A beat by every conventional measure
On the numbers alone, Datadog's second quarter of fiscal 2026 was a strong result. Revenue reached $1.12 billion, up 36% year over year and 11% sequentially, the fastest quarter-over-quarter growth the company has posted since the second quarter of 2022. Non-GAAP earnings per share came in at $0.65 against a forecast of $0.58, a beat of more than 12%, and free cash flow hit $279 million on a 25% margin, according to the company's press release and its August 6 earnings call.
Customer metrics moved in the right direction too. Datadog ended the quarter with roughly 33,400 total customers, up from 31,400 a year earlier, and 4,720 customers generating more than $100,000 in annual recurring revenue, up 23% year over year. New logos contributed 30% of year-over-year growth, up from 25% in the prior quarter, and 58% of customers now use four or more Datadog products, evidence that the platform-consolidation strategy the company has pursued for years is still working.
The sentence that erased the gains
None of those figures stopped the stock from falling roughly 15.6% in premarket trading, from $283.17 to around $239. The trigger was a specific disclosure buried in an otherwise strong call: CEO Olivier Pomel confirmed the company had renewed its largest customer, but acknowledged a user reduction starting in the third quarter, which he said was already incorporated into guidance. When Morgan Stanley analyst Sanjit Singh pressed for whether the reduction reflected churn or a pricing renegotiation, Pomel declined to give specifics, saying only that the company had fully de-risked guidance.
The market's reaction tells you something about how investors read observability economics right now. A single customer trimming usage was enough to overshadow 36% top-line growth and a double-digit earnings beat, because it raised the question of whether Datadog's largest, most sophisticated customers are starting to actively manage down their observability bills rather than letting usage scale automatically with infrastructure growth. That is a different kind of risk than slowing new-customer acquisition, and investors priced it as the more serious one.
Why usage-based pricing cuts both ways
Datadog's entire growth engine over the past several years has run on usage-based pricing tied to hosts, containers, log volume, and now AI inference calls. That model is what let the company post 36% growth in a quarter when the AI infrastructure buildout is pushing telemetry volumes to record highs. It is also, by design, a model where a single large customer optimizing its bill, migrating workloads, or consolidating tooling can move the needle on Datadog's reported growth rate in a way a fixed-seat SaaS contract never could.
Pomel's own framing of the customer relationship, that people buy software to make more money or save money, is an acknowledgment that Datadog's largest customers now actively manage observability spend as a cost center rather than treating it purely as a monitoring necessity. Datadog's response has been to build features like infinite cardinality metrics specifically to address unpredictable billing from fine-grained data tags, an attempt to keep customers from feeling the need to cut usage in the first place by making the pricing model less punishing.
The AI story underneath the headline
Pomel used the call to argue that the customer-specific headwind obscures a larger opportunity: AI inference workloads generating observability demand at every layer of the stack, from GPU infrastructure through the application layer to agent-level outcomes. Datadog counts more than 750 AI-native customers, including 31 spending over $1 million annually and, in Pomel's words, all 10 of the top 10 AI leaders. Non-AI revenue also accelerated into the high-20s percentage range year over year, evidence that AI adoption alone is not the only thing keeping growth elevated.
The company backed that narrative with an acquisition and a product launch during the quarter: it bought Adaptive ML, a reinforcement-learning operations startup, and unveiled more than 100 new capabilities at its DASH 2026 conference, including autonomous incident detection and remediation and an AI Guard feature aimed at prompt-injection protection. Barclays analyst Raimo Lenschow argued the inference opportunity could ultimately outweigh the customer headwind, since observability opportunities exist at every layer of the inference stack.
What CFOs and CTOs should take from the stock reaction
If you run a large Datadog contract, this earnings call is worth reading in full, because it shows what a customer looks like when it successfully renegotiates leverage with Datadog. The unnamed largest customer clearly had enough scale and enough alternatives, whether in-house tooling, a competitor, or simple usage discipline, to force a usage reduction that Datadog had to publicly disclose as material. That is the kind of leverage every enterprise buyer wants and few actually have, and it is worth benchmarking your own contract terms against what evidently worked for that customer.
The more immediate lesson is about volatility in usage-based vendor relationships generally. If your organization's observability, data warehouse, or streaming bill scales with infrastructure usage, that bill will keep growing as AI workloads generate more telemetry, more logs, and more inference calls, and the vendor has every incentive to let it. CTOs should run the FinOps exercise Datadog's largest customer apparently already ran, well before renewal season forces the conversation on them.
The consolidation trade-off this puts back on the table
Datadog's statistic that 58% of customers use four or more products is the flip side of the cost story. Consolidating monitoring, security, and now AI observability onto a single platform genuinely reduces tool sprawl and integration overhead, which is exactly why so many customers keep adding products even as their bills grow. But consolidation onto a usage-based platform means cost exposure grows with every product added, not just with infrastructure footprint, a materially different risk profile than adding a flat-fee tool.
The practical takeaway for a CTO deciding whether to expand a Datadog footprint, or any usage-based observability platform, is to model the cost curve under an AI-driven telemetry growth scenario before signing, not after the first surprising invoice arrives. Datadog's own largest customer apparently ran that math and pushed back hard enough to move the needle on a $1.12 billion quarter. Enterprise buyers with meaningfully smaller scale should assume they hold less leverage than that customer did, and negotiate protections into the contract now.
