Let’s do some free association. I say “yellow”, you think… “sun”, maybe? It varies person to person. Now let’s do one that cuts across all people and cultures. I say “Datadog”, you think “expensive”, obviously.
It may surprise you that Datadog is not actually profitable, though. Take Richard Artoul, co-creator of Datadog’s storage engine, who tweeted:
November 2, 2026TIL Datadog has 4% net margins
wtf
wtf indeed. Is Datadog not making money?
No. Datadog makes a lot of money. In 2025, they brought in $3.4 billion.
Then where does all that revenue go? Since Datadog is public, we can answer.
It’s not the infra
Let’s just get this out of the way. Datadog is not spending all its money on infra. I estimate its infra spending at only ~20–25% of revenue.1 This is normal. Datadog’s gross margin is ~80%2, which is better than peers like Snowflake and Cloudflare.
That’s not to say that infra is cheap. If you naively throw everything on SSDs, storage costs will destroy you. Datadog puts tremendous effort into optimizing its infra. Most analytics and observability databases have converged on one architecture, which I covered in The New State of the Art in Observability. (TLDR: use object storage and make your workers stateless.) It’s not easy, but it’s possible.
So, it’s not the infra. It’s people.
Datadog has a giant staff
If running Datadog costs ~20–25% of its revenue, where does the rest go? People. 8,000 of them: 3,900 in R&D, 3,600 in sales/marketing, and a few hundred in other roles.
The salaries and stock grants dwarf infra spending.
I think Datadog has a large staff for two main reasons:
- they optimize for growth
- Datadog performs well, so there is little investor/analyst pressure to downsize
Investors expect Datadog to focus on growth
If you or I were starting a new company, we’d never dream of using Datadog. It’s just hilariously expensive. (Maybe you’d use Telemetry Machine, which isn’t? 🥺)
You’d be forgiven for thinking Datadog’s not growing. But it is, at like 30% a year.
When a company is growing this fast, investors don’t expect it to be profitable. Think of it like this: if Datadog stopped paying its employees, it’d be able to give its income to investors. But where would the investors put the money? Into a quickly growing company.
So, right now, investors have no reason to ask Datadog to downsize. That will surely change when Datadog’s growth slows. Just look at Dynatrace. After its growth slowed, activist investor Starboard Value started arguing that it should cut its sales/marketing org.3
Wait, but how does “optimize for growth” translate into “hire a bunch of people”?
Datadog loves enterprise companies
To answer that, first consider who’s buying Datadog. Giant enterprises. >90% of Datadog’s revenue comes from accounts paying >$100k a year. Just 34 customers paying >$10 million a year account for at least 9% of its revenue. One customer (a common guess is OpenAI) moves Datadog’s numbers so much that Datadog’s CFO has to discuss it on earnings calls.4
Enterprise companies love salespeople
Now I can tell you what all the fine people at Datadog do.
I already railed about this in Observability Doesn’t Have to Be Expensive, but if you want to win giant accounts that spend millions of dollars, you need a giant sales staff.
Why so many? Datadog’s own 10-K explains it. Before a big company buys, it “may require considerable time to evaluate and test our solutions and those of our competitors”. Then big customers “often begin to deploy our products on a limited basis, but nevertheless demand configuration, integration services and pricing negotiations.” Each of those takes people: sales engineers, solutions architects, account executives.
To be clear, the size of its sales staff is normal for B2B SaaS. What’s unusual is R&D: 45% of Datadog’s revenue goes to R&D (nearly twice Cloudflare’s share).
Enterprise companies love new products
The sales/marketing team lands the giant customers, but the R&D staff squeezes every penny out of them.
Datadog is expensive not just because its billing model is crazy (it’s fucking crazy – ask your local SRE how gross they feel running giant Kube nodes to fit into Datadog’s per-host billing), but because it keeps adding tempting new products.
Until 2017, it sold one product (infrastructure monitoring). It sells 26 now. LLM Observability! SIEM! APM! Cloud Cost Management! Synthetic Monitoring! Feature Flags!
It’s working. 61% of its customers use 4+ products, and 16% use 10+. SaaS companies talk a lot about churn. Datadog has negative churn: after accounting for the customers it loses, it would keep growing even if it stopped onboarding new customers.
I’m not merely speculating, by the way. Adding new SKUs, CEO Olivier Pomel told analysts in 2021, “is a way basically to maintain price and increase price as we grow and as we ship that functionality.”5
Summing up the narrative
So, why isn’t Datadog that profitable? I think it goes like this:
- Datadog still has a lot of room to grow. Investors want it to put its revenue towards growth.
- The “room to grow” is mostly giant companies.
- Winning and supporting giant companies takes a large sales/marketing staff.
- Giant companies have a lot of money, so Datadog needs lots of R&D staff to build products that get more money from them.
Datadog’s investors want it to keep pumping its revenue into the staff that finds new customers and squeezes more out of them, as long as it keeps growing. Until, of course, it doesn’t.
Coda: why is Datadog so expensive in the first place?
After all this analysis, it’s tempting to say that Datadog is expensive because they have to pay for all those people (and infra). But that’s not how B2B SaaS pricing works. The way it works is that you throw a price out there and see if people pay. You raise and lower it and see how customers react.
This is exactly what Datadog did: they planned to launch infrastructure monitoring at $12 per host per month. The night before, they went with $15 instead.6
There are obviously many enterprises still out there willing to pay a lot. There are even more companies that aren’t. If you are a savvy SRE, engineer, or CTO who wants amazing observability without paying horrifying prices, give Telemetry Machine a spin. If you are curious about our philosophy, I wrote about what a company built from the ground up for you looks like.