Datadog Stock After Q3 2026: The Flat Guide Is Mostly One Customer's Floor

I think the market sold Datadog on a number that mostly describes one contract. The Q4 2026 guide is flat with Q3, and the stock fell about 12% before the open on November 4. But CFO David Obstler said on the call that the guide prices the company's largest customer at the least it is contractually bound to spend, and that growth excluding that customer is guided in the low to mid 30s. The total is 25% to 26%.

Datadog sells observability: software that watches a company's servers, apps and logs and tells engineers what broke. It doesn't name the largest customer or say how big it is, and I won't guess. What it did say is enough to take the flat quarter apart.

Obstler's prepared remarks said the largest customer "reduced their usage in the third quarter, while their usage remained above their minimum commitment." In the Q&A he said the customer "spent substantially above the minimum commitment" in Q3, and that it renewed. Then the line that explains the guide: "our Q4 guidance contemplates their revenue at their minimum commitment."

The spread between those two Q4 growth rates is the clearest number in the print. Take the largest customer out and Datadog is guiding growth several points above the total. Some of the flat sequential line is also the slow late-December stretch Obstler mentioned, and he wouldn't size either piece, so I can't split them exactly. But the account explains most of it, and Obstler said any spend above the minimum would be "an implicit beat, mathematically."

Datadog's record makes that believable. It beat the top of its revenue guide by $37 million to $52 million in each of the last four quarters, and it says it guides on recent trends with conservatism built in.

In an August piece I said I'd be wrong if Q3 printed at or below its guide and Q4 failed to re-accelerate, leaving the full year below the range. Q3 beat and the range went up, so that test is still holding. The Q4 guide does press on its second half, and the spread above is why I don't think it breaks it.

Datadog bills mostly on usage. A customer commits to a minimum, then pays for what its servers, logs and traces actually send. That's why one account can move the line inside a quarter. It's also why the minimum matters: it's the part of the bill a customer can't optimise away until the contract ends.

The switching cost lives somewhere else. It's the monitors, dashboards and incident workflows a team builds, all fed by the one piece of Datadog software running on each server. On the call, CEO Olivier Pomel gave the breadth figures in the table below, and every one rose from a year earlier. Gross retention, the share of last year's revenue that stayed, held in the mid to high 90s. On my reading the bundle got deeper this quarter. The framework still rates it a notch below its top grade, because until now Datadog hadn't disclosed how many customers buy several products.

The same split runs through the other usage-billed software I cover. Snowflake and MongoDB also sell capacity that customers commit to and then draw down, so a big customer tuning its workloads lands in revenue before it lands in retention. Cloudflare sells more of its revenue as contracted subscriptions, where a cut waits for the renewal. I don't have a sourced large-customer episode for those three this quarter, so they're here for the billing model and the live scores, not as evidence.

The strongest objection is that the largest customer is the kind of buyer that can squeeze a vendor, and Datadog is adding more of them. Pomel said all 15 of what he called the top 15 AI leaders are customers. Datadog added about a hundred AI customers in Q3 alone, and 39 of the cohort now spend over $1 million a year. These companies have deep engineering benches, huge telemetry bills and every reason to treat observability as a cost to cut, or to build in-house. If the biggest of them cut usage and still renewed, the lesson may be that Datadog's switching cost is weakest exactly where its growth is fastest.

The renewals point the same way. Pomel said early renewals with large AI customers create "headwind into the short term" on revenue, and Obstler said they come at higher commitments in exchange for volume discounts. A bigger floor at a lower unit price is still a lower unit price. If more of the base renews on those terms, the high-20s growth I'm leaning on could be rolling into cheaper contracts a year from now.

A renewal also doesn't rule out a customer building its own tools alongside Datadog and spending down to the floor until the next contract. Obstler said Datadog is "not in the business of projecting" what the customer does next, and the 2027 guide in February is when this gets settled.

I take that seriously, and it's why my test is about the base and not the customer. If AI labs can push Datadog's prices down, it'll show up first as growth excluding the largest customer slowing, because the other labs are inside that number. So far it has done the opposite.

1. 1February 2027, the Q4 2026 call: growth excluding the largest customer. Below 30% means the base slowed and I'm wrong.
2. 2The same call: the 2027 guide, and whether non-AI customer growth is still described as accelerating.
3. 3Product breadth: the share of customers on four or more products, last at 61%. A stall there would worry me more than any single account.
4. 4Any disclosure of the largest customer's share of revenue or contract term, which would let me size the floor instead of inferring it.