Datadog has spent the past decade and a half building what is quietly the most complete software observability franchise on the planet — and the AI era is turning out to be its moment. Founded in 2010 in New York City, the company now ships twenty-six revenue-generating products spanning infrastructure monitoring, application performance, security, and increasingly the machine-learning stack itself. The first quarter of fiscal 2026 marked a milestone: revenue crossed the billion-dollar mark for the first time, growing 32% year over year and accelerating for the fourth consecutive quarter. That is the signature of a company whose platform is compounding, not merely growing.
The compounding shows up in the land-and-expand mechanics. Datadog now counts five products with more than $100 million in annual recurring revenue, three more in the $50-to-100-million band, and eighteen earlier-cycle products still in their infancy. Usage depth tells the same story: 56% of customers use four or more products, 35% use six or more, and 20% use eight or more. Net revenue retention sits in the low 120s, gross retention in the mid-to-high 90s, and Datadog has captured roughly 14% of the Gartner ITOM observability market — against a field that includes Cisco's Splunk, Dynatrace, and New Relic, not to mention the open-source gravity of Grafana and the ever-present threat of AWS CloudWatch. The moat is not any single product; it is the breadth of a platform that makes leaving progressively more painful the deeper a customer goes.
The AI opportunity is where the thesis gets interesting. More than 6,500 customers — representing about 80% of Datadog's ARR — now use its AI integrations, and the company counts 22 AI-native customers spending over $1 million annually, with five above $10 million. The product surface is expanding in lockstep: LLM Observability for tracing prompts and model behavior, Agent Observability for the autonomous-agent wave, GPU Monitoring for the infrastructure layer, and the Bits AI suite of copilots that turn telemetry into answers. At DASH 2026 in June, Datadog announced more than 100 product launches, including autonomous Bits AI agents that investigate incidents on their own. Days later it closed the acquisition of Adaptive ML, adding reinforcement-learning operations — RLOps — that let AI teams fine-tune models the same way they monitor production. Datadog is positioning itself as the control plane for AI applications, not merely an observer of them.
The financials support the story. Roughly 4,550 customers generate more than $100,000 in annual recurring revenue, up 21% year over year, and that cohort now represents about 90% of total ARR — a base of large, sticky accounts that keeps expanding. Free cash flow came in at $289 million for the quarter, a 29% margin, on $335 million of operating cash flow, while remaining performance obligations hit $3.48 billion, up 51% year over year — a forward indicator that revenue growth has room to run. Management raised full-year guidance to $4.30–4.34 billion, implying 25–27% growth, on non-GAAP operating income of $223 million and a gross margin just above 80%. The balance sheet is fortress-like: $4.76 billion in cash against $1.29 billion in debt.
None of this means the stock is cheap, and the market has said so. Shares trade at a premium multiple in the high twenties on sales — a valuation that drew downgrades from Bernstein and Jefferies in July even as the business itself kept compounding. The competitive picture carries one genuinely open question: OpenTelemetry, the open standard for telemetry data, is weakening the historical lock-in that made Datadog's premium pricing stick. The bull case is that Datadog converts the standard into a distribution advantage; the bear case is that it commoditizes the moat. That tension makes the next earnings report — due the morning after this piece publishes, with second-quarter revenue guided to $1.07–1.08 billion — the most consequential Datadog print in years.
There are quieter tailwinds worth noting. FedRAMP High certification opens the federal door, and a new UK data center deepens the European footprint at a moment when sovereign-cloud requirements are tightening. Datadog's own State of AI Engineering report found that 69% of organizations now run three or more AI models in production and that roughly 5% of AI requests fail — failures that, in a production AI world, are exactly the kind of problem observability was invented to solve. The irony of the AI boom is that it makes software harder to understand, and harder software is Datadog's core business. For investors willing to look past a demanding multiple, the compounding platform underneath it remains one of the cleanest ways to own the AI application layer — with the caveat that a premium valuation leaves little room for execution error.
Disclosure: The Signal holds no position in DDOG. Positions may change. This is not financial advice.
