Datadog is expanding into AI training
- Datadog runs a usage-based SaaS model, which means customers pay as they monitor more hosts or index more data.
- The core story is land and expand: customers start with one tool, then add more products over time.
- Large customers matter most, with about 4,550 customers over $100,000 in ARR making up 90% of ARR as of March 31, 2026.
- Multi-product use is rising fast, with 35% of customers using six or more products as of March 31, 2026.
- The main debate is whether AI training adds a new growth leg, or creates more usage swings from customers that grow fast and then optimize spending.
AI adds a new test
Datadog is a high-quality cloud software business with a clear expansion engine. Its dollar-based net retention rate, which measures how much existing customers spend after upsells and churn, was in the low-120% range as of March 31, 2026. That means the existing customer base is still spending more over time.
The strongest sign is product depth. As of March 31, 2026, 35% of customers used six or more products, 20% used eight or more, and 11% used ten or more. That makes Datadog harder to replace because it becomes part of how developers, operations teams, and security teams work each day.
The new twist is AI training. Management said training workloads, not only inference workloads, are becoming a real market for Datadog. The company landed deals with 2 of the world's biggest AI research teams and said hyperscalers are using Datadog in super intelligence labs to monitor training runs and GPUs.
The stock still carries a price problem. The business is strong, but valuation is not cheap. If AI demand wobbles, if customers cut cloud spending, or if hyperscalers bundle their own tools more aggressively, investors may not give Datadog much room for mistakes.
Paid by cloud usage
Datadog sells subscriptions to its cloud-based observability and security platform. Subscription terms are mostly monthly or annual. Revenue is tied to usage, mainly the number of hosts monitored or the amount of data indexed.
This model can grow quickly when customer workloads grow. It can also slow quickly when customers optimize, which means they look for ways to reduce data volume, monitored hosts, or cloud costs. That is why Datadog's usage trends matter as much as customer count.
The sales motion is simple: land with one product, prove value fast, then expand into more products. This is working best in large accounts. As of March 31, 2026, customers with more than $100,000 in ARR represented 90% of total ARR, up from 88% a year earlier.
Datadog also sells through a direct sales force and cloud-provider marketplaces. That helps reach cloud buyers, but it keeps the company tied to the same hyperscalers that can also compete with it.
One platform, many hooks
Infrastructure monitoring
This tracks servers, containers, databases, and cloud resources. It is often an easy first Datadog product for engineering teams.
Application performance monitoring
APM helps teams find slow or broken parts of an app. It is core to Datadog's value because it links app health to the rest of the tech stack.
Log management
Log tools collect and search machine data from apps and systems. This can scale with usage, but it is also an area where customers may optimize data volumes.
User experience monitoring
These tools show how real users or test users experience an app. They help Datadog reach teams that care about uptime, speed, and customer impact.
Cloud security
Security products extend Datadog from monitoring into finding risks and threats. This supports the multi-product expansion story.
GPU monitoring and LLM observability
These products target AI workloads, including training runs and large language model apps. The open question is how much they can add to total growth.
Cloud Prem and bring-your-own-cloud
These offerings let Datadog run on customer infrastructure. They are meant to serve data residency and sovereign AI needs.
Mostly North America
Datadog reports one operating segment. The mix below uses customer billing geography for the three months ended March 31, 2026, when revenue from outside North America was about 28% of total revenue, down from 30% a year earlier.
What could break
AI usage swings
High impact · Medium oddsAI-native customers have shown a pattern of fast usage growth followed by optimization. This cohort represented high single digits of year-over-year revenue growth for the quarter ended March 31, 2026, so even small swings can matter. AI training may help, but it may also make usage harder to predict.
Hyperscaler bundling
High impact · Medium oddsDatadog competes with AWS, Azure, and Google Cloud, plus their native monitoring tools. Those cloud providers can bundle tools into broader contracts. Datadog also relies on cloud ecosystems for distribution and infrastructure, which makes the relationship both helpful and risky.
Expansion slows
High impact · Medium oddsThe model depends on customers using more Datadog products over time. If dollar-based net retention falls out of the 120% range, the core thesis weakens. A slowdown in six-plus, eight-plus, or ten-plus product adoption would be an early warning.
Cloud cost and margin pressure
Medium impact · Medium oddsDatadog hosts its platform on third-party cloud infrastructure. Higher cloud infrastructure costs have already been called out as a margin pressure in prior filings. New AI products may require more investment before they reach full margin scale.
Security or privacy failure
High impact · Low oddsDatadog handles sensitive operational and security data for customers. A major breach could hurt trust, create legal costs, and slow sales. Global privacy rules also add compliance risk as the company grows outside North America.
In one breath
What does Datadog do?
Datadog sells software that helps companies watch their cloud apps and systems in real time. It covers infrastructure, app performance, logs, user experience, and cloud security.
How does Datadog make money?
Datadog sells subscriptions to its SaaS platform. Many subscriptions are usage based, so revenue rises when customers monitor more hosts or index more data.
Why does AI matter for Datadog?
AI workloads create complex systems that need monitoring, especially GPUs and training runs. Management now says AI training is becoming a market for Datadog, after earlier focusing more on inference.
What is the biggest risk for Datadog investors?
The biggest risk is that usage growth slows while the stock still prices in strong growth. AI-native customers can grow fast, then optimize spending, which can make revenue less predictable.