Dynatrace pricing 2026: how DPS actually bills

Dynatrace publishes its pricing per host-hour, which suggests a per-host meter analogous to Datadog. The reality is more nuanced. Underneath the per-host headline is the Davis Platform Subscription, a consumption-based pool that customers commit to in advance. Each product (Full-Stack, Infrastructure, Application Security, Log Management, Synthetics, Business Analytics) draws down from the pool at a published capability-unit rate. The host-hour pricing is one way to express the draw rate for the most common product (Full-Stack Monitoring); the underlying meter is units, not hosts.

The practical consequence is that estimating a Dynatrace bill from list pricing alone underestimates two things and overestimates one. It underestimates flexibility (the customer can shift consumption between products without renegotiating, which is genuinely useful). It underestimates total cost when usage exceeds the pool (overage rates apply). It overestimates the rate the customer actually pays at scale (enterprise discounts on multi-year DPS commitments are routinely 30 to 50 percent below published list).

The host-memory banding is the second source of confusion. A standard 8 GB host pays the published rate. A 16 GB host pays roughly twice. A 32 GB host pays roughly four times. Customers running large database hosts or container hosts with high memory allocation discover that the bill scales with memory rather than logical host count, which differs from Datadog's flat-rate model.

For log management, Dynatrace shifted in 2023 from a per-byte indexing model to a query-time billing model with hot/warm/cold storage tiers. Logs in hot storage are immediately queryable; logs in cold storage cost less to store but bill per query when accessed. This is structurally cheaper for archival workloads (compliance retention, audit trails) and structurally similar to Datadog Flex Logs for searchable workloads.

Dynatrace is the strongest fit for three customer profiles. The first is the large enterprise (1,000+ hosts) running a heterogeneous stack across on-premises VMs, Kubernetes, mainframe, and managed cloud services. OneAgent's auto-discovery is genuinely best-in-class for environments where the customer cannot mandate a specific instrumentation library. A bank with 200 different application teams, half on COBOL, half on Spring Boot, all needing single-pane observability, is the canonical Dynatrace use case.

The second is the regulated industry customer (banking, healthcare, pharma) that needs platform-level certifications (FedRAMP High, HIPAA, PCI DSS at the platform level) and is willing to pay the premium for the certification overhead. Datadog and New Relic also carry these certifications, but Dynatrace's enterprise-led sales motion and 24x7 dedicated support tend to land better with regulated procurement.

The third is the AI-led observability customer that values Davis as a meaningful differentiator. Davis has been investing in causal AI and root-cause inference since 2018, longer than any peer in the market. For incident response teams that need automated root-cause attribution rather than dashboards-as-investigation, Davis is the strongest argument.

Dynatrace is less suited to teams below 100 hosts (the per-host rate becomes uncompetitive without the enterprise discount), teams that want a free or freemium starting point (Dynatrace has only a 15-day trial, no permanent free tier), startups that prefer self-service procurement (sales-led model is mandatory above the trial), and teams with deep Prometheus/OpenTelemetry expertise that prefer the open-source-aligned vendor (Grafana Cloud is the alternative).