AI Is Hitting Operational Limits as Companies Rush to Scale, Datadog Report Finds

Datadog (NASDAQ: DDOG) finds AI operational complexity, not model intelligence, is the main barrier to scaling AI reliably.

Rhea-AI Summary

Datadog (NASDAQ: DDOG) finds AI operational complexity, not model intelligence, is the main barrier to scaling AI reliably.

Key data: 69% of companies use three+ models, ~5% of AI requests fail in production, and ~60% of those failures are caused by capacity limits. OpenAI holds 63% provider share while Google Gemini and Anthropic Claude rose by 20 and 23 percentage points. Agent framework adoption doubled year‑over‑year and request token volumes rose substantially for median and heavy users.

Positive

• Multi-model adoption at 69% of companies
• Agent framework adoption doubled year-over-year
• OpenAI provider share remains high at 63%

Negative

• Production failure rate near 5% of AI requests
• Capacity limits cause ~60% of production failures
• Token payload growth doubled for median users and quadrupled for heavy users

Argus Apr 21 session

-0.35% close to close Open Argus

Details

News Market Reaction – DDOG

On Apr 21, the day this news came out, DDOG closed 0.35% below the previous close.

Data tracked by StockTitan Argus for the Apr 21 session.

Key Figures

AI request failure rate: Nearly 1 in 20 requests Failure rate percentage: Around 5% of AI model requests Capacity-related failures: Nearly 60% of failures +5 more

AI request failure rate

Nearly 1 in 20 requests

Production AI requests in Datadog report

Failure rate percentage

Around 5% of AI model requests

Production failure rate cited in report

Capacity-related failures

Nearly 60% of failures

Share of AI request failures due to capacity limits

Multi-model adoption

69% of companies use 3+ models

Organizations running AI in production

OpenAI usage share

63% share

Most widely used AI provider in Datadog dataset

Google Gemini adoption change

Grew by 20 percentage points

Change in provider adoption in report period

Anthropic Claude adoption change

Grew by 23 percentage points

Change in provider adoption in report period

Token growth for heavy users

Quadrupled tokens per request

90th percentile AI usage teams

Previous AI Reports

5 past events · Latest: Mar 23

Same Type 5 events

1. Mar 23

   AI product launch

   24h Move

   +3.3%

   Launch of Bits AI Security Analyst to accelerate threat investigations.

2. Mar 09

   AI infrastructure launch

   24h Move

   +2.2%

   GA of MCP Server for secure, real-time observability access for AI agents.

3. Feb 18

   AI conference announcement

   24h Move

   -0.6%

   Announcement of DASH 2026 focused on AI observability and security.

4. Dec 03

   AI cloud collaboration

   24h Move

   -0.4%

   Strategic Collaboration Agreement with AWS highlighting AI capabilities.

5. Dec 02

   AI agent launch

   24h Move

   -0.9%

   Launch of Bits AI SRE agent for autonomous incident investigation.

24h Move is the share-price change in the day after each event; other market factors may also have contributed.

Key Terms

agent workflows, agent framework, ai observability, agentic infrastructure, +2 more

6 terms

agent workflows technical

"companies (69%) now use three or more models alongside increasingly complex agent workflows."

Agent workflows are sequences of automated tasks carried out by software “agents” that follow preset rules or learn from data to complete business processes, such as routing customer requests, analyzing documents, or executing routine trades. For investors, they matter because they can speed operations, cut labor costs, reduce human error and scale activity quickly—affecting a company’s productivity, profit margins and operational risk much like adding an extra efficient shift on an assembly line.

agent framework technical

"Agent framework adoption doubled year-over-year, accelerating development but also introducing more moving parts"

An agent framework is a software system that coordinates autonomous programs (agents) to carry out tasks like gathering information, making decisions, or executing trades. Think of it as a team manager that assigns jobs, passes messages, and enforces rules so each software “worker” can act on its own but still contribute to a larger goal. Investors care because these frameworks can speed research, automate trading and compliance, and scale operations—affecting costs, speed, and risk.

ai observability technical

"In this new era, AI observability becomes as essential as cloud observability was a decade ago."

A set of tools and practices that let organizations monitor and understand how their AI systems behave in real time — tracking things like model accuracy, unexpected outputs, data changes, errors, and resource use. For investors, observability is like a health dashboard for AI: it reduces the chance of hidden failures, regulatory problems, or costly mistakes, and signals whether management is effectively managing operational and reputational risk tied to AI deployments.

agentic infrastructure technical

"We built agentic infrastructure at Vercel because agents need the same production feedback loops"

Agentic infrastructure is the collection of software, hardware, networks and rules that let autonomous digital agents—software programs that can make decisions and act—operate reliably at scale. It matters to investors because it is the foundation that enables automation, new revenue streams and efficiency gains, while also concentrating technological and regulatory risks much like roads and traffic signals determine how cars can move and where congestion or accidents may occur.

llm technical

"Unlike traditional software, agents have control flow driven by the LLM itself, making observability not just useful, but essential."

A large language model (LLM) is an advanced computer system trained on vast amounts of written text to understand and generate human-like language, similar to a very fast, well-read assistant that can summarize documents, draft messages, or answer questions. Investors care because LLMs can speed up research, automate customer support, and reduce costs, while also creating new product opportunities and risks around accuracy, bias, and regulatory oversight that can affect a company’s performance.

gpu utilization technical

"need real-time visibility across the entire stack – from GPU utilization to model behavior to agent workflows."

GPU utilization is the percentage of a graphics processing unit’s available computing power that is being used at a given time. For investors, it signals how much demand or efficiency there is for workloads that need heavy computing—like AI training, data processing, or graphics rendering—and can affect revenue potential, operating costs and capacity planning; think of it as how full a delivery truck is when measuring how well resources are being put to work.

AI-generated analysis. How Rhea-AI works. Not financial advice.

See more from StockTitan in Google Search and AI answers. Adds StockTitan as a preferred source · opens Google

Add on Google Not now

Nearly 1 in 20 AI requests fail in production as capacity limits become the primary bottleneck to scaling AI reliably

NEW YORK, April 21, 2026 (GLOBE NEWSWIRE) -- As AI adoption accelerates, operational complexity – not model intelligence – is becoming the primary barrier to reliable AI at scale, according to new data from Datadog, Inc. (NASDAQ: DDOG), the AI-powered observability and security platform.

Datadog’s State of AI Engineering 2026 report, based on real-world data from thousands of organizations running AI in production, highlights a compounding complexity challenge as AI systems scale. Nearly seven in ten companies (69%) now use three or more models alongside increasingly complex agent workflows. Around 5% of AI model requests fail in production, with nearly 60% of those failures caused by capacity limits – leading to slowdowns, errors, and broken experiences in AI-powered applications.

Additional key findings:

• Multi-model is now the norm: OpenAI remains the most widely used provider at 63% share, alongside rising adoption of Google Gemini and Anthropic Claude which grew by 20 and 23 percentage points, respectively.
• Agent framework adoption doubled year-over-year, accelerating development but also introducing more moving parts into production systems.
• The amount of data sent to AI models per request is also rising: the average number of tokens more than doubled for ‘median use’ teams (50th percentile of usage volume) and quadrupled for heavy users (90th percentile).
  

“AI is starting to look a lot like the early days of cloud,” said Yanbing Li, Chief Product Officer at Datadog. “The cloud made systems programmable but much more complex to manage. AI is now doing the same thing to the application layer. The companies that win won’t just build better models - they’ll build operational control around them. In this new era, AI observability becomes as essential as cloud observability was a decade ago.”

Speed Requires Control

Competitive pressure is accelerating AI deployment across startups and large enterprises alike. But as systems scale, speed without control creates risk. Failures are increasingly driven by system design, including fragmented workflows, excessive retries, and inefficient routing.

"The next wave of agent failures won't be about what agents can't do but what teams can't observe,” said Guillermo Rauch, CEO at Vercel, the company behind Next.js and a leading platform for building AI-powered web applications. “We built agentic infrastructure at Vercel because agents need the same production feedback loops as great software. Unlike traditional software, agents have control flow driven by the LLM itself, making observability not just useful, but essential.”

“Innovation alone isn’t enough,” added Li. “To scale AI with confidence, organizations need real-time visibility across the entire stack – from GPU utilization to model behavior to agent workflows. Visibility and operational control are what allow teams to move fast without sacrificing reliability or governance. At scale, how you operate AI may matter more than the models you choose.”

Read the full report - The State of AI Engineering 2026 - and learn how Datadog is investing in AI observability to help teams operate and scale AI systems in production here.

Report Methodology
Datadog analyzed anonymized usage data from thousands of customers using LLMs in production environments, with global coverage across industries and geographies.

About Datadog

Datadog is the AI-powered observability and security platform. Our SaaS platform integrates and automates infrastructure monitoring, application performance monitoring, log management, user experience monitoring, cloud security and many other capabilities to provide unified, real-time observability and security for our customers' entire technology stack. Datadog is used by organizations of all sizes and across a wide range of industries to enable digital transformation and cloud migration, drive collaboration among development, operations, security and business teams, accelerate time to market for applications, reduce time to problem resolution, secure applications and infrastructure, understand user behavior and track key business metrics.

Forward-Looking Statements

This press release may include certain “forward-looking statements” within the meaning of Section 27A of the Securities Act of 1933, as amended, or the Securities Act, and Section 21E of the Securities Exchange Act of 1934, as amended including statements on the benefits of new products and features. These forward-looking statements reflect our current views about our plans, intentions, expectations, strategies and prospects, which are based on the information currently available to us and on assumptions we have made. Actual results may differ materially from those described in the forward-looking statements and are subject to a variety of assumptions, uncertainties, risks and factors that are beyond our control, including those risks detailed under the caption “Risk Factors” and elsewhere in our Securities and Exchange Commission filings and reports, including the Quarterly Report on Form 10-Q filed with the Securities and Exchange Commission on February 18, 2026, as well as future filings and reports by us. Except as required by law, we undertake no duty or obligation to update any forward-looking statements contained in this release as a result of new information, future events, changes in expectations or otherwise.

Contact:
press@datadoghq.com

FAQ

AI-generated questions and answers. How Rhea-AI works. Not financial advice.

What does Datadog's April 21, 2026 report say about AI production failure rates for DDOG customers?

About 5% of AI model requests fail in production. According to the company, nearly 60% of those failures stem from capacity limits, causing slowdowns, errors, and broken experiences in AI applications.

How common is multi-model usage among organizations in Datadog's State of AI Engineering 2026 report?

69% of companies now use three or more models in production. According to the company, multi-model deployments and agent workflows are becoming the standard, increasing operational complexity.

What provider market shares does Datadog report for AI model usage affecting DDOG customers?

OpenAI holds 63% share among providers, with Google Gemini and Anthropic Claude rising significantly. According to the company, Gemini and Claude grew by 20 and 23 percentage points, respectively.

How has agent framework adoption changed per Datadog's April 21, 2026 findings and what does that imply for DDOG?

Agent framework adoption doubled year-over-year, accelerating development. According to the company, this increases moving parts in production and raises the need for stronger observability and operational controls.

What does Datadog report about token usage per AI request and implications for DDOG customers?

Token volumes rose markedly: median teams doubled tokens and heavy users quadrupled them. According to the company, larger payloads increase compute and capacity demands, stressing infrastructure and costs.

How do capacity limits affect AI reliability according to Datadog's State of AI Engineering 2026?

Capacity limits are the leading cause of production failures, responsible for about 60% of failed requests. According to the company, capacity-related slowdowns and retries degrade user experiences and require operational fixes.