DDOG — Datadog, Inc.

Datadog has built the most comprehensive unified observability platform in cloud infrastructure — 32,000 enterprise customers, a 120% net revenue retention rate, and 80% gross margins — and the OpenAI churn controversy that drove the stock down 35% from its November 2025 high has obscured what the operating data shows: the platform is winning in enterprise and the AI complexity wave is expanding its footprint rather than contracting it. The business is excellent; the stock at $128.82 trades at 45 times enterprise value to reported free cash flow, a multiple that embeds years of compounding growth as the entry price rather than as the upside. Good business, meaningfully overpriced.

---

Enterprise technology spending in 2026 has entered a new discipline phase. The unconstrained cloud adoption cycle of 2020 to 2022 — when engineering teams added observability tools, security platforms, and data products without meaningful cost scrutiny — has given way to FinOps mandates, observability budget reviews, and a systematic rationalization of point solutions. Chief information officers who once approved an AppDynamics deployment alongside a Splunk instance alongside a New Relic rollout are now demanding consolidation onto fewer platforms with demonstrable return on investment. The immediate effect on observability vendors is pricing pressure from customers who are over-deployed and expansion resistance from buyers evaluating total cost of ownership for the first time. The second-order effect is a structural advantage for the platform that survives the rationalization — the vendor still installed after the budget committee finishes its review is the vendor that captured the switching cost.

The second macro current affecting observability vendors in 2026 is the OpenAI effect, which is the specific narrative that caused Datadog's stock to fall from $199.72 in November 2025 to its current $128.82. OpenAI was Datadog's largest single customer, with annual recurring revenue estimated at approximately $240 million by mid-2025. OpenAI began building internal observability tooling during 2025 and accelerated that transition through year end; its ARR with Datadog is projected to decline to approximately $80 million by the end of 2026, a reduction of roughly $160 million. On a revenue base of $3.43 billion, this represents approximately five percentage points of growth headwind in 2026 — real but not existential, and specifically isolated to a customer whose engineering capacity and infrastructure spend exist nowhere else in the commercial market. The question the market conflated with this event is whether OpenAI's in-house build signals what every major cloud-native company will eventually do. The operating data for the other 31,999 customers suggest it does not.

The observability and application performance monitoring market is a $28 billion addressable opportunity as of 2026, growing at approximately 15 to 16 percent annually, and is structurally different from most enterprise software categories in one critical way: the cost of not monitoring is catastrophic. A financial services firm running unreliable trades, a retail platform losing transactions during peak load, a healthcare system with degraded application response times — these are not inconveniences. They are existential events for the businesses involved. This necessity-product characteristic means demand is not optional and is not the first line item cut when budgets tighten. It means pricing power is real. It means the platform already instrumented into production environments has the same structural advantage as any indispensable infrastructure provider: replacement requires engineering effort, risk, and downtime that no organization undertakes without serious cause.

The industry has consolidated meaningfully since 2020. New Relic, once Datadog's primary competitor in full-stack observability, was taken private by Francisco Partners and TPG in late 2023 — evidence that at the current pace of platform competition, point-solution vendors cannot sustain public-market valuations. Dynatrace remains the most direct public peer, with approximately $1.5 billion in annual subscription revenue and a highly automated platform focused on AI-assisted monitoring. Splunk, historically the log management leader, was acquired by Cisco for $28 billion in March 2024 — at a 31-times revenue acquisition multiple — establishing a valuation reference for the category's most data-rich platform. The consolidation of observability around a small number of comprehensive platforms is the structural force that makes Datadog's addressable market expansion credible: enterprises simplifying from five vendors to one are specifically choosing what Datadog has spent fifteen years building.

Datadog was founded in 2010 by Olivier Pomel and Alexis Lê-Quôc, both former systems engineers who had experienced the fragmentation of monitoring tooling in cloud infrastructure firsthand. Their original insight was that infrastructure monitoring, application performance monitoring, and log management were three facets of the same problem — what is happening in production right now — and that the industry's practice of running them on three separate platforms with three separate data models made troubleshooting unnecessarily slow. The company spent its first decade building a unified platform that ingested metrics, traces, and logs into a single data model, allowing engineers to move between infrastructure dashboards, distributed traces, and log streams within one interface. By 2026, the platform covers fifteen product categories: infrastructure monitoring, APM, log management, real user monitoring, synthetic testing, security monitoring, database monitoring, network performance monitoring, CI visibility, incident management, workflow automation, cloud cost management, and an AI observability layer for LLM application monitoring. Each new product layer is built on the same underlying data model, which means instrumentation installed for infrastructure monitoring also feeds APM alerting, which also feeds security incident detection. The integration is what creates the switching cost, and the switching cost is what creates the platform economics.

The revenue model is consumption-based: customers license Datadog by the number of hosts monitored, log volume ingested, metrics retained, and sessions tracked. This creates a revenue structure that scales automatically with customers' infrastructure — as an enterprise grows its cloud footprint, its Datadog spend grows proportionally, without requiring a renegotiation. Approximately 75 percent of Datadog's revenue growth comes from expansion within existing customers rather than new logo acquisition. The 120% net revenue retention rate — meaning the existing customer cohort's spending grew 20% over the past year without any new customers added — is the direct financial expression of this model's mechanics.

The moat is the unified platform's switching cost structure. Datadog does not merely monitor; it becomes the operational nervous system of its customers' cloud environments. Custom dashboards, alert policies, incident response workflows, and diagnostic procedures are built on top of Datadog's data model and would need to be rebuilt entirely on any alternative. Historical data — the baselines and anomaly detection patterns that make monitoring operationally useful — is stored in Datadog's infrastructure and cannot be portably migrated. As customers expand from one product to multiple products, each additional layer deepens the switching cost further: an enterprise using infrastructure monitoring, APM, log management, and security monitoring has four separate operational processes that would need to be simultaneously migrated and validated before any cutover could succeed. Eighty-five percent of Datadog customers use two or more products; 45 percent use four or more. This multi-product adoption rate, measured consistently over six years of disclosed data, is the most direct evidence that the switching cost structure is real and deepening.

	Datadog (DDOG)	Dynatrace (DT)
Net revenue retention rate	~120%	~110–115%
Subscription gross margin	~80%	~83%
Revenue growth (FY2025)	+28%	~+15%
$1M+ ARR customers	603	~160 (est.)
Multi-product adoption	85% use 2+; 45% use 4+	Predominantly 1–2

The comparative data makes the competitive position concrete. Datadog's 120% NRR against Dynatrace's approximately 110 to 115% looks like a modest gap in percentage points but represents a compounding advantage: over five years, Datadog retains and expands its original customer cohort to roughly 249% of original spend while Dynatrace grows it to approximately 185%. The 603 enterprise customers with $1 million or more in annual recurring revenue — against Dynatrace's estimated 160 — is the clearest statement of which platform large enterprises are choosing when they standardize their observability stack. The comparable gross margins confirm that the competitive advantage is not being purchased through pricing discounts; both platforms sustain approximately 80 to 83 percent subscription margins, reflecting pricing power across the category rather than a commodity dynamic.

Datadog's FY2025 financial results show $3.43 billion in total revenue, growing 28 percent year over year, with a non-GAAP operating margin of 22 percent and reported free cash flow of $915 million, representing a 27 percent FCF margin. These are the headline figures that appear in investor presentations and that drive the EV/FCF multiple of approximately 44 times on reported numbers. The GAAP figures require more careful examination. GAAP net income for FY2025 was approximately $107 million — not $915 million — because free cash flow adds back stock-based compensation of approximately $640 million annually. This reconciliation item is not merely accounting convention. It represents real economic dilution: at $640 million in annual SBC on a diluted share count of approximately 337 million, Datadog is issuing approximately $1.90 per diluted share annually in economic value to employees — a figure that must be deducted from apparent cash generation to arrive at what the shareholder actually retains. Adjusted for SBC, real economic free cash flow is approximately $275 million. The EV to real cash generation multiple is approximately 147 times.

The non-GAAP operating metrics are informative about the platform's unit economics — 80 percent gross margins reflect genuine pricing power and scalable infrastructure costs — but the SBC adjustment is the reason GAAP net income and reported FCF diverge by $808 million on a $3.43 billion revenue base. The company's FY2026 guidance for non-GAAP operating income of $840 to $880 million does not resolve this. If SBC scales proportionally with revenue — and at approximately 19 percent of revenue it has been remarkably stable — the GAAP operating income at the guided revenue level of $4.08 billion will be approximately $85 to $100 million, implying pre-tax income of roughly $235 million after adding interest income from the company's $3.2 billion net cash position. The GAAP earnings yield at the current stock price is less than half of one percent.

Olivier Pomel has led Datadog from founding through its 2019 IPO to $3.4 billion in revenue without a CEO transition, without a capital-destructive acquisition, and without the over-hiring cycle that forced painful workforce reductions at most software peers in 2022 and 2023. The company went public at $27 per share with no debt, managed the COVID demand surge without compromising operating discipline, and expanded FCF margins from approximately 15 percent in FY2021 to 27 percent in FY2025 while sustaining growth rates that few software businesses at this scale have matched. The capital allocation record is clean: no significant acquisitions, no dividend program, no buyback program of scale — the $3.2 billion net cash position accumulates on the balance sheet. The absence of buybacks while issuing $640 million annually in SBC is the one meaningful tension: the company is growing the diluted share count rather than shrinking it, which is a rational choice for a high-growth business but means shareholders are funding employee compensation through dilution rather than through cash expense.

The growth runway is best understood through the metrics that track whether the land-and-expand model is functioning as designed.

Year	Revenue ($B)	YoY Growth	Net Revenue Retention	$1M+ ARR Customers	FCF Margin
FY2021	$1.03	+63%	130%+	~135	~15%
FY2022	$1.68	+63%	130%+	216	~16%
FY2023	$2.13	+27%	~119%	317	~20%
FY2024	$2.68	+26%	~114%	461	~23%
FY2025	$3.43	+28%	~120%	603	~27%
FY2026E	~$4.08	+19%	—	—	~28%

The table tells three concurrent stories. Revenue growth accelerated through FY2022 as cloud migration surged, decelerated sharply in FY2023 as enterprise cloud budgets hit their first optimization cycle, stabilized at 26 to 28 percent for two years, and is now guided lower to 19 percent in FY2026 — of which approximately five percentage points reflect the OpenAI ARR headwind. Net revenue retention declined from 130-plus percent at peak to approximately 114 percent in FY2024 as cloud spending rationalization suppressed expansion, then recovered to approximately 120 percent in FY2025 as customers resumed expanding their observability footprint. This NRR recovery — happening during the same period the OpenAI narrative was most visible — is the key evidence that the churn concern is customer-specific rather than platform-wide. The $1 million-plus ARR cohort grew from 461 to 603 customers in a single year, a 31 percent increase; customers with $10 million or more in annual spending grew over 60 percent year over year to 34. These are the bellwether numbers. When the largest and most sophisticated enterprise accounts are accelerating their spend on a platform during the period of maximum competitive anxiety about that platform, the competitive anxiety is wrong.

The structural reason the growth trajectory has been durable is that cloud infrastructure complexity is not transitory. The AI infrastructure layer — GPUs, LLM inference endpoints, model training pipelines, vector databases — is a new observability challenge being added on top of the existing cloud complexity Datadog already monitors. Fourteen of the top twenty AI-native companies by infrastructure spend are Datadog customers. The 650 AI-native companies in the customer base are growing their Datadog spending faster than the overall population. This creates a structural tailwind that will not peak in 2026: the number of AI agents, model endpoints, and GPU clusters requiring monitoring will compound for years as enterprise AI deployment scales from pilot to production. Datadog's AI Observability product — purpose-built for LLM trace monitoring, token cost tracking, and model performance analysis — is designed to capture the incremental spend that comes with this complexity growth. A FedRAMP High authorization in process as of early 2026 adds the U.S. federal government as an incremental segment whose scale and observability spend could be material once authorization is completed.

Datadog has captured approximately 32,000 paying customers against an estimated addressable universe of 500,000 global enterprises with meaningful cloud infrastructure — a penetration rate of 6.4 percent. At $3.43 billion in revenue against a $28 billion addressable market, revenue penetration is approximately 12 percent. The nearly 468,000 enterprises that have not yet adopted the platform include millions of developers currently running fragmented stacks of open-source tooling — Prometheus for metrics, Grafana for dashboards, the ELK stack for logs — who will eventually face the operational complexity of coordinating those tools at scale and choose a managed platform. International markets represent approximately 30 percent of current revenue against a global cloud infrastructure base that is roughly 50 percent non-North American; Europe and Asia-Pacific are in earlier stages of cloud adoption than the United States, with new sovereign data regions in Germany and Japan targeting regulated industries that have historically underinvested in observability relative to their infrastructure complexity.

At $128.82 per share, Datadog trades at a market capitalization of approximately $43.5 billion. Net cash of approximately $3.2 billion implies an enterprise value of $40.3 billion. Against FY2025 reported free cash flow of $915 million, the EV/FCF multiple is 44 times. Against real economic free cash flow — after deducting $640 million in annual SBC — the EV to cash generation multiple is approximately 147 times. On normalized pre-tax earnings per share — using FY2026 guided non-GAAP operating income of $860 million, adjusted for SBC of approximately $775 million running at 19 percent of guided revenue, plus interest income of approximately $150 million, divided by 340 million diluted shares — normalized pre-tax EPS is approximately $0.69 per share. At $128.82, the normalized pre-tax multiple is 187 times. The price at which that multiple would reach 15 times is approximately $10 per share. This arithmetic is not a buy signal at $10; it is an illustration of what the normalized earnings framework tells you about a business where SBC runs at 19 percent of revenue — which is that the standard earnings multiple methodology produces a result that captures the shareholder cost of SBC but not the business quality that justifies paying any meaningful multiple for growth. The right way to look at this is EV/FCF on reported figures: 44 times, for a business growing 19 percent and generating real cash on a pre-SBC basis, is expensive but not absurd. Adding back SBC to arrive at 147 times — which is the honest figure — is expensive and absurd.

The most intelligent bear on this stock argues that Datadog's platform advantage is more fragile than the NRR suggests: OpenTelemetry, the vendor-neutral observability standard that collects metrics, logs, and traces in a portable format, is systematically reducing the instrumentation switching cost that has historically anchored customers. If engineers can redirect their telemetry pipeline from Datadog to Grafana Cloud or Honeycomb with three days of configuration work rather than three months of migration, the 120% NRR represents inertia rather than conviction. The counter is that switching the collection layer is the easy part; switching the alert configurations, dashboard architecture, incident response workflows, and historical analysis built on Datadog's specific data model and query language remains prohibitively complex regardless of how the underlying telemetry is collected. OpenTelemetry makes data portable. It does not make the platform portable. The $1 million-plus ARR cohort growing 31% during a period of maximum OpenTelemetry adoption and maximum competitive anxiety suggests customers with the engineering capacity to evaluate the alternatives have evaluated them and chosen to spend more.

For the thesis to change, either the price needs to come down meaningfully or GAAP earnings need to accelerate dramatically — and ideally both. If Datadog sustains 20-plus percent revenue growth for four years, FCF margins expand to 35 percent, and SBC as a percentage of revenue declines from 19 percent toward 12 percent as the business matures, normalized pre-tax EPS could reach $3 to $4 by FY2030. At 25 times normalized pre-tax on that earnings base, intrinsic value would be approximately $75 to $100 — below the current price. The business quality is excellent. The runway is long. The price does not wait for the runway to be covered.

The platform is exceptional. The price tags the quality and then charges for the next decade of compounding before it happens.