($0.05 process + $0.40 write); $0.10/GB per extra 30-day retention block | Indexes labels, not full text, much cheaper than Splunk-style indexing |
| Traces (Tempo) | 50 GB/month | $0.45/GB ingest; $0.10/GB per extra 30-day retention block | OpenTelemetry-native; sampling at the source recommended |
| Profiles (Pyroscope) | 50 GB/month | $0.45/GB ingest; $0.10/GB per extra 30-day retention block | Continuous profiling integrated since the 2023 Pyroscope acquisition |
| Synthetics + k6 load testing | Synthetics 100K API + 10K browser test executions; k6 500 virtual user hours | $5 per 10K API test executions; k6 from $0.15/virtual user hour | Synthetics meters by test execution; k6 meters separately by virtual user hour |
| User seats + platform fee | 3 users | $8/active user (Grafana visualization) + $19/mo Pro platform fee | Viewer seats free; the flat platform fee applies to every Pro stack |
Three scenarios
What real teams pay
Scenario
Startup, 10 hosts
- Active series (10 hosts x ~150 series)1,500 series
- Logs (10 GB/mo)0 GB billable
- Traces (10 GB/mo)0 GB billable
- Users (3 max)$0
Total: $0/month
10K active series, 50 GB logs, 50 GB traces and 3 users cover small Kubernetes clusters comfortably. Grafana's free tier is the most generous in the market for Prometheus-shaped stacks.
Scenario
Mid-market, 100 hosts + 50 GB/day logs
- Active series (100 hosts x ~150 series + 200 custom each)35,000 series
- Billable series (above 10K)25,000 series
- Series cost (25K x $6.50/1,000)$163
- Logs (1,500 GB/mo)1,450 GB billable
- Logs cost (1,450 x $0.45)$653
- Traces (200 GB/mo)$68
- Pro user seats (~5 editors) + $19 platform fee$59
Total: ~$950 to $1,400/month
Cardinality is the main lever. A team that keeps Kubernetes labels disciplined can run 100 hosts for under $1,000.
Scenario
Enterprise, 1,000 hosts + 500 GB/day logs
- Active series (~1.2M total)1.2M series
- Series cost (1.19M x $6.50/1,000)$7,735
- Logs (15,000 GB/mo)$6,728
- Traces (2,000 GB/mo)$878
- Pro user seats (~15 editors)$120
Total: ~$15,000 to $20,000/month list
Negotiated rates apply. Grafana Cloud Enterprise contracts start at a $25,000/year spend commitment and discount on annual commitments.
Where it bites
Three cardinality traps that blow up Grafana Cloud bills
The Kubernetes pod label
The high-cardinality status code
The third-party exporter without limits
Where the model rewards
When Grafana Cloud is the right call
Grafana Cloud is the natural choice for three customer profiles. The first is the team already running open-source Prometheus, Loki, or Grafana on-premises and wanting to migrate the storage tier to managed without rewriting dashboards, alert rules, or instrumentation. The dashboards transfer one-for-one, the PromQL queries are identical, and the migration risk is low. Teams with 50 to 200 hosts on self-managed Prometheus often find Grafana Cloud cheaper once they account for the engineering time spent on Prometheus federation, long-term storage with Cortex or Thanos, and the on-call rotation for the observability stack itself.
The second is the small startup running a Kubernetes-native stack with disciplined labelling. The free tier is large enough to cover real production workloads, the paid tier is honest about how cardinality drives cost, and the operational complexity is lower than self-hosting. A seed-stage company with 20 hosts and three engineers can run on the free tier indefinitely, then graduate to Pro at $200 to $500 per month as the business scales.
The third is the platform engineering team standardising the entire organisation on OpenTelemetry. Grafana Labs has invested heavily in OpenTelemetry support since 2024, and the Tempo trace backend is purpose-built around OTel native ingest. Teams that have already adopted OTel for vendor independence find Grafana Cloud the most natural commercial backend, with no proprietary agent lock-in.
Grafana Cloud is less suited to teams that want zero-config monitoring with one auto-instrumented agent (Datadog and New Relic are stronger here), teams without Prometheus or PromQL comfort, teams running deep APM workloads on legacy languages where vendor instrumentation is more mature than community OTel libraries, and teams with regulatory requirements that make the multi-tenant SaaS model non-viable (Grafana Enterprise self-managed is the answer there, not Grafana Cloud).
Tier deep-dive
Grafana Cloud Enterprise (formerly Advanced): what enterprise teams actually need it for
The Free and Pro tiers are usage-priced and self-serve; the Enterprise tier (which Grafana Labs previously marketed as "Advanced") shifts to a custom-quoted contract starting at a $25,000 per year spend commitment, and adds the controls that procurement, security, and platform-engineering teams need before they can roll Grafana Cloud out across a regulated organisation. The differences from Pro are mostly governance and trust features, not monitoring features.
What Enterprise adds over Pro:
- SAML / SSO integration with the corporate identity provider (Okta, Azure AD, Google Workspace), so engineers do not authenticate against a separate Grafana Cloud login. This is the single most common gate-keeper requirement from internal IT before a Grafana Cloud rollout is approved beyond the platform team.
- Fine-grained role-based access control across organisations, folders, teams, and individual dashboards. Pro role assignment is folder-level at best; Advanced supports the per-resource control that compliance reviews look for.
- Audit logs of admin and editor actions exportable to a SIEM. Required for SOC 2 Type II evidence collection for many organisations.
- Enhanced support SLA with named technical account management, response-time guarantees, and access to escalation paths beyond the standard ticket queue. The exact response-time commitment is negotiated in the contract.
- Premium and enterprise-only data source integrations (Snowflake, Databricks, ServiceNow, Splunk, Oracle, SAP HANA, and similar). These are not available on the Pro tier and are often the line-item reason a team upgrades.
- Mutual TLS, customer-managed encryption keys, IP allow-listing, and VPC-peering options for organisations with stricter network-isolation requirements than the default multi-tenant SaaS provides.
How Enterprise is priced: Grafana Labs does not publish per-unit list pricing for Enterprise beyond the entry point: contracts start at a $25,000 per year spend commitment. Pricing is negotiated based on total committed usage (active series, GB of logs / traces / profiles, user count) plus the feature bundle the customer requires. The published per-unit Pro rates ($6.50 per 1,000 active series, roughly $0.45 per GB of ingest on logs / traces / profiles, $8 per active visualization user) are the reference baseline; the Enterprise quote folds those usage charges into an annual commit with discount, plus the enterprise feature bundle. There is no canonical "Enterprise is X% more than Pro" multiplier - the answer depends entirely on negotiated volume and which feature bundles you take.
What to ask for in an Enterprise quote:
- Annual commit discount on the usage-based components (series, log GB, trace GB). 15-30% is common at meaningful volume; ask explicitly.
- Burst protection on series ingestion so a runaway cardinality bomb does not blow the annual commit in a single quarter. Get the per-component overage rate in writing.
- Named-account technical contact and a defined escalation path for production incidents, not just a ticket queue.
- Data residency commitment (US, EU, or APAC region) if you have a regulatory requirement, and the contractual language that data does not leave that region.
- A documented migration path to Grafana Enterprise self-managed if compliance or cost considerations later push you off SaaS. Avoiding lock-in is a real negotiation lever.
When Enterprise is overkill: if the only thing you need from it is one premium data source connector, ask whether that connector is now available on the Pro tier - the line between Pro and Enterprise shifts as Grafana Labs rolls features down. For small teams (under 25 engineers) without SOC 2 obligations, Pro is almost always the right answer; below the $25,000 per year commitment floor, Enterprise procurement overhead exceeds the value.
Cost reduction levers
Three ways to cut a Grafana Cloud bill
Drop labels at scrape time
Recording rules for hot paths
Loki structured metadata
Verify before you buy
Cross-references
Related pages
/datadog-pricing
Datadog pricing breakdown
/new-relic-pricing
New Relic pricing breakdown
/datadog-vs-grafana-cloud
Datadog vs Grafana Cloud
/new-relic-vs-grafana-cloud
New Relic vs Grafana Cloud
/datadog-vs-prometheus-grafana
Datadog vs Prometheus + Grafana
/comparison
Six-vendor comparison
/calculator
Multi-vendor cost calculator
/open-source-vs-paid
Open source vs paid TCO
/kubernetes-monitoring
Kubernetes monitoring cost mechanics