Back
"No human wants to look at billions of traces": Dynatrace closes $915M Arize deal
SiTech AI Team2 min read

"No human wants to look at billions of traces": Dynatrace closes $915M Arize deal

Dynatrace has closed its $915 million acquisition of Arize, pairing infrastructure monitoring with tools that trace, evaluate and debug AI models and agents. Both companies say agents, not humans, will be the primary consumers of observability data.

Dynatrace has completed its $915 million acquisition of Arize, pairing infrastructure monitoring with tracing and debugging for AI models and agents. Dynatrace CPO Steve Tack and Arize CPO Aparna Dhinakaran explained the combination to The New Stack.

Classic observability shows what applications and infrastructure are doing and where they fail. AI agents complicate that: the app can run flawlessly while the agent on top gives the wrong answer or fails its task.

Two sides of observability

Dynatrace has more than two decades in application performance monitoring (APM). Founded in Austria in 2005, it was bought by Compuware for $256 million in 2011 and listed on the NYSE in 2019. Davis, launched in 2017, became its causal AI engine.

Arize launched in 2020 as an AI observability startup for production models; as LLMs and agents spread, it expanded into tracing agent behavior and catching failures conventional monitoring misses. Its Signal agent reviews traces and surfaces fixes.

Signal by Arize

The platforms are complementary: Dynatrace covers infrastructure and SRE teams; Arize tracks model and agent behavior. "We used to have one side of the coin. Now we have both sides to be able to debug," Dhinakaran says.

Who reads billions of traces

The bigger shift, Dhinakaran says, is who consumes the data: telemetry far exceeds human inspection. "No human wants to go look at billions of traces. So how do agents go read your telemetry data?"

Arize already runs this internally: Signal reviews the traces of Alyx, its in-product assistant, spots recurring failures and opens fix PRs; Arize accepts around 65% to 70%; Alyx's engineer now reviews PRs, not traces.

Anthropic used agentic search to review millions of conversations while investigating incidents where Claude reached real third-party systems. Dhinakaran's goal is "self-sustaining, self-maintaining software, but also kind of self-improving," with humans doing the final review.

Buy vs. build, and what comes next

The companies were not formal partners before the deal; Tack says some shared customers had already asked for the two platforms to operate as one.

Arize's assets settled the build-versus-buy question: Phoenix, a source-available platform for tracing AI agents, and OpenInference, which adds AI conventions to OpenTelemetry. "Those are not easy things to build," Tack says; a partnership could not have delivered the same seamless experience.

Arize will keep supporting Phoenix and AX, its enterprise platform, and remains standalone, so it works without Dynatrace. Over time its capabilities will be folded into Dynatrace, linking build-time evaluation with production behavior.

SSiTech

SiTech — AI-powered web development

We build fast, modern websites and bring AI into real business workflows. Have a project or a question? We'd love to help.