Can Opentelemetry Save Observability In 2026?

Observability is not surviving up to expectations. As prices proceed to rise, nan symptom is progressively felt.

This twelvemonth has a foreboding consciousness of accrued infrastructure complexity to negociate astatine scale. In guidance to skyrocketing nationalist unreality pricing, galore organizations person shifted toward backstage cloud, often resulting successful moreover much complexity.

At scale, organizations now progressively look multi-cloud and heterogeneous environments spanning nationalist cloud, backstage unreality and on-premises systems. In galore cases, nan costs buildup is seen arsenic prohibitive.

At nan aforesaid time, observability remains a necessity for organizations some ample and small. In 2025, observability players and suppliers reflected a displacement successful really they thief customers usage their solutions, peculiarly by helping them debar astonishment observability bills of $130,000 per period aliases more.  (Stay tuned astir really this happens, since observability metrics mostly scale  overmuch much than compute does.)

Up to 84% of existent observability users struggle pinch nan costs and complexity of their regular monitoring responsibilities, Gartner analysts Pankaj Prasad and Matt Crossley wrote successful their annual hype rhythm study connected observability and monitoring, published successful July. Still, they concluded, observability is uniquely positioned to reside these exertion challenges.

“The halfway rumor is that nan economics of observability person been upside-down for years. Costs turn linearly pinch telemetry volume, but nan worth doesn’t,” Tom Wilkie, CTO astatine Grafana Labs, told me. “What we need, and what we’re now yet seeing, is simply a exemplary wherever costs scales pinch insight, not ingestion.”

Rethinking nan ‘Big Data Lake’ Model

In 2025, nan observability marketplace was often seen arsenic top-heavy and centralized, creating precocious levels of vexation and costs issues for customers while nan worth seemed to beryllium declining.

Large vendors complaint for indexing and retention by throwing everything into a large strategy aliases information lake, Bob Quillin, laminitis and CEO of ControlTheory, told me.  A much contrarian model, he said, involves putting much intelligence astatine nan separator and utilizing a feedback system.

“By distilling information astatine nan edge, companies only nonstop up what nan AI devices need, allowing nan strategy to tally adjacent to existing devices alternatively of trying to beryllium connected apical of them,” Quillin said. “The measurement nan marketplace has developed complete nan past 20 years has been to propulsion everything successful a large strategy and users salary for indexing, retention, each that. But now that we person much intelligent systems, you don’t request to put each nan information up there.”

Indeed, galore platforms still promote teams to cod everything while providing constricted aliases nary transparency into which telemetry information is really valuable.

Bill Hineline, section CTO for Chronosphere, told me: “The resulting astonishment bills and accrued operational overhead required to negociate them person created nan belief that observability itself was a bad investment, erstwhile successful reality nan finance was poorly governed.”

How AI (Plus Context) Can Help

AI, of course, travel into play successful 2025, forcing organizations and platform engineering teams to take really to merge AI. At nan very least, and particularly successful nan devices I’ve tested — specified arsenic Grafana’s AI — it does nary harm. In fact, it tin help, if utilized properly.

However, AI tin lend peculiarly bully support for interpreting log data. Logs are fundamentally a language, and “if you propulsion nan correct accusation into a large connection model,” Quillin said, it tin thief explicate what happened.

Users should beryllium capable to “throw a log into their favourite LLM via back-end APIs,” Quillin said. “This helps bubble up patterns, specified arsenic repeated captious metrics, and assists successful shape study and power charts to abstracted nan awesome from nan noise.

“As much codification is created by AI, nan complexity of logs is not getting simpler, necessitating a measurement backmost to look astatine nan fundamentals of really information is analyzed. Since logs are connection and if you propulsion nan correct accusation into an LLM, it tin really thief explicate what happened pinch nan log – it looks crossed each your logs and bubbles that up and looks astatine a batch of shape analysis.”

In galore ways, Wilkie said, AI further “inflated and mis-set expectations,” successful 2025.

“AI tin beryllium genuinely helpful, but only erstwhile it’s grounded successful discourse alternatively than treated arsenic a magic furniture radiated connected apical of observability,” he said. “The correction galore vendors make is assuming AI adds worth simply because it exists, erstwhile successful practice, AI only useful if nan underlying telemetry is high-quality, well-structured and connected crossed systems.”

The existent opportunity is what was observed successful 2025, Wilkie said: “AI that turns observability from a specialist-only subject into thing each technologist — and moreover non-engineers — tin use.”   

“And importantly, alternatively of burying teams nether much data, AI helps them extract much worth from nan information they already have,” he said. “AI won’t switch operators successful 2026, but it will yet thief them support up pinch nan complexity curve.”

AI applied superficially tin consciousness for illustration a productivity boost without delivering existent operational clarity, David Jones, vice president of NORAM solution engineering astatine Dynatrace, told me.

“Many devices furniture generative AI connected apical of earthy telemetry, which whitethorn assistance pinch querying but doesn’t fundamentally amended understanding,” Jones said. “That’s wherever disorder creeps in: AI without discourse amplifies uncertainty alternatively than reducing it.”

OpenTelemetry: Poised to Make a Big Impact

With nan added complexity astatine each level, and pinch environments becoming harder to negociate done observability, nan obstruction to introduction remains high—not only to understand really to make usage of logs, metrics and traces, which I would reason are not alleged pillars of observability, but essential, intertwined components, on pinch different metrics, that must beryllium utilized together.

This is wherever OpenTelemetry (OTel) has made huge strides successful 2025, becoming much accessible not only crossed different programming languages, but besides crossed metrics, logs and now traces. Thanks to standardization, it has importantly lowered nan introduction threshold, which is why we are seeing a number of caller and increasing observability players offering their solutions.

Modern systems are becoming harder to operate, and that rising complexity is simply a large logic nan obstruction to introduction has stayed high.

“OpenTelemetry has genuinely helped here: by standardizing instrumentation crossed languages and awesome types, it’s removed a tremendous magnitude of friction,”  Wilkie said. “Teams nary longer person to portion together vendor-specific SDKs, and caller observability players tin participate nan marketplace much easy because instrumentation is nary longer nan moat.”

OpenTelemetry helps by giving organizations vendor-neutrality, Wilkie said, allowing orgs to move much easy to vendors that really perceive to and enactment connected their customers’ costs and worth concerns.

Indeed, OpenTelemetry helps “break nan aged dependence connected proprietary agents and opaque pricing,” he said. “It’s erstwhile you brace OTel pinch intelligent systems — which tin trim information by 80 to 90% while expanding nan worth of what remains — you commencement to flip nan equation truthful that costs scales pinch value, not telemetry volume. Observability hasn’t failed; nan business exemplary did.”

However, standards are only useful if they summation wide adoption, while bequest systems “stick astir successful organizations for decades,” Wilkie said.

“The costs of re-instrumentation is prohibitive,” he said. “OTel is making greenfield activity easier, but customers still request to weave together caller and aged systems into a coherent image – and request observability systems that don’t fastener customers into 1 measurement of doing things.”

My stake is that OpenTelemetry will play a important domiciled if its acquisition tin beryllium improved. The complexity of nan devices themselves besides needs to beryllium reduced truthful that observability is not usable only by an organization’s prima in-house observability expert.

Vendors are opening to aftermath up to nan truth that non-technical users should beryllium capable to make usage of observability and telemetry information arsenic good — but we are not location yet. Maybe, 1 day, AI agents will show observability metrics and troubleshoot and foretell early failures pinch nary input from humans. But that time is not going to travel successful 2026.

Closing nan Usability Gap

In nan meantime, nan non-technical stakeholders must besides put clip and effort into learning to construe telemetry data, sloppy of really elemental to usage observability devices become.

“OpenTelemetry is foundationally changing nan imaginable for nan early by democratizing really observability is done. It has go a de facto modular that lowers nan obstruction to entry, arsenic organizations are now standardizing connected it alternatively than conscionable considering it,” Quillin said. “Using an unfastened root instrumentality built connected these standards allows for a little commercial, much charismatic speech astir nan broader problems successful nan industry.”

Wilkie echoed nan conception that OpenTelemetry solves 1 portion of nan problem: standardizing and generating telemetry.

“But nan usability gap, particularly for non-experts, is still there,” he said. “Instrumentation coming still feels excessively overmuch for illustration a tax: excessively galore configuration knobs, excessively galore semantic conventions to remember, excessively galore separator cases erstwhile services behave successful astonishing ways. Standards lick interoperability, not usability. So yes, OTel will play a bigger role, but only if it becomes simpler.”

OpenTelemetry whitethorn not yet “save” observability, but it should spell a agelong measurement by improving standardization for instrumentation and interoperability betwixt different tools. It should besides service arsenic a springboard for providers arsenic they interest little astir integration and tin create existent solutions that empower what observability tin and should do, for not only nan in-house master but each stakeholders.

“When OpenTelemetry is treated arsenic foundational plumbing crossed nan enterprise, observability platforms tin present much consistent, out-of-the-box insights without relying connected heavy customized dashboards,” Hineline said. “Today, galore dashboards are still built by experts for experts, which leads to dashboard sprawl and limits who tin efficaciously usage observability. When mixed pinch a beardown strategy, OpenTelemetry tin materially amended usability by enabling much consistent, reusable views crossed teams.”

Group Created pinch Sketch.