The Infrastructure Debt Agentic AI Is About to Call In

I’ve spent more than 20 years building enterprise software, at BMC Software, CA Technologies, Splunk, and now Apica, and one pattern holds across all of it: You do not find out your foundation is wrong during the crisis. You find out when the technical debt comes due. Agentic AI is about to call in a very large debt for a lot of organizations. This week, Apica released findings from a new Omdia research study, commissioned by Apica and fielded by Omdia/Informa TechTarget, surveying 300+ enterprise IT decision-makers across North America and Western Europe. I want to share what the data told us and more importantly, what it means for every organization that has committed to making agentic AI work in production.

The Crisis Is Already Present-Tense

The headline that should stop every CIO in their tracks: 54% of enterprises saw their telemetry data volume triple in the past 12 months. Not in two years. Not as a projection. Right now, in the environments these organizations are running today. The estimated average growth is 3.7x year over year. AI and machine learning workloads now account for approximately 43% of that expansion, making AI the single largest driver of telemetry volume in enterprise environments. And 83% of enterprises rank AI observability as a top priority for 2026, scoring it an 8 or higher on a 10-point scale.

“This is just the beginning of the growth curve. We are at the beginning of the hockey stick.” — Keir Walker, Senior Market Research Analyst, Omdia

Walker is right. The 3.7x growth of the past 12 months is not the peak. It is the warm-up. Because agentic AI hasn’t even arrived at scale yet.

Agentic AI Is the Accelerant. The Numbers Are Staggering.

The finding I keep coming back to, the one that I think every technology leader needs to sit with: Survey respondents anticipate an average 9.5x increase in telemetry data from agentic AI workloads within two years. And 44% expect growth of 6x to 100x. I want to draw your attention to that range, not the average, but the spread. Six times to one hundred times. That is not a forecast; it is an honest admission that nobody can accurately bound this problem. When your most sophisticated enterprise IT leaders cannot put a ceiling on the data volume they expect, that tells you the architecture has to be built for extremes, not averages. Here is what makes it worse: Despite this anticipated surge, nearly two-thirds of organizations are only “somewhat prepared” or worse for the data volumes agentic AI will generate. More than one in five haven’t even considered the data implications.

35% of enterprises report widespread agentic AI deployment — despite the technology having existed as an enterprise category only since 2024.

When I see that number, I think about what it actually means. The Omdia analysts flagged it directly: Organizations claiming broad deployment are reflecting competitive pressure, not genuine infrastructure readiness. Every CEO is saying their company has to be agentic. But the pipelines underneath those agents — the systems that route, filter, govern, and contextualize the telemetry those agents produce — were not built for this moment. The gap is measurable. Organizations unfamiliar with agentic AI are 4.5x less likely to be prepared for the data volumes it generates. The organizations most at risk are the ones least likely to know it.

Observability Costs Have Already Crossed a Line

I have been in technology long enough to know that cost surprises do not come from where you expect them. For agentic AI projects, the surprising cost is not compute. It is not talent. It is observability. The research confirms: In 69% of agentic AI projects, observability costs already exceed compute and infrastructure costs combined. The average enterprise spends $3.17M annually on observability. Nearly 20% exceed $5M. And 81% are actively looking for cost-cutting alternatives — not because they want less visibility, but because legacy platforms cannot sustainably absorb AI-scale data volumes at their current pricing models.

Organizations with 10,000+ employees are 2.8x more likely to spend $5M or more annually on observability, with an estimated average spend of ~$5M at that scale.

Observability budgets are growing 28% year over year on average. For more than a third of enterprises, that growth exceeds 52% annually. Combined with the anticipated 9.5x agentic AI data surge, you have a cost trajectory with no natural ceiling, unless you change the architecture underneath it. The business consequences are already materializing: 59% of organizations have terminated or delayed at least one agentic AI deployment because monitoring costs were too high. The agents most likely to be shelved are not experimental ones — cybersecurity, legal and compliance, and fraud detection top the list of paused use cases, meaning the cost of leaving them unmonitored is measured in business risk, not just IT budget. This is no longer an IT budget conversation. When uncontrolled observability costs are causing organizations to leave AI agents unmonitored or undeployed, you have an enterprise risk management problem. And Torsten Volk from Omdia put it as plainly as I’ve heard it said:

“Scalability is the main reason why people can’t have agentic projects. They have no way of deploying them without exposing themselves to operational, legal, and security risk.” — Torsten Volk, Principal Analyst, Omdia

The Solution Is Already Validated. The Market Just Needs to Move.

Here is what gives me optimism alongside the urgency: The industry has already identified the answer. 54% of enterprises have already implemented a telemetry pipeline solution, and 97% have implemented or are actively evaluating one — confirming the pipeline not as an emerging concept, but as the market’s consensus response to the telemetry data crisis. The performance advantage of pipeline adoption is measurable:

Pipeline adopters are 50% more likely to anticipate, and be prepared for, the scale of data growth agentic AI will demand within 24 months.

Pipeline adopters are 80% more likely to have avoided the operational cost challenges that constrain organizations still in earlier stages of agentic AI deployment.

That second number is the one I find most important. Pipeline adoption is not correlated with readiness by coincidence. It is a defining characteristic of the organizations that have successfully scaled agentic AI. The mature cohort got there, in large part, because they built the pipeline layer first. Pipeline solutions are also delivering across every dimension enterprises care about: multi-cloud and hybrid support (78%), cost reduction (77%), performance at scale (76%), and data quality (76%). These aren’t soft benefits. They are measurable outcomes from architectural decisions that the most advanced enterprises in this study made before the wave hit. The top capabilities buyers are prioritizing in their pipeline evaluation: multi-destination routing (47%), data reduction and sampling (43%), and content-based routing (41%). Every one of those is a core capability of Apica Flow. And nearly 25% of buyers say existing vendor relationships are not a significant factor in their decision. The pipeline evaluation market will be won on capability and thought leadership, not on incumbency.

What Organizations Need to Do Now

The action window is open. 68% of enterprises plan to evaluate changes to their observability solutions within six months. 70% plan to evaluate telemetry pipeline solutions in the same window. The organizations reading this are, in large majority, already in or entering an active evaluation cycle. Based on what the data tells us, and what I’ve seen across 20+ years of enterprise infrastructure, here is what I would tell any technology leader right now:

• Audit your current telemetry architecture against the anticipated 9.5x surge. Most legacy architectures were not designed for this. The audit will tell you where the debt is.
• Treat observability cost as an enterprise risk metric, not just an IT line item. If cost growth is forcing trade-offs on AI agent coverage, that is a risk conversation for the C-suite.
• Evaluate pipeline-first architecture before the action window closes. The data is clear: Organizations that build the pipeline layer first are materially better positioned for the agentic AI scale-up. The 68–70% evaluation window is six months. That is your planning horizon.
• Look for vendor-neutral, multi-destination routing capability. The organizations that are winning are not locking themselves to a single observability platform. They are building an intelligent layer that routes, filters, and governs telemetry before it reaches any downstream system.

The Foundation Comes First

In more than 20 years of enterprise technology, the pattern holds: Organizations that pull ahead in transformative technology cycles are the ones that solve the infrastructure problem before it becomes a budget crisis. They build the foundation that lets them scale without re-platforming. For agentic AI, that foundation is a telemetry data pipeline. An intelligent, vendor-neutral layer that routes, filters, enriches, and governs telemetry data before it reaches any observability or analytics platform. Not an alternative to observability investment — the mechanism that makes continued observability investment viable at agentic AI scale. The research we’re releasing today is not a warning about a future problem. It is a measurement of a present one. 300+ enterprise IT decision-makers told us exactly where the gap is, how large it is, and how fast it’s growing.

The window to build the right foundation is open. The data shows it will not stay open long.

Download the full Omdia/Informa TechTarget research report to benchmark your organization’s agentic readiness and see how leading enterprises are closing the gap: https://www.apica.io/state-of-agentic-ready-observability-infrastructure-report-2026/