OpenTelemetry (OTel) was supposed to standardize observability and kill vendor lock-in. In 2026, OTel is the standard... but vendors have adapted.
OpenTelemetry Successes:
- CNCF's 2nd most active project behind K8s
- Native support in all major languages (auto-instrumentation improving rapidly)
- Collector architecture enables flexible pipelines
- eBPF integration for zero-code instrumentation
The 2026 Reality:
- Most orgs use OTel for data COLLECTION
- But still send to commercial backends (Datadog, Honeycomb, Grafana Cloud)
- Vendors now compete on analysis, AI-powered insights, and UI
- Open source backends (Jaeger, Prometheus, Tempo, Loki) for cost-sensitive teams
Vendor Landscape 2026:
- Datadog: Most complete, most expensive, best for unified observability
- Honeycomb: High-cardinality champion, best for debugging complex systems
- Grafana Cloud: LGTM stack (Loki, Grafana, Tempo, Mimir), cost-effective
- New Relic: Reinvented with better pricing, strong APM
- Chronosphere: For truly massive scale (DoorDash, Snap, Uber)
Debate Questions:
- Is OTel actually reducing vendor lock-in or just shifting the lock-in point?
- What's your observability stack cost vs value?
- Are you using eBPF? What's the experience like?
- Logs vs metrics vs traces - what's your priority?
What's your 2026 observability setup? 📊