Elastic surveyed IT decision-makers across 15 countries for its 2026 observability report. The headline finding: 97% of organizations experienced unexpected monitoring and observability costs in the past year. Not a slim majority. Ninety-seven percent.
The number shouldn’t surprise anyone who has reviewed a Datadog invoice after adding a few microservices. But it reveals something important about how FinOps teams allocate attention. Most organizations have dedicated processes for optimizing compute, storage, and even SaaS licenses. Monitoring costs? Those get waved through procurement as “necessary infrastructure” and rarely see the same scrutiny.
The result: a mid-market company with 100 engineers and 50 services pays approximately $124,000 a year on monitoring and observability tooling. For enterprises, the figure climbs to $500,000 or more. And with AI workloads generating 40 to 200 percent more telemetry data than traditional applications, those numbers are accelerating.
What a $124,000 Monitoring Bill Actually Looks Like
A typical mid-market monitoring stack breaks down roughly like this, based on published vendor pricing at the 100-host scale:
| Category | Monthly Cost |
|---|---|
| Infrastructure monitoring | $1,150 |
| Container monitoring | $920 |
| Custom metrics | $1,450 |
| Log management | $1,875 |
| APM and distributed tracing | $3,275 |
| Real user monitoring and synthetics | $1,290 |
| Database monitoring | $350 |
| Total | $10,310/month ($123,720/year) |
That figure is before overages. And overages are the norm: Elastic’s survey found that 67% of organizations report cost surprises on a regular basis. Large enterprises with 20,000+ employees experience them at nearly five times the rate of smaller companies.
The problem is structural. Tools like Datadog price across multiple dimensions simultaneously: per host, per product module, per GB of log data, per million spans, per thousand user sessions. Each dimension scales independently. Your monitoring bill doesn’t grow linearly with your infrastructure. It grows faster.
Add a staging environment. Increase log retention from 7 to 15 days. Turn on database monitoring for three more clusters. Each of these reasonable operational decisions triggers a pricing multiplier that compounds across billing dimensions.
The Seven-Tool Stack Compounds the Problem
The $124K figure understates the real cost for many organizations because it assumes a single vendor. Most mid-market companies run five to seven monitoring tools simultaneously:
| Tool | Function | Monthly Cost |
|---|---|---|
| Datadog | Infrastructure, APM, logs | $5,490 |
| PagerDuty | Alerting and on-call management | $3,075 |
| Sentry | Error tracking | $442 |
| StatusPage | Status communication | $399 |
| Grafana Cloud | Dashboards and visualization | $299 |
| Loggly | Log aggregation | $349 |
| Pingdom | Uptime monitoring | $249 |
| Total | $10,303/month ($123,636/year) |
Whether you consolidate on one vendor or spread across seven, mid-market teams land at roughly the same $124K annual number. The multi-tool path adds hidden costs on top: approximately 12 integration failures per year (each requiring 2 to 8 hours to diagnose), context switching during incidents (10 to 23 minutes of cognitive reload per tool change), and vendor management overhead estimated at 100+ hours annually across compliance, procurement, and security reviews.
The total cost of ownership runs 2 to 3 times higher than sticker prices alone. If you’ve conducted a SaaS audit recently and didn’t include monitoring tools, you missed one of the largest line items in your operational SaaS spend.
AI Monitoring Breaks the Cost Model
Traditional infrastructure monitoring generates a predictable volume of telemetry. A web application running 50 services on 100 hosts produces a relatively stable number of metrics, logs, and traces month over month.
AI workloads destroy that predictability.
A single LLM-powered support bot processing 50,000 daily messages generates roughly 400 million tokens, 1 million spans, and 4 million custom metrics per day. Teams that add LLM observability to their existing monitoring stack report bill increases of 40 to 200 percent.
AI application monitoring requires tracking dimensions that didn’t exist two years ago. Token consumption per model. Latency per inference call. Prompt and completion ratios. Cache hit rates. Model routing decisions. Cost attribution per AI feature. Each of these dimensions hits a billing meter in your monitoring platform.
The 2026 State of FinOps survey confirms 98% of FinOps teams now manage AI spend. But most of those teams focus on the inference bill (the tokens consumed by the model) and overlook the monitoring bill (the telemetry generated by observing those tokens). As enterprises scale AI deployments, the monitoring cost footprint scales alongside them. Often faster, because each AI feature multiplies the number of dimensions your observability platform tracks and bills for.
Three Patterns That Cut Monitoring Spend 40 to 70 Percent
Elastic’s survey found 54% of IT leaders face increasing executive pressure to justify observability expenses, and 96% of organizations are actively implementing cost controls. The organizations that succeed tend to follow three patterns.
Consolidate the stack. Moving from a multi-tool collection to a unified platform delivers the largest savings. At the 150-engineer scale, the cost differences are dramatic: Datadog runs approximately $42,500/month, Grafana Cloud costs roughly $21,000/month, SigNoz comes in at $12,200/month, and Uptrace at just $2,000/month. Grafana Cloud’s open-source foundation (built on Prometheus, Loki, Tempo, and Mimir) offers full-stack monitoring at roughly half the cost of Datadog at equivalent scale. For teams comfortable with OpenTelemetry-native tooling, the savings are steeper. Consolidation alone delivers 40 to 70 percent direct cost savings before accounting for reduced integration maintenance and vendor management overhead.
Tier the monitoring by environment and criticality. Elastic’s survey found 41% of organizations now deploy lower-cost or open-source tools for less-critical systems. Production gets full APM, distributed tracing, and real user monitoring. Staging gets infrastructure metrics and log aggregation only. Development environments get self-service Grafana or Prometheus stacks with no vendor billing attached. Another 42% have begun deploying observability only to critical environments, skipping non-production entirely. The principle mirrors what FinOps teams already apply to cloud storage: not every workload deserves the premium tier.
Cap and route telemetry before it hits the billing meter. Log volume is the single largest cost driver in most monitoring stacks. Sampling high-volume, low-signal logs (debug output, health check pings, routine cron jobs) before ingestion can reduce log costs 20 to 40 percent without sacrificing visibility into the events that matter. Datadog’s Flex Logs, Grafana’s adaptive logs, and pipeline tools like Cribl and Observo all support this pattern natively. When paired with shorter retention windows for low-priority data, storage costs drop an additional 25 to 50 percent. The key is making these decisions before data enters the billing pipeline, not after.
Monitoring Belongs in Your FinOps Practice
In 20 years of managing IT operations budgets, I’ve watched monitoring costs treated as a rounding error, right up until they weren’t. The inflection point arrives when you add AI workloads, expand to multi-cloud, or reach the three-year mark on a vendor contract with annual price escalators baked in. By that point, the monitoring line item is larger than several compute categories, and nobody noticed because it was split across seven vendors and four cost centers.
The FinOps for SaaS movement is gaining traction because unmanaged SaaS spend is where the hidden money lives. The 2026 State of FinOps data shows 90% of FinOps teams now manage SaaS (up from 65% in 2025). Monitoring tools are one of the largest categories within operational SaaS spend, yet they sit in an awkward gap: too operational for the SaaS management team to claim, too SaaS-like for the infrastructure FinOps team to notice.
The fix is straightforward. Run a SaaS audit on your monitoring and observability stack. Total the actual spend across every tool that collects, stores, alerts on, or visualizes telemetry data. Compare that number to your compute spend. If monitoring exceeds 10% of your infrastructure bill (and for many mid-market teams it does), it deserves the same optimization discipline you apply to reserved instances and right-sizing.
Observability is not optional. Overspending on it is.