What is Datadog's unnamed eight-figure AI customer deal?

On Datadog's fourth-quarter earnings call in February, CEO Olivier Pomel disclosed that the company had signed an "eight-figure annualized" contract with "one of the largest AI foundational model companies" — one of the biggest new logos in Datadog's history. The customer was migrating off more than five tools, including open-source, commercial, hyperscaler, and in-house solutions, onto Datadog. Pomel did not identify the customer by name, and Datadog has not confirmed the identity since.

Didn't an Anthropic employee say the company doesn't use Datadog?

In a July 2025 Tegus interview, an unnamed Anthropic employee reportedly said the company "didn't even touch Datadog ever." Hunterbrook presents this alongside the documented evidence of Datadog running inside Claude products as a contradiction, and frames it as an example of the bear thesis — that AI-native companies would build observability tooling in-house — falling apart under scrutiny. Hunterbrook's report does not identify or verify the employee's role, and the claim is attributed to that Tegus interview.

What public ties already exist between Datadog and Anthropic?

Anthropic President Daniela Amodei announced a joint product integration with Datadog at Datadog's DASH 2024 conference. Datadog subsequently published a dedicated Anthropic monitoring product and a Cloud Cost Management integration for tracking Anthropic API spend. Datadog is also listed as a named member of Anthropic's Claude Design research beta. Hunterbrook frames these public ties as consistent with a major commercial relationship, though none of them individually confirm the identity of Datadog's unnamed earnings-call customer.

Has either Datadog or Anthropic confirmed this relationship?

Neither company has confirmed that Anthropic is the customer Pomel referenced on the February earnings call. Hunterbrook's report does not describe a formal request-for-comment process, and no direct response from either company is quoted beyond the absence of confirmation. The customer identification is Hunterbrook's inference from technical and circumstantial evidence, not a disclosed fact.

Hunterbrook: Datadog's Mystery Eight-Figure AI Customer Is Almost Certainly Anthropic, and the Evidence Is in Claude's Own Source Code

Q: What is the Claude Code Datadog feature flag?

Claude Code's source contains a feature flag named `tengu_log_datadog_events` that controls whether the application sends telemetry events to Datadog. The flag is managed through GrowthBook, a standard A/B testing and feature rollout tool, suggesting a deliberate phased deployment rather than an accidental or experimental integration. A user-submitted configuration dump showed the flag turned on, with batching settings consistent with a production telemetry pipeline.

The case builds from multiple independent sources. Datadog RUM is classified as "essential" by Anthropic and enabled by default across Claude products, with enterprise customers required to actively opt out. Claude Code's source contains a hard-coded client token, a dedicated integration module, and a GrowthBook-managed feature flag named tengu_log_datadog_events – consistent with a phased production rollout rather than an experiment. Three public bug reports document the depth of the wiring: one showed 1.2 million failed log delivery attempts in 80 minutes, another showed Datadog's endpoint failing simultaneously with api.anthropic.com, suggesting it sits on the production critical path. Anthropic's enterprise Cowork feature sends structured model-interaction events including user_prompt, api_request, and tool_decision to Datadog's OTLP endpoint. The circumstantial layer: OpenAI is an existing Datadog customer and wouldn't qualify as a "new logo," Grafana Labs publicly named Anthropic as a customer in September 2025, and Anthropic is precisely the kind of technically sophisticated company Pomel described as "already running in-house tools" before consolidating. Neither company has confirmed the identification.

Ticker: DDOG (Datadog, Inc.)
Research Firm: Hunterbrook Media
Report URL: https://hntrbrk.com/datadog-anthropic/?ref=shortreport.fyi
Position Disclosure: Hunterbrook discloses a long position in Datadog and states it may benefit financially if the share price increases.

Thesis

Hunterbrook Media argues that Anthropic is the unnamed AI foundation-model company behind Datadog's "eight-figure annualized" deal disclosed on the February fourth-quarter earnings call — and that the relationship is deeper and more operationally embedded than the market appears to recognize.

Live RUM Deployment: A live inspection of Claude.ai found Datadog's Real User Monitoring SDK initialized with service name claude-ai, environment production, and a sample rate of 100%, meaning every consumer session is captured by Datadog. The same SDK is bundled in the Claude desktop app.
Default Infrastructure: Anthropic classifies Datadog RUM as "essential," enabling it by default across Claude products. Enterprise customers must actively opt out.
Expanding Telemetry Rollout: Claude Code source contains a hard-coded Datadog client token, a dedicated Datadog integration module, and a feature flag named tengu_log_datadog_events controlling whether users send telemetry to Datadog. The flag is managed through GrowthBook, a standard A/B rollout tool, consistent with a phased production deployment.
Bug Report Trail: A January 6 user report found Claude Code calling datadoghq.com every few seconds; the issue was closed as "not planned." On March 13, a user reported 1.2 million error log lines generated in roughly 80 minutes while Claude Code repeatedly attempted delivery to Datadog's log ingestion endpoint. A March 18 outage report showed that Datadog endpoint failing in lockstep with api.anthropic.com.
Enterprise Product Instrumented: A March 26 bug report on Anthropic's enterprise Cowork feature references an OpenTelemetry exporter sending event types including user_prompt, api_request, api_error, tool_result, and tool_decision to Datadog's OTLP logs endpoint — suggesting Datadog telemetry extends beyond consumer products.
Accidental Sourcemap Leak: On March 31, Anthropic accidentally shipped a sourcemap file inside a Claude Code update. Independent researchers including Karan Prasad and Engineer's Codex analyzed it and found a file path confirming a dedicated Datadog integration module alongside other analytics services.
Customer Identity — Circumstantial: Pomel described the mystery customer as "one of the largest AI foundational model companies," a new logo consolidating from "more than five open-source, commercial, hyperscaler, and in-house tools." Grafana Labs publicly named Anthropic as a customer in a September 2025 ARR announcement, fitting the open-source observability piece of that description. OpenAI, frequently cited by analysts as Datadog's largest single customer, would not qualify as a "new logo." The identification of Anthropic as that customer remains circumstantial.
Bear Thesis Undermined — Circumstantial: A July 2025 Tegus interview with an Anthropic employee claimed the company "didn't even touch Datadog ever." The report argues the evidence above, combined with Pomel's statement that the mystery customer had been using in-house tools before consolidating, suggests even the AI lab most capable of building its own observability tooling chose to buy instead.
Financial Upside: CFO David Obstler said 2026 guidance modeled the contract with "very conservative" assumptions given consumption-based pricing uncertainty. Wells Fargo estimated last year that Datadog's OpenAI contract represented "4–6% of DDOG's revenue run-rate," or roughly $200 million annualized — offered as a benchmark for what a major AI observability contract can be worth.

Notable Details

Anthropic's Cowork enterprise feature was instrumented to send specific event types — user_prompt, api_request, api_error, tool_result, and tool_decision — to Datadog's OpenTelemetry endpoint, per a March 26 bug report. The event taxonomy suggests Datadog is receiving structured data about how users and the model interact, not just error logs.
The March 13 bug report showed Claude Code generating 1.2 million error log lines in roughly 80 minutes — every line a failed attempt to deliver telemetry to Datadog's log ingestion endpoint. The volume implies Datadog is wired deeply enough into the product that a delivery failure created catastrophic log spam.
On March 18, Datadog's log ingestion endpoint and api.anthropic.com failed simultaneously, per an outage report. That timing raises the possibility that Datadog sits on the production critical path for users with the telemetry flag enabled — not just a side analytics feed.
Anthropic accidentally shipped a sourcemap file in a March 31 Claude Code update, giving outside researchers a rare view of the internal source tree. The file revealed a dedicated Datadog integration module sitting alongside Anthropic's other analytics services by name.
Despite all of the above, neither Datadog nor Anthropic has confirmed that Anthropic is the customer Pomel referenced on the earnings call.

"Anthropic, the frontier-AI customer most capable of doing exactly that kind of vibe coding — and, according to Pomel, already running in-house tools before the deal — seemingly chose instead to consolidate at least five tools onto Datadog."

— Sam Koppelman, Hunterbrook Media

FAQs

Why does Hunterbrook think the mystery Datadog customer is Anthropic?

Hunterbrook's case rests on two layers of evidence. The documented layer: Datadog's monitoring software is actively running inside Claude products, and Claude Code contains source-level Datadog integration code with feature-flag-controlled rollout. The circumstantial layer: Pomel's earnings-call description of a customer consolidating multiple observability tools fits Anthropic's known stack, Grafana Labs publicly named Anthropic as a customer in a September 2025 announcement, and OpenAI — the other commonly cited candidate — is already an existing Datadog customer and would not qualify as a "new logo." The customer identification remains an inference, not a confirmed fact.

What did Hunterbrook actually find in Claude.ai?

Hunterbrook's engineering team conducted a live inspection of Claude.ai and found Datadog's Real User Monitoring SDK loaded in the browser, initialized with service name claude-ai, environment production, and a sample rate of 100%. A 100% sample rate means every consumer session is captured. The same SDK was also found bundled inside the Claude desktop app. Anthropic classifies Datadog RUM as "essential" and enables it by default, requiring enterprise customers to actively opt out.

What is the Claude Code Datadog feature flag?

Claude Code's source contains a feature flag named tengu_log_datadog_events that controls whether the application sends telemetry events to Datadog. The flag is managed through GrowthBook, a standard A/B testing and feature rollout tool, suggesting a deliberate phased deployment rather than an accidental or experimental integration. A user-submitted configuration dump showed the flag turned on, with batching settings consistent with a production telemetry pipeline.

What happened with the Claude Code sourcemap leak?

On March 31, Anthropic accidentally shipped a sourcemap file inside a Claude Code software update. Sourcemaps are normally used by developers to debug minified code, but when shipped publicly they expose the internal source tree structure. Independent researchers including Karan Prasad and Engineer's Codex analyzed the file and found a path indicating a dedicated Datadog integration module alongside Anthropic's other analytics services — giving outside observers an unusually direct look at Anthropic's internal tooling choices.

What do the Claude Code bug reports show about Datadog?

Three public bug reports document Datadog activity in Claude Code across early 2025. On January 6, a user reported Claude Code was calling datadoghq.com every few seconds throughout each session; the issue was closed as "not planned." On March 13, a user reported that during a memory issue, Claude Code generated 1.2 million error log lines in roughly 80 minutes, each a failed attempt to deliver telemetry to Datadog's log ingestion endpoint. On March 18, an outage report showed that Datadog endpoint failing simultaneously with api.anthropic.com, suggesting possible production-path dependency for users with the flag enabled.

Is Anthropic also sending Datadog telemetry from enterprise products?

A March 26 bug report references Anthropic's enterprise Cowork feature using an OpenTelemetry exporter to send logs to Datadog's OTLP endpoint. The event types listed in the report include user_prompt, api_request, api_error, tool_result, and tool_decision — structured data describing model interactions, not just infrastructure errors. This suggests Datadog telemetry extends beyond Anthropic's consumer products into its enterprise offerings.

How big could the Anthropic deal be for Datadog financially?

The deal is described as "eight-figure annualized" and appears to be consumption-priced, meaning revenue will rise or fall with Anthropic's actual usage of Datadog's services. Datadog CFO David Obstler said the company modeled the contract with "very conservative" assumptions in its 2026 guidance because of that pricing uncertainty. As a benchmark for scale, Wells Fargo estimated last year that Datadog's contract with OpenAI represented roughly 4–6% of Datadog's revenue run-rate, or approximately $200 million annualized.

Why does the Datadog-Anthropic story matter for the DDOG bear thesis?

A common bear argument on Datadog has been that AI-native companies — which write code fluently and operate their own infrastructure — would simply build observability tooling themselves rather than pay a vendor like Datadog. Hunterbrook's report argues that Anthropic, one of the most technically sophisticated AI labs in the world and a company that Pomel said was already running in-house tools before the deal, appears to have concluded that consolidating onto Datadog was the better path. Whether that generalizes to other AI-native companies is not something the report claims to establish.

Disclaimer: This summary is not primary research and does not constitute investment advice. It is a brief overview of a detailed equity research report authored by the firm, organization, or source referenced in this article or at https://hntrbrk.com/datadog-anthropic/?ref=shortreport.fyi, which contains extensive evidence, regulatory filings, and analysis; readers are encouraged to review the full report there for a comprehensive understanding. The content provided in this publication is not authored or originated by us — we act solely as a distributor and do not endorse, verify, or take responsibility for the accuracy, completeness, or reliability of the information presented. This publication is for informational purposes only and should not be construed as legal, business, investment, or tax advice. Always conduct independent due diligence and consult qualified professionals before making any decisions based on the information contained herein. We disclaim all liability for any loss or damage arising from reliance on third-party content, and the views expressed are solely those of the respective source and do not necessarily reflect our own.