Guides / llm cost in datadog
How to see LLM cost in Datadog
Datadog answers LLM cost twice. LLM Observability monitors token usage and cost across agents in real time, and Cloud Cost Management ingests OpenAI as a SaaS spend source and forecasts it. Culpa, a local-first LLM cost, margin, and forecast ledger, prices each call and sets it against the revenue the customer who triggered it pays you.
Why this happens
The common assumption about general observability platforms, that they see an LLM call as an HTTP span and lose the cost, is out of date for Datadog and this page won't repeat it. Its LLM Observability product says it monitors latency, token usage and cost across all AI agents in real time, with span-level inputs and outputs across prompts, retrieval, tool calls and decisions. Separately, Cloud Cost Management lists AI Costs alongside Snowflake and Databricks, with OpenAI named as an ingested spend source, plus reporting, budgets and spend forecasting. So Datadog can tell you what a call cost and what the OpenAI invoice will look like. The gap is between those two answers rather than in either of them. One is built from your traces and the other from a vendor bill, they live in different products with different pricing, and neither carries what your customer pays you, so neither can divide one by the other. Worth knowing before you build on it: Cloud Cost Management Pro is priced at $5 per $1,000 of cloud and SaaS spend per month, so the cost of watching scales with the spend being watched.
What this usually looks like
- You can see cost per agent in one product and the OpenAI invoice in another, and nobody has joined them.
- Your spend forecast covers the vendor bill and says nothing about which customer drove it.
- The observability bill grew because you instrumented more, not because you spent more on models.
- A cost spike is visible and the customer behind it takes a day to find.
- Nobody can say whether your largest AI customer is profitable.
Free, no card, no account
Run the free Cost Leak Scan
It shows your most expensive conversation before you install anything.
Mistakes that cost the most
| Mistake | Why it hurts | Do instead |
|---|---|---|
| Assuming an APM can't track LLM cost. | Datadog's LLM Observability tracks token usage and cost per agent in real time, so this argument loses on the facts. | Check what your existing tooling already does before buying anything to fill a gap it doesn't have. |
| Reading a Cloud Cost Management forecast as a unit-economics forecast. | It projects an ingested vendor invoice, which has no customer, feature or prompt version in it. | Forecast from your own per-call ledger if you need the answer broken down by anything. |
| Ignoring what the cost tooling itself costs. | Cloud Cost Management Pro is published at $5 per $1,000 of cloud and SaaS spend, so it scales with what it watches. | Price the cost tool against the spend it covers before assuming it pays for itself. |
| Instrumenting every span because storage feels free. | Datadog bills ingestion and custom metrics, so richer LLM traces raise a bill that has nothing to do with your model spend. | Decide what each span is for, and price the telemetry as a line of its own. |
Run this check tonight
- Open LLM Observability and find cost for one agent, then open Cloud Cost Management and find the OpenAI line.
- Try to reconcile the two, and note how long it takes.
- Ask what your top AI customer costs you and what they pay you, and see which system holds each half.
- Price your Cloud Cost Management tier against the spend it watches.
What the cost tool costs
Datadog publishes Cloud Cost Management Pro at $5 per $1,000 in cloud and SaaS spend per month, read from its pricing page on 2026-08-03. Take a team with $40,000 a month of combined cloud and AI spend and price the watching against the watched. The spend figure is modelled and the rate is published.
Half a percent is a fair price for cost visibility and it still costs, and it scales with the thing it measures rather than with the work you do. Worth putting in the same model as the spend itself, which almost nobody does.
Every number, with its confidence and source
| Figure | What it means | Confidence | Source |
|---|---|---|---|
| $200.00 per month | Datadog Cloud Cost Management Pro on a modelled $40,000 of monthly cloud and SaaS spend | calculated | $5 per $1,000 in cloud and SaaS spend per month, published on datadoghq.com/pricing and read 2026-08-03, applied to a modelled $40,000 monthly spend: 40 x $5.00 = $200.00, which is 0.5% of the spend covered. The rate is published and the spend figure is modelled. |
What a generic answer can’t know
This page has to concede more than most in the cluster. Datadog tracks LLM cost properly, forecasts spend, and publishes on-prem options, so neither cost tracking nor forecasting nor local-first is what separates the two. One thing does. Datadog holds telemetry and invoices, and neither contains what your customer pays you, so no query it can run divides cost by revenue. Culpa carries that column, prices each call from a versioned price book in exact decimal, and keeps the ledger on your own infrastructure, so cost per customer becomes margin per customer and the forecast is built from your own per-call history rather than from an ingested vendor bill. Run Datadog for the system. Run the ledger for the economics.
Questions founders ask next
Does Datadog track LLM costs?
Yes, and this page says so plainly rather than implying otherwise. Datadog LLM Observability states it monitors latency, token usage and cost across all AI agents in real time. Separately, Cloud Cost Management lists AI Costs with OpenAI named as an ingested SaaS spend source.
Can Datadog forecast my AI spend?
Cloud Cost Management publishes spend forecasting, budgets and anomaly detection, and its free tier is limited to Datadog's own costs while paid tiers cover cloud and SaaS spend. That forecast is built from an ingested vendor invoice, so it projects the bill rather than the economics underneath it.
What does Datadog's cost tooling cost?
Cloud Cost Management Pro is published at $5 per $1,000 in cloud and SaaS spend per month, which works out at 0.5% of the spend it reports on. A free tier exists, limited to Datadog costs only. Read from its pricing page on 2026-08-03.
So why would I run Culpa alongside Datadog?
For the one question neither of Datadog's two cost products can reach. Both work from telemetry or from invoices, and neither holds the revenue a customer pays you, so cost per customer stays a cost rather than becoming a margin.
On your infrastructure
Culpa runs on your infrastructure. Your prompts and responses never leave it. Culpa counts calls to run your plan, and it fails open, so if it ever breaks your app keeps running.
How Culpa works
Find the culprit. Not just the total.
Your dashboard shows what you spent. It stops short of who spent it. Culpa shows the conversation, the user and the feature behind it.
Your prompts stay local.
Culpa runs on your own infrastructure. What you send to a model reaches us at no point.
Every dollar has a name.
Follow any charge to the conversation, the user, the feature and the customer behind it.
See the bill before it lands.
Cost your next feature before you ship it. You get the likely bill and the worst case, at best, median, p90 and p99.
Three steps to your first answer.
Change one base URL.
Or drop in the Python or TypeScript library.
Find your most expensive conversation.
In the first session, not the first week.
Cost your next feature before you ship it.
Why the bill went up
Example dashboardCalls traced
418,209
across 3 projects
Spend this week
$378.41
+ $182 vs last week
Failed calls
312
74% retried, and you paid for all of them
+ $182 this week traced to one culprit
Spend over 14 days
Most expensive users
Next week forecast
Graded against reality. Accuracy shown as results land.
Keep reading
Sources: Datadog LLM Observability, Datadog pricing. Last reviewed 2026-08-03, rates effective 2026-07-02. Plain text version.