Observability’s new era – AI needs and the developer’s role in purchasing open new fronts for Dynatrace in a $915 million gambit

Rick McConnell

We’re aggressive shoppers but very, very disciplined buyers.

An interesting behavioral characterisation from Rick McConnell, CEO of observability specialist Dynatrace, coming as it does off the back of the firm’s announcement of a $915 million takeover of Arize, an AI observability platform provider, a move he, inevitably perhaps, posits as “game-changing”. 

Is it? Assuming this all jumps through the requisite regulatory hoops and closes by  Q3, what’s Dynatrace getting for its money? As per the official announcement blah blah:

AI observability spans the full lifecycle of an AI application, from experimentation and evaluation before release, to tracing how LLMs (Large Language Models), agents, and orchestration layers behave in production. It connects that behavior with application performance, GPU utilization, infrastructure health, and business processes. AI observability delivers the insights AI engineers, developers, SREs (Site Reliability Engineers), and platform teams need to debug and optimize AI behavior and keep AI-powered services accurate, reliable, and cost-efficient.

It is also one of the fastest-growing categories in observability, projected to exceed $10 billion by 2030. Or as McConnell puts it:

On the one hand, we have the evolution of this $82 billion market for what we would think of as core observability or traditional observability, which is now getting supplemented by the end of the decade [with] this $10 billion AI observability space.

Arize, headquartered in San Francisco, provides AI observability and evaluation tools for AI applications to enable teams to detect issues, troubleshoot problems, and improve performance across AI deployments. Again from the corporate blah blah:

Today, AI software delivery is fragmented. AI engineering teams evaluate model and agent behavior in one set of tools, while the teams running the applications and infrastructure beneath them work in another. There is often no shared system connecting how an AI application is evaluated to how it behaves in production, so when output quality slips or a customer transaction fails, the cause can sit anywhere from the prompt to the infrastructure, and there is little feedback to developers. Dynatrace’s acquisition of Arize will eliminate that fragmentation and provide end-to-end observability from development to production.

Following the closing, Arize’s two founders, Jason Lopatecki and Aparna Dhinakaran, will join Dynatrace where Lopatecki will continue to lead the Arize team, reporting directly McConnell.

It matters because…?

As noted above, for his part, the Dynatrace CEO positions this as a big deal. His mile-high pitch: 

From a strategic standpoint, we believe this acquisition will increase our exposure to the high-growth AI observability market, extend our end-to-end observability leadership position, advance our AI observability road map, expand our reach with developers, accelerate our ARR (AnnuAL Recurring Revenue) growth and add a world-class AI-native team to Dynatrace.

The synergy with Arize was simply better than anybody else in the market “by far”, he argues:

We do see them as the category leader in AI observability from a pure-play standpoint. That gives us capabilities, for example, in LLM experimentation, LLM observability, so targeting both AI workloads as well as agentic workflows. It brings us the AI software delivery life cycle of components that we need to supplement the capabilities that we have at Dynatrace in model operations, cloud and infrastructure operations and elsewhere…AI observability spans the full life cycle of an AI-powered application from continuous evaluation before and after release to real-time visibility into how LLMs, agents and orchestration layers behave in production to the impact they have on applications, infrastructure and business outcomes.

Big questions

From a basic user needs perspective, there are three core questions that AI observability needs to address, says McConnell, beginning with whether it is  working just as with traditional workloads, followed by that critical query – is it accurate and delivering output that can be trusted and relied upon with confidence?  Finally,  are agentic systems delivering the outcomes for which they were built?

That last one is a big question for all enterprises to wrestle with. McConnell says:

Enterprises are deploying agents to build software at a pace that wasn’t possible before, and code must be built well, shipped safely and run reliably. Gartner describes the difference in AI observability well in their Innovation Insight on AI observability [behind paywall] from May of this year. They say AI systems fail differently from traditional software. A mis-configured application typically produces an error that is visible, traceable and re-producible. An AI system can fail silently. It can generate confident, plausible sounding outputs that are biased or factually wrong, with no corresponding alert in an infrastructure dashboard or application log.

From Dynatrace’s perspective, this has implications, he continues:

As AI evolves from simple prompts to autonomous multi-step systems and increasingly operates without a human in the loop, visibility becomes increasingly critical. AI observability provides AI engineers, SREs (Site Reliability Engineers), and platform teams with the insights they need to de-bug, optimize performance, control costs and improve accuracy, helping to ensure AI-powered services remain reliable, efficient and trustworthy.

So, given the clear importance of this emerging space, why has Dynatrace chosen to ‘buy, not build’? It’s not that binary a choice, it seems. According to McConnell:

Acquiring Arize will accelerate our own AI observability portfolio and road map. Through this acquisition, Dynatrace will have an even stronger end-to-end observability solution from pre-production to production. Arize adds leading capabilities in AI and agent evaluation, experimentation and agentic workflow optimization across development and production. These complement Dynatrace’s existing strengths in application and AI infrastructure observability, model performance and business impact.

Together, these capabilities position Dynatrace to provide observability across the full AI stack, connecting model and agent quality with application performance, infrastructure health, AI usage and cost and business outcomes. This is not a point solution. It is a portfolio expansion that positions Dynatrace to better capture a greater share of AI spending in this rapidly emerging category.

Bring me…the developer! 

And there’s another important aspect to the acquisition to be factored in, namely that Arize will bring with it a developer audience that the more enterprise-focused Dynatrace would like to get a slice of. McConnell explains:

We see the developer as being a quintessential part of this acquisition that we can then access not just for AI observability, but observability more broadly. We see developers being a core persona, if you will, in the observability decision….The developer is becoming more critical day-by-day in the overall observability decision because of this ‘shift left’ phenomenon or ‘extend left’ phenomenon in the observability decision-making.

It isn’t just about top-down selling to the CXO and IT ops anymore. It is about top-down selling combined with bottom-up selling from the developer. So gaining access to the developer does enable us, we believe, to sell more core observability or traditional Dynatrace observability into the market as well by having a better end-to-end or a more fulsome end-to-end portfolio.

This is a big asset that Arize has to offer, he notes:

Through its widely-adopted open source community, Arize has already earned the trust of AI developers who increasingly influence enterprise technology decisions. This acquisition will give Dynatrace direct access to this important audience while strengthening our position with AI-native companies. Arize’s strong engagement with developers, coupled with Dynatrace’s deep enterprise relationships and global reach will create new opportunities to land, expand and deliver greater value as customers operationalize AI.

Through millions of monthly downloads of Arize’s open source Phoenix platform, we believe there is a significant opportunity to accelerate ARR growth over time by further extending Dynatrace’s reach into developer-led buying motions and AI-native workloads.

There is some overlap among customers between Dynatrace and Arize today, although that’s estimated as being around 20-30%. but there’s also new market opportunities here for the latter, adds McConnell:

We expect access to Dynatrace’s customer base to further accelerate Arize’s trajectory.

Overall, he concludes:

We absolutely see observability entering a new era. And it is a new era in which it isn’t just about business resilience, it is about AI reliability as well, and it is about ensuring that models and agents are working as expected. These areas together are going to represent an explosion in observability opportunities as we look ahead. And for us, we wanted to play offense, and this is an opportunity to accelerate our ability to do so.

My take

We see this as a game-changing moment for Dynatrace.

Some confident assertions from the never-knowingly-under-selling McConnell and on the face of it, this looks like an interesting fit with wins for both parties. The identification of the developer as a critical component of the buy-side element of the observability equation is an interesting thesis that bears further scrutiny over time. Let’s get through the regulatory hoop-jumping and come back to that one later in the year.

Onward!