A Beat-and-Raise Quarter That Still Cost 17 Percent
Datadog's stock closed down about 17 percent on August 6, 2026, on a morning when nearly every headline number the company reported had beaten Wall Street's expectations. The observability and security software company said revenue for its second quarter, ended June 30, 2026, reached 1.12 billion dollars, up 36 percent year over year, with non-GAAP earnings of 0.65 dollars per diluted share, about 0.07 dollars ahead of what analysts had modeled. Operating cash flow came in at 316 million dollars and free cash flow at 279 million dollars, and the company said it now had about 4,720 customers paying it 100,000 dollars or more a year, up 23 percent from roughly 3,850 a year earlier.
Datadog did not just beat the quarter it had already guided to - it raised the outlook for the rest of the year. Full-year 2026 revenue guidance moved to a range of 4.45 to 4.47 billion dollars, with non-GAAP operating income guided to 1.01 to 1.03 billion dollars and non-GAAP earnings per share to 2.50 to 2.54 dollars, all filed the same morning in the company's own earnings release. A beat on the quarter just reported plus a raise on the year ahead is the combination boards and investors normally read as an unambiguous green light. Instead, the stock lost roughly a sixth of its value in a single session, which is the puzzle this piece is actually about.
The Nine-Figure Renewal That Came With a Usage Cut
The answer sat inside the same earnings call, in a customer Datadog would not name. Co-founder and chief executive Olivier Pomel told analysts the company's single largest customer, which he described as a longtime account using 17 different Datadog products, had reduced its usage of the platform starting in the third quarter, and that Datadog had 'fully de-risked' its third-quarter and full-year guidance against that reduction before publishing it. Chief financial officer David Obstler confirmed the cut had already been built into both the third-quarter revenue range and the full-year raise, rather than being a risk still sitting in front of the company. In the same breath, Pomel disclosed that the same customer had just signed what he called a 'nine-figure renewal' - a multi-year commitment measured in the hundreds of millions of dollars, arriving in the identical quarter as the usage cut.
Analysts pushed on the details Datadog was not offering. Morgan Stanley's Sanjit Singh asked directly whether the reduction reflected the customer churning workloads away or simply renegotiating a lower unit price, and Pomel declined to separate the two, repeating only that guidance was already de-risked. Datadog itself never named the customer in the release or on the call, describing it only as a leading AI company; multiple outlets covering the call have reported it is widely believed in the market to be OpenAI, an identification Datadog has not confirmed. What is confirmed, in the company's own words, is the shape of the event: a renewal and a cutback from the same counterparty, disclosed in the same ten minutes.
What a Beat-and-Raise Quarter Does Not Tell You
Total revenue growth is a single number, and a single number can rise even while the base underneath it is consolidating around fewer, larger counterparties whose own decisions a vendor does not control - Datadog's quarter is the concrete proof of that mechanism, not an exception to it. The 36 percent growth investors cheered in Q2 and the 28 to 29 percent implied by Datadog's own Q3 guidance are both true at once, and the gap between them is a company-supplied deceleration number, not analyst guesswork; it exists because one customer's usage curve is now pulling against everyone else's growth.
The detail worth isolating is the one Pomel volunteered without being asked twice: strip the largest customer out of Datadog's growth entirely, and the rate for the rest of the business is, in his words, 'pretty much the same' as the reported total. That single sentence is doing more work than the guidance raise itself, because it is the test that tells an outside observer whether Datadog's growth is broad or borrowed - and in this case the answer came back broad, which is the actual reason the raise held up despite the cut. Most companies reporting a beat-and-raise quarter never run that test in public, and most investors never ask for it before they price the stock on the top-line number alone.
The Concentration Test for Your Own Business
Any owner or operator who sells infrastructure, tooling, or platform services to a small number of very large customers can run the same test Datadog ran on itself, and should run it before the next board update rather than after a customer forces the question: take this quarter's total revenue growth, subtract the growth contributed by your single largest customer, and compare the two rates side by side. A small gap means your growth is broad, the way Datadog's turned out to be. A large gap - most of this quarter's growth coming from one counterparty - means your topline is consolidating around a customer whose own business model, budget cycle, or usage pattern you do not control, no matter how healthy your aggregate number looks.
The cadence matters as much as the calculation. Datadog re-ran this test every quarter and had already priced the cut into two separate guidance ranges before a single analyst asked about it; a business that only checks its concentration at contract renewal time, once a year or once every few years, finds out the same way Wall Street found out about Datadog on August 6 - after the number is already moving. For any owner selling into a handful of large AI-lab or hyperscaler accounts specifically, where budget cycles and internal usage optimization can shift a counterparty's spending inside a single quarter, running this test quarterly rather than annually is the concrete, checkable habit this story argues for.
Read next: GIC and Macquarie Become Anthropic's Landlords | 530 US Counties Now Restrict New Data Centers
