Dynatrace has signed a definitive agreement to acquire Arize for $915 million, in a deal aimed at extending its observability platform into one of the fastest-growing areas in enterprise AI deployment: understanding not only whether infrastructure and applications are functioning correctly but also how models, agents, and systems based on large language models (LLMs) behave once they go into production.
The key points of Dynatrace’s acquisition of Arize in 20 seconds
- Dynatrace has agreed to acquire Arize for $915 million, primarily in cash.
- Arize develops tools to evaluate and monitor AI models, LLMs, and agents.
- The transaction aims to connect AI evaluation, applications, infrastructure, and business outcomes.
- Dynatrace estimates the AI observability market will surpass $10 billion by 2030.
- The closing is expected this quarter or early in Dynatrace’s third fiscal quarter.
The transaction is still subject to regulatory reviews and other customary conditions, so it should be regarded as an agreed-upon deal pending completion, rather than a finished acquisition. Dynatrace expects to pay approximately $815 million in cash, in addition to assuming certain employee stock incentives for those joining the company from Arize.
Beyond the amount, the deal illustrates how the observability market is changing. For years, these platforms focused on answering relatively well-known questions for operations teams: which application is failing, where latency appears, which resource is saturated, or which dependency is causing an issue.
Generative AI adds another layer. An application can be technically available but still deliver incorrect responses, experience degraded model quality, or trigger an unexpected spike in consumption and costs.
From server monitoring to understanding what AI agents do
Arize’s proposal specifically aims to cover that gap.
The company is developing an AI observability and LLM evaluation platform geared toward both traditional machine learning and generative systems. Its tools allow analyzing model and agent behavior, detecting issues, investigating causes, and tracking performance over time.
Dynatrace wants to integrate this layer with its existing observability of applications and infrastructure.
The goal is to be able to follow an AI application throughout a broader lifecycle: from testing and evaluation before deployment to its behavior once it begins serving real requests.
This is especially important for agents.
Traditional applications often run relatively deterministic flows. Agent-based systems, however, can operate differently: a model decides which tool to use, queries information, calls external services, performs actions, and then reasons over the results.
When something goes wrong, pinpointing the cause becomes more complex.
The problem may lie in the prompt, the chosen model, data retrieved via retrieval-augmented generation (RAG), an external tool, application code, a database, or directly in the infrastructure.
Dynatrace specifically highlights that fragmentation as one of the issues they aim to address with Arize. Currently, AI engineers can evaluate models and agents using some tools while SRE (Site Reliability Engineering), platform, and operations teams monitor applications and servers with others.
This creates a difficult-to-investigate zone between the two worlds.
What Dynatrace aims to achieve with Arize
The planned integration seeks to provide a common context that connects AI behavior with application and infrastructure performance.
For example, if a service based on an LLM degrades, such a platform could help determine whether the root cause relates to response quality, data changes, an external call, application latency, or issues with underlying compute resources.
Infrastructure again gains importance here.
AI services consume GPUs, CPUs, memory, storage, and network bandwidth. A response time increase might originate from completely different issues—such as a model hallucination or infrastructure bottleneck—but both affect the end user of the application.
Dynatrace envisions that the integration will enable relating these layers and linking them to business processes.
The company outlines three main capabilities once the acquisition is complete: continuous coverage from experimentation to production; a unified context connecting AI behavior with infrastructure; and an enterprise database to analyze large-scale AI workloads.
It’s important to distinguish between the current features available on both platforms and the capabilities Dynatrace expects to develop after completing and integrating the deal. The company describes the latter as future objectives and explicitly notes that these projections could change.
Arize also contributes an open-source community
The acquisition includes an additional component beyond enterprise technology.
Arize maintains Phoenix, an open-source project used for AI observability, traceability, and evaluation.
This gives Dynatrace a foothold among AI developers and engineers who may not initially choose a corporate observability platform.
This trend is increasingly common in developer tools: technical decisions often start within a team via an open project, then expand into enterprise products as governance, support, security, or scaled deployment needs emerge.
Dynatrace believes that this developer presence complements its position among large enterprises.
For Arize, this access to a broader organization with significant commercial infrastructure and a sizeable client base is valuable.
Its founders, Jason Lopatecki and Aparna Dhinakaran, will join Dynatrace upon deal closure. Lopatecki will continue leading the Arize team and will report directly to Dynatrace CEO Rick McConnell.
AI observability is becoming a distinct market
Dynatrace estimates the AI observability market will exceed $10 billion by 2030. This forecast, included in the announcement, should not be interpreted as a realized figure but as an expected growth trajectory.
However, the growth of this niche addresses a real challenge.
Testing chatbots internally is relatively simple. Managing thousands or millions of conversations, model calls, and agent executions while maintaining quality, latency, and cost control is much more complex.
Traditional monitoring can confirm an API responds with HTTP 200 and that infrastructure operates within expected parameters.
But it doesn’t automatically verify if the generated responses make sense.
That’s why specialized tools are emerging to evaluate aspects like responses, traces, hallucinations, reply quality, agent performance, and model consumption.
As companies transition projects from testing to production, these metrics increasingly blend with traditional monitoring.
Making an additional call to a model can improve a response but also increase latency and costs. Using a smaller model might reduce expense but must be checked for sufficient quality for the use case.
Thus, observability is moving beyond simply asking if the service is working. In AI applications, it’s also about understanding what the system is doing, the quality of responses, latency, and costs involved.
An acquisition anticipated to impact Dynatrace’s financials
Dynatrace has also outlined how it expects the acquisition to influence its results.
The company estimates the operation will add approximately roughly 200 basis points to annual recurring revenue (ARR) growth during its fiscal year 2027.
At the same time, it predicts a negative impact of about 175 basis points on its non-GAAP operating margin over the same period. Dynatrace plans to recover margins gradually during 2028 and subsequent years.
These projections are company estimates and depend, among other factors, on the deal closing and smooth integration.
Dynatrace states it will fund the deal through available cash and/or existing credit lines. It also does not anticipate a material impact on its forecasts for Q2 of the 2027 fiscal year or on its current share buyback program.
The deal comes at a time when observability must adapt to significantly more complex application architectures.
Models no longer necessarily operate as isolated APIs within a single program. They coordinate tools, data, and other systems via agents, increasing the number of potential failure points and complicating root cause analysis.
Dynatrace is betting $915 million that monitoring infrastructure and applications alone will no longer suffice for a growing segment of its clients.
If completed, Arize will allow them to pose one more question to operations teams: beyond checking if the application runs, whether the AI is functioning properly.
Frequently Asked Questions
How much will Dynatrace pay for Arize?
Dynatrace has valued the deal at $915 million. Approximately $815 million will be paid in cash, with the remainder in stock incentives for Arize employees joining Dynatrace.
What does Arize do?
Arize develops observability and evaluation technology for AI. Its tools allow analyzing machine learning models, LLM applications, and AI agents during development and after deployment.
Has Dynatrace already acquired Arize?
No. The companies have signed a definitive agreement, but the acquisition has yet to complete regulatory reviews and other closing conditions. Dynatrace expects to finalize it during this quarter or early in its third fiscal quarter.
What is AI observability?
It encompasses techniques and tools designed to analyze the behavior of AI models and applications. This can include traceability of agents and LLMs, response evaluation, problem detection related to quality, and connections to application and infrastructure performance.
via: dynatrace
