Datadog Slides After Top AI Customer Flags Q3 Usage Cut

A beat-and-raise quarter that took a fifth off the stock is the sort of print worth pausing on, because the mechanics say something about how the AI build-out is landing on the vendors underneath it. Datadog reported Q2 revenue up 36% to $1.12 billion, ahead of the roughly $1.08 billion consensus, and raised its full-year outlook to $4.45-$4.47 billion. Shares still fell around 14-18%, according to the Wall Street Journal and follow-on coverage.

The reason is concentration. The company told investors its largest AI customer, an account that recently signed a nine-figure renewal, will cut usage in the back half of the year. Q3 revenue is now guided to $1.135-$1.145 billion, which implies growth slowing to roughly 29% from the 36% just posted, and some analysts read the resulting fourth-quarter walk as closer to 25%. OpenAI is reportedly Datadog's biggest single customer, and AI-native accounts made up about 11% of Q2 revenue, up from 4% a year earlier, so a single whale trimming consumption shows up in the print immediately.

Why this matters if you are not a Datadog shareholder: this is the first clean read on what happens to a usage-priced infrastructure vendor when a frontier lab decides to optimize. Cloud consumption billed by the log line or the ingested metric is beautiful on the way up and unforgiving on the way down, and the same properties that made Datadog the cleanest AI beneficiary in B2B software are now the properties that punish it when one customer changes its mind. Rivals in observability will be pitching 'no single-customer cliff' by Monday.

The honest caveat is that the reporting here is thin on why the customer is cutting back, whether the workload is being brought in-house, moved to a competitor, or simply run more efficiently against the same underlying models. What the coverage does not give you is how much of the raised full-year guide already bakes in this pullback versus assumes it stabilizes, or how concentrated the rest of the AI-native book is behind the top account.

The forward read is that consumption pricing meets frontier-lab economics is going to be a recurring earnings-day theme for anyone whose meter spins when GPUs do, and the vendors that get ahead of it with disclosure and product packaging, Bits AI and the Adaptive ML acquisition in Datadog's case, will be the ones that trade on their platform story instead of their top-customer story.