Datadog Grew Faster and Got Cheaper — The Discount Is the Story

Q3 accelerated to 35% growth, yet the stock sits 36% below its October peak, because the AI cohort powering the upside is the same one now renegotiating its price.

DDOG · Earnings Call · 2026-11-04

The quarter that accelerated into a drawdown

Datadog's Q3 was, on the numbers, its best operating quarter in years. Revenue of $1.197 billion grew 35% year over year, above the high end of guidance. Free cash flow was $251 million at a 21% margin. Customers above $100,000 in ARR rose to about 4,860 from 4,060 a year ago, and those customers now generate roughly 91% of ARR. Perhaps most important, “Revenue growth from our non AI customers accelerated for the 5th straight quarter was in the high-20s percent year over year.” — Olivier Pomel, Cofounder and CEO · 2026-11-04 That is a five-quarter acceleration in the unglamorous part of the business, the part that isn't a frontier lab. And yet the tape tells an entirely different story. Datadog shares are down 27.5% over the last 90 days, and sit 35.6% below the October 9 peak of $293.26. The entire decline came in a violent three-day, 20% swoop just before this report. A reader encountering only the price chart would assume something broke; a reader encountering only the transcript would assume everything is working. Both are true, and the gap between them is the whole investment question. Part of the answer is that the software was simply expensive. The company still trades at roughly 23.6x trailing revenue, and it ran up 86% in five weeks before the fall. But part of the answer is specific to Datadog, and management told us so — just quietly.

The largest customer, the early renewals, and the guidance gap

The operative phrase this quarter was minimum commitment. Datadog's single largest customer reduced usage during Q3, though it stayed above its contract floor. Management has decided to guide Q4 at that floor and nothing more: “our Q4 guidance contemplates their revenue at their minimum commitment.” — David Obstler, CFO · 2026-11-04 The implication, acknowledged on the call, is that any usage above the floor is an implicit beat. Q4 revenue guidance of $1.19–1.20 billion implies 25–26% growth — a visible deceleration from Q3's 35% — but excluding that one customer the guide implies low-to-mid-30s growth. Nothing has actually deteriorated; the optics have. The genuinely new wrinkle is early renewal. Several large AI customers renewed ahead of schedule, accepting higher dollar commitments in exchange for better volume pricing. This is mathematically good and sequentially awkward: the same traffic now bills at a lower rate, so revenue growth dips before it re-accelerates. Management has never had to explain this dynamic before, and it visibly strained the call. Analyst after analyst circled the same question — how do you model the sequentials? — and the CEO's honest answer was that these customers are scaling faster than anything the industry has seen, so "normal" is not a useful word.

they get better pricing... But, the volume is going. Kind of going crazy there, but it kind of creates a short term headwind because the same volume is now on a lower price. But then the volume grows.
David Obstler, CFO · 2026-11-04

Note what shifted in the language. In August, the CFO's posture was defensive: “we did chose to fully derisk our largest customer.” — Olivier Pomel, Co-Founder and CEO · 2026-08-06 This quarter the tone was that the renewals are "all good news" — and the keyword record agrees. The company's conservatism and growth trend keywords both fell off sharply in the latest quarter's ranking, replaced by mechanical words: minimum, commit, renewal, annualized. Management is no longer talking about caution. It is talking about arithmetic.

Bits AI Scout: Datadog quietly became a model builder

The most under-covered item on the call was not financial. Datadog trained its own reasoning model, Scout, using in-house reinforcement learning, and deployed it to power alert investigations. The claim is striking: “it performed on par or even better than frontier models. On alert investigations. At a fraction of the inference cost.” — Olivier Pomel, Cofounder and CEO · 2026-11-04 This is a software company deciding that generic frontier models are too expensive for the volume of machine-to-machine work observability requires, and building a narrow specialist instead. It shows up as fresh, company-unique vocabulary in the keyword set — AI research, open weight models, and inference cost — none of which were load-bearing themes even two quarters ago. It also reframes the training-versus-inference debate the company has narrated for a year. In May, the CEO was explicit that training was a new market: “last year, when we reported earnings, we said we're mostly interested in inference workloads and training is not a real market for us yet. Now we actually see training becoming a market.” — Olivier Pomel, Co-Founder and Chief Executive Officer (CEO) · 2026-05-07 By this quarter Datadog was booking an eight-figure annualized expansion with the AI division of a large technology company for AI training workloads, and notes GPU devices monitored up eightfold this year. The Infrastructure monitoring suite crossed $2 billion in ARR growing mid-30s — and the CEO was careful to say GPU monetization is largely not yet in that number.

What the market missed while watching the data-center complex

The sell-off in AI infrastructure has been brutal and broad — memory, optics, and data-center names are down 25–40% across the last 30 days in the global tape. Datadog got swept into that de-rating despite having almost no capex and a cloud-hosted cost structure; its R&D line keeps rising, up 23% year over year. The company is, in effect, a beneficiary of the complexity that the AI build-out creates rather than a participant in its capital cycle. Three fresh, company-specific proof points deserve highlighting. Defense Agency — a five-year, seven-figure deal with a U.S. federal defense agency, landing nine products, made possible by FedRAMP High and a roadmap toward IL5. Japanese banks — two of Japan's largest, modernizing core systems onto the full platform in the same quarter. And Cloud Cost Management, where AI spend tracking has become a genuine wedge: more than a third of CCM customers now monitor AI costs, up more than 11x in customer count. Agent Observability is the sleeper. It rose from a peripheral mention to a top-five keyword this quarter, and the CEO noted the demand is no longer just AI natives: “what is most interesting here is the non-AI-natives are consuming more and more... we think that this is like, crossing the chasm, so to speak.” — Olivier Pomel, Cofounder and CEO · 2026-11-04 That is the same non-AI cohort accelerating for five straight quarters — the two trends are the same trend, and the company is being valued as if only the first one matters.

The verdict

The bear case writes itself: growth decelerates to 25% in the guide, one customer controls the swing factor, and early renewals inject noise into every sequential print. The bull case is that every distortion this quarter is a mechanical artifact of customers who are growing too fast — and that a company which just taught itself to build cheaper models for its own data plane is compounding an advantage rivals cannot buy. The market chose the first story in three days of trading. The transcript argues for the second. Somewhere in that gap is the actual opportunity.