A survey of 300 enterprise IT decision-makers in North America and Western Europe, published today, finds respondents are anticipating on average a 9.5x increase in the amount of telemetry data being collected over the next two years as organizations move into the agentic artificial intelligence (AI) era.

Conducted by Omdia/Informa TechTarget on behalf of Apica, a provider of a platform for managing telemetry data, the survey finds 44% are expecting anywhere from a 6x to 100x growth in telemetry data. In total, well over a third of respondents (35%) report that deployments of AI agents are already widespread.

About two-thirds of respondents said they are either “somewhat prepared” or unprepared to manage telemetry data at scale, with 22% saying they have not begun to consider the data implications.

Andi Mann, chief product and technology officer for Apica, said the survey makes it clear that telemetry data management costs are already starting to spiral out of control. As those costs rise, many organizations are going to discover that the cost of observability will increase to unsustainable levels, he added. In fact, if IT leaders don’t get ahead of reining in those costs, many of them will ultimately wind up losing their jobs, said Mann.

In general, more than half of respondents (54%) report that their telemetry data volume has tripled in the past 12 months, with estimated average growth of 3.7x year over year. AI and machine learning workloads now account for approximately 43% of that growth.

Not surprisingly, 83% of respondents ranked AI observability as a top priority for 2026, with 68% working for organizations that plan to evaluate observability changes within six months. A total of 70% also plan to evaluate telemetry pipeline tools and platforms, with 54% of respondents already having some type of telemetry data pipeline in place.

On average, survey respondents are spending $3.17 million annually on observability, with 81% actively looking for ways to cut those costs. Observability budgets are growing an average of 28% year over year, the survey finds.

Most troubling still, 69% of respondents said observability costs for agentic AI projects already exceed compute and infrastructure costs combined. Well over half (59%) said that as a result their organization has either terminated or delayed an agentic AI deployment because observability costs were too high.

It’s not clear to what degree IT leaders are proactively managing the deployment of AI agents versus viewing them as yet another class of applications they will be tasked with managing after they have been deployed. The one thing that is certain is that they will ultimately be held responsible for managing the total cost of the IT infrastructure needed to build, deploy, manage and secure agentic AI applications.

The challenge, of course, is that no one really has a firm handle on what those costs might actually be, which makes planning for an uncertain future today nearly impossible. In the meantime, however, IT leaders might want to assume the worst because, as far as most CFOs are concerned, the only thing worse than overestimating the cost of anything is underestimating it.