Datadog (DDOG) Q1 FY26 earnings call: guidance and Q&A

How much of the exploding code produced by code generators (Claude Code, Codex, Cursor) is reaching production and driving Datadog activity, and does the growing heterogeneity of custom silicon (Trainium, Graviton, TPUs, Maia) act as a tailwind?

Olivier said many more applications are being created and moving to production across both AI-native and non-AI companies, visible in rising AI-product data volumes. On silicon, heterogeneity plays in Datadog's favor since customers need someone to tie GPU and non-GPU infrastructure together; training is democratizing beyond a few players, and Datadog is now landing hyperscalers' super-intelligence labs to monitor and accelerate training runs — a shift from a year ago when it saw only inference as a market.

“Interestingly, last year, when we reported earnings, we said we're mostly interested in inference workloads and training is not really a market for us yet. Now we actually see training becoming a market.”

asked by Mark Murphy · answered by Olivier Pomel

Any impact from geopolitical tensions or consumer-discretionary softness in e-commerce/retail/travel; and how does the category evolve — and pricing — when agents rather than humans do triaging and investigation?

David said the quarter was strong across industries and geographies with strong SMB, no observed effect yet in consumer/e-commerce/travel businesses, and guidance discounts trends across the board plus extra conservatism on the largest customer. Olivier said Datadog's usage-based model is agnostic to whether usage comes from humans or agents; it currently sees both stratospheric agent usage (MCP server, customer agents) and rising human web-interface usage working hand-in-hand.

“we're usage-based, and it doesn't really matter where the usage is coming from, from that perspective.”

asked by Sanjit Singh · answered by David Obstler

What underpins the confidence in a Q2 guide implying the largest sequential dollar add in company history, and any color on the ramp of the large research labs?

David said the biggest near-term indicator is prior-quarter ARR growth, which was a record and very broad-based (not concentrated); guidance takes signed ARR and discounts growth trends. Olivier reinforced that adds were broad-based — even excluding the top revenue-contributing Q4 land, Q1 was still a record ARR-add quarter — and several Q1 lands not yet contributing revenue are expected to be big future contributors.

“even if you take out the customer we land in Q4 that added the most revenue in Q1, we still had a record quarter in terms of ARR adds.”

asked by Karl Keirstead · answered by Olivier Pomel

Why is the training opportunity inflecting now, and does observability spend as a percentage of AI spend change given the new training data?

Olivier explained training is shifting from artisanal, one-off research work done by very few companies into always-on production workloads scaling by orders of magnitude — where every failed training run cedes a week to competitors — making it a far more interesting market. David said the 6,500 AI-integration customers (20% of customers, 80% of ARR) evidence attach that is still primarily inference-driven, with training an early but promising contributor.

“every failure you have in your training runs is a week you give away to the competition.”

asked by Gabriela Borges · answered by Olivier Pomel

What is really happening with open-source tooling consolidation, and how much of accelerating non-AI growth is cloud-migration industry trend versus Datadog being better positioned?

Olivier said most companies keep some open source, but come to Datadog to unify fragmented tooling (often 4-25 tools across business units) for better results, automation and cost savings — and this quarter Datadog even displaced in-house builds at hyperscalers. David cited platform adoption breadth and the migration of open-source and competitive point solutions onto the platform as durable revenue drivers, plus successful ramp of sales capacity without hurting productivity.

“they have the same issues. When it comes to going as fast as they can and being as efficient as they can with their resources, like they come to us to replace some of the things that we were using before.”

asked by Raimo Lenschow · answered by Olivier Pomel

With telemetry volumes going parabolic and new training workloads, how does Datadog keep capex and gross margins efficient, and how does it address rising sovereign/data-residency requirements?

Olivier said most workloads run on cloud, so investment shows in opex not capex, keeping capex low; investments are ramping in R&D and self-trained models but nothing moves the numbers materially yet, and Datadog would flag any model change. On data residency/sovereignty, investment is going into more geographies and certifications (UK data center, FedRAMP High, higher federal levels) and bring-your-own-cloud products that run on customers' infrastructure.

“we run most of our workloads on cloud, meaning you'll see all of that in OpEx, not in CapEx.”

asked by Fatima Boolani · answered by Olivier Pomel

Clarify the 'higher degree of conservatism' on the largest customer — is it relative to other cohorts or to prior-quarter methodology?

David said it is both — the same methodology as prior quarters, with the largest customer discounted more than the rest of the base; there is no change in methodology versus prior quarters. Olivier downplayed reading too much into specific wording, noting both executives simply had raspy voices with no hidden meaning.

“for this particular customer, we took a higher degree of conservatism than the other part of the customer base and discounted it more.”

asked by Patrick Edwin Colville · answered by David Obstler

Given hyperscalers typically build observability in-house, what about AI workloads makes Datadog attractive and persistent for them?

Olivier said the same reasons all customers use Datadog apply — high stakes, high complexity, and non-core to their differentiation — and the urgency of the AI race forces even hyperscalers (who historically had near-unlimited staffing and could set their own timelines) to decide what is core and outsource the rest to maximize speed and chances of success.

“I think the urgency of their development efforts focuses the mind.”

asked by Peter Weed · answered by Olivier Pomel