Published October 23, 2026
Observability platforms collect the telemetry that applications and infrastructure emit, mainly metrics, logs and traces, and let engineers query it to find out why a system is slow or failing. Application performance monitoring (APM) products use distributed tracing, which follows requests as they pass through the services of a distributed system. Platforms are sold as hosted services, as open-source software to run yourself, or both.
Where the market stands
The Cloud Native Computing Foundation (CNCF) announced the graduation of OpenTelemetry on May 21, 2026. OpenTelemetry standardizes how metrics, logs and traces are collected, processed and exported. Its profiles signal is in alpha.
Four vendors changed owners:
- Francisco Partners and TPG completed their acquisition of New Relic on November 8, 2023, at $87.00 per share in cash, about $6.5 billion in equity value.
- Cisco completed its acquisition of Splunk on March 18, 2024, at $157 per share in cash, about $28 billion in equity value.
- Palo Alto Networks completed its acquisition of Chronosphere on January 29, 2026. It plans to integrate its Cortex AgentiX product with Chronosphere's platform. Chronosphere Telemetry Pipeline remains available as a standalone product.
- Snowflake acquired Observe on February 2, 2026. Snowflake said the deal sets up an observability architecture based on Apache Iceberg and OpenTelemetry.
What OpenTelemetry does and does not do
OpenTelemetry generates, collects and exports telemetry. It is vendor-neutral and works with many backends. It is not a backend itself: storing and displaying the data is left to other tools. Its Collector receives, processes and exports telemetry. Its wire protocol is OTLP.
Each platform in this edition documents OpenTelemetry support:
- Datadog accepts OTLP directly, from a standard Collector, or through the DDOT Collector embedded in the Datadog Agent.
- Dynatrace accepts OTLP directly, from a standard Collector, or from its own Collector distribution. Dynatrace says frontend data sent through OpenTelemetry SDKs does not get its RUM-specific features, such as user sessions and Session Replay.
- New Relic recommends native OTLP ingest as its preferred method.
- Honeycomb receives OTLP.
- Splunk ships a distribution of the Collector based on the Contrib version. Grafana Alloy is a Collector distribution with built-in Prometheus pipelines.
- Elastic ships OpenTelemetry distributions tailored to Elastic.
Dashboards, alerts and saved queries are built in each vendor's own tools. Plan to rebuild them when the backend changes.
Four ways to buy
- Hosted suites priced per host and per product. Datadog and Splunk Observability Cloud charge per monitored host per month, with APM, logs and other products charged separately or bundled in tiers.
- Hosted platforms priced by consumption. Dynatrace bills memory-hours, host-hours and data against an annual commitment. New Relic bills data ingested plus users, or compute. Grafana Cloud bills active metric series and GB. Elastic Cloud Serverless bills GB ingested and retained. Honeycomb bills events and metrics data points.
- Open-source software you run yourself. Grafana, Loki (logs), Tempo (traces), Mimir (metrics) and Pyroscope (profiles) are licensed under the GNU AGPLv3. You carry the hosting, upgrades and on-call.
- Telemetry pipelines in front of any backend. Datadog Observability Pipelines runs in your own infrastructure and filters low-value logs before they are stored. Chronosphere Telemetry Pipeline filters logs, removes fields and routes data to platforms such as Datadog and Splunk, or to Amazon S3 and Azure Blob storage.
What each platform costs
List prices as each vendor published them on October 17, 2026.
- Datadog. Infrastructure Pro is $15 per host per month billed annually, or $18 on demand. Enterprise is $23 billed annually, or $27 on demand. APM is $31 per host per month with infrastructure, billed annually. Log ingest is $0.10 per GB. Indexing with 15-day retention is $1.70 per million events billed annually, or $2.55 on demand. On the high-water mark plan, hosts are counted hourly, and the bill uses the highest count after the top 1% of hours is dropped.
- Dynatrace. The Dynatrace Platform Subscription is a minimum annual spend commitment drawn down at rate-card prices. Dynatrace says larger commitments get larger discounts. Full-Stack Monitoring is $0.01 per memory-GiB-hour, or $58 a month for an 8 GiB host. Infrastructure Monitoring is $29 per host per month. Log ingest is $0.20 per GiB, with retention billed separately. Real user monitoring is $2.25 per 1,000 sessions.
- New Relic. The first 100 GB of data a month is free. After that, data costs $0.40 per GB, or $0.60 per GB for Data Plus, which adds up to 90 days of extra retention for most data types. On Standard, full platform users cost $10 a month for the first and $99 for each additional user, up to 5. On Pro they cost $349 a month each with an annual commitment, or $418.80 paid monthly. Pro and Enterprise can instead use compute pricing with no user fees. No compute rate is published.
- Splunk Observability Cloud. Infrastructure starts at $15 per host per month billed annually, App & Infra at $60 and End-to-End at $75. Only End-to-End includes real user monitoring and browser synthetics.
- Grafana Cloud. The free tier covers 10,000 active metric series, 50 GB each of logs, traces and profiles, 14-day retention and 3 active Grafana visualization users. Pro has a $19 monthly platform fee. Metrics start at $6.50 per 1,000 active series. Logs start at $0.05 per GB processed, $0.40 per GB written and $0.10 per GB retained. Enterprise has a minimum commitment of $25,000 a year.
- Elastic Cloud Serverless. The Complete tier starts at $0.09 per GB ingested and $0.019 per GB retained per month for logs and traces. Metrics ingest starts at $0.023 per GB under pricing effective July 1, 2026. Logs Essentials starts at $0.07 per GB ingested.
- Honeycomb. Free up to 20 million events a month. Pro starts at $150 a month for 50 million events, up to 750 million. Seats are unlimited.
What makes the bill grow
In Prometheus, every unique combination of label names and values is a new time series. Prometheus advises against labels with many different values, such as user IDs. Datadog counts a custom metric as each unique combination of metric name, host and tags. Grafana Cloud bills per 1,000 active series. Under both, one tag with many values multiplies the count.
Traces are controlled by sampling. Head sampling decides early, without seeing the whole trace. Tail sampling decides after seeing all or most of a trace's spans. The components that do it must store a large amount of data.
Set sampling, log filtering and per-team budgets before broad rollout. Model the bill at the data volumes you expect in three years, not at the pilot's.
What a change of platform requires
Move instrumentation to OpenTelemetry first, through a Collector you control. The backend can then change without touching application code.
Run the old and new platforms side by side on a few critical services. Compare alerts during real incidents and the projected bill at full volume. Rebuild dashboards, alerts and runbooks, and train the on-call staff, before switching off the old platform.
Check which features need the vendor's own agent rather than OpenTelemetry, and what you lose without it.
Questions to settle first
- Which pricing unit matches how your estate grows: hosts, memory, GB, series, events or users?
- Which signals do you send today, at what daily volume, and which do you keep longer than 30 days?
- Will you instrument with OpenTelemetry, a vendor agent, or both?
- Who owns the telemetry pipeline and each team's budget?
- If the vendor has changed owners, what does the contract say about price at renewal?