Verdict up front: CloudWatch wins for single-cloud AWS shops under 10 hosts. Datadog wins for multi-cloud, container-heavy, or APM-required workloads past 10 hosts. Grafana + Prometheus (self-hosted or Grafana Cloud) is the middle option that consistently saves 60-80% vs Datadog at the cost of more DevOps work. Don't default to Datadog because everyone else does — audit your actual needs first.
Datadog real pricing in 2026
- Infrastructure Pro: $15/host/month (basic metrics, 15-month retention)
- Infrastructure Enterprise: $23/host/month (live process monitoring, network)
- APM: $31/host/month per APM host (separate from infra host count)
- Logs: $0.10 per million events ingested + $1.06 per million events for 7-day retention (more for longer)
- RUM: $1.50 per 10k sessions
- Synthetics: $5 per 10k API tests, $12 per 1k browser tests
For a typical 10-host startup with APM, 10GB logs/day, 100k sessions/month:
- Infrastructure: $150/month
- APM: $310/month
- Logs ingestion (10GB/day ≈ 30M events/day): $300/month ingest + $1,000/month retention = $1,300/month
- RUM: $15/month
- Total: ~$1,775/month = $21,300/year
The log retention bill is the surprise. Teams budget $200-300/month for Datadog, then discover their actual bill is 5-10x higher because logs are billed separately.
CloudWatch real pricing in 2026
- Metrics: $0.30 per metric/month, free for AWS-native metrics
- Custom metrics: $0.30/metric/month up to 10k, $0.10/metric past that
- Logs ingestion: $0.50/GB
- Logs storage: $0.03/GB/month (much cheaper than Datadog)
- Dashboards: $3/dashboard/month past 3 free
- Alarms: $0.10/alarm/month (standard), $0.30 (high resolution)
- Container Insights: $0.0072/instance/hour ≈ $5/instance/month
- Synthetics Canaries: $0.0012/run
For the same 10-host startup:
- Default AWS metrics: free
- Custom metrics (~500): $150/month
- Logs ingestion (10GB/day = 300GB/month): $150/month
- Logs retention (90 days = 900GB): $27/month
- Container Insights: $50/month
- Synthetics (1k runs/day = 30k/month): $36/month
- 20 alarms: $2/month
- Total: ~$415/month = $4,980/year
CloudWatch is 4x cheaper at this scale. You're losing some features — Datadog's APM is genuinely better than X-Ray, dashboards are nicer — but at $16k/year savings, the trade-off is real.
Where Datadog earns its price
- APM: distributed tracing, code-level latency breakdown, error tracking. AWS X-Ray exists but is significantly weaker.
- Multi-cloud: if you run on AWS + GCP + on-prem, Datadog is one pane of glass. CloudWatch is AWS-only.
- Out-of-box dashboards: 600+ pre-built integrations (Postgres, Redis, Kubernetes, Kafka). CloudWatch needs you to build these.
- Alert quality: Datadog's anomaly detection and forecast-based alerts are better than CloudWatch's static thresholds.
- Log search: Datadog's log search is genuinely fast (Elasticsearch-backed). CloudWatch Insights is slower for ad-hoc queries.
Grafana + Prometheus: the middle option
Self-hosted (1 EC2 instance for Grafana + Prometheus):
- EC2 t4g.small: $13/month
- EBS storage: $5/month
- Snapshots: $2/month
- DevOps time to set up: 1-2 weeks one-time
- Ongoing maintenance: 2-4 hours/month
- Total: ~$20/month + DevOps time
Grafana Cloud (managed):
- Free tier: 10k metrics, 50GB logs, 3 users
- Pro: $19/month base + $8 per active user + usage
- For typical 10-host setup: $200-400/month including logs
The trade-off: more DevOps work to set up dashboards, fewer pre-built integrations than Datadog. But the cost savings at scale are substantial.
Side-by-side
| Feature | Datadog | CloudWatch | Grafana Cloud |
|---|---|---|---|
| Cost (10-host startup) | ~$1,775/month | ~$415/month | ~$300/month |
| Hosts (basic metrics) | $15/host | Free (AWS metrics) | Free (10k metrics) |
| APM | $31/host (best) | X-Ray ($5/M traces) | Tempo (good) |
| Logs ingestion | $0.10/M events | $0.50/GB | $0.50/GB |
| Log retention | $1.06/M events for 7 days | $0.03/GB/month | $0.50/GB indexed |
| Multi-cloud support | Yes (AWS, GCP, Azure) | AWS only | Yes |
| Pre-built integrations | 600+ | AWS services only | 200+ |
| Anomaly detection | Yes (built-in) | Yes (limited) | Via plugins |
| Dashboards | Polished, fast | Functional | Industry standard |
| Setup time | 1-2 days | Continuous as you build | 1-2 weeks for full |
| Vendor lock-in | High | AWS-bound | Low (open source) |
When to choose Datadog
- 10+ hosts, multi-cloud, mixed workloads
- APM is critical (latency-sensitive product)
- Team doesn't have DevOps capacity to build/maintain monitoring
- You need 24/7 support contract with a SaaS vendor
- Annual budget for observability is $20k+ and you want best-in-class
When to choose CloudWatch
- AWS-only, under 10 hosts
- Simple metrics + logs needs (no distributed tracing required)
- You want monitoring on the same bill as the infra
- Cost matters more than dashboard polish
When to choose Grafana + Prometheus
- You have DevOps capacity to invest in setup
- You want to avoid vendor lock-in (Prometheus is the de facto standard)
- Multi-cloud or hybrid (on-prem + cloud)
- Cost is a key constraint and Datadog is out of budget
Datadog bill making your AWS bill look reasonable?
We audit Datadog usage, identify the 10% of metrics generating 60% of cost. Typical bill cut: 40-60% without losing visibility.
Book a discovery call