Why Does Observability Feel so Expensive? (Because it Is)

If you’ve ever stared at your observability bill and whispered “there’s no way this is real,” congratulations! Your instincts are working. As Charity Majors puts it, “it’s fucking expensive.” In her recent webinar with James Governor of RedMonk, Rethinking Observability Budgets in 2026: Optimizing for ROI, she shared that most companies now spend somewhere between 20 and 25% of their entire infrastructure budget on their observability tools… the range goes from around 10 to 50%.” So if the spend feels wildly out of control, it’s because it is.

Did you miss the webinar? Watch it here!

A “metric fuckton”

Charity reminded us that today’s systems generate a “metric fuckton” of data, which is perhaps the most accurate unit of measurement ever invented. Observability has evolved far beyond the early Nagios-and-Ganglia era, when one Gartner customer spent a quaint $50,000 a year in 2009. Fast forward to today: that same customer now spends $24 million, a number growing 48% year over year for 15 years straight. The complexity of modern systems just isn’t something human brains were built to reason about unaided. As Charity joked while showing a tangled microservices diagram, “nobody can keep this in their head.”

If all observability did was tell you whether a system is up or down, then sure, these costs would be hard to defend. But that’s the whole point: observability should be doing far more than that. The opportunity isn’t operational hygiene; it’s understanding your product in a way that directly drives business outcomes.

Half of observability ties to revenue

Charity talked about how the external, customer-facing half of observability ties directly to revenue. Walmart famously found that “for every 100 milliseconds of page load improvement, it boosted revenue by 1%.” Fast systems convert better, retain better, and frustrate fewer people. Pretty simple. But the internal half of observability—the developer experience side—is where she sees the biggest untapped potential. This is the part that determines whether your teams ship confidently or spend their nights spelunking across five tools trying to understand whether the code they deployed actually works.

She offered Intercom as a north star: “They ship 300 to 500 times a day […] It takes them four minutes from when a developer commits a diff to when it’s tested and rolled out to production.” That kind of speed came from a desire to build systems in a way that developers don’t waste time searching for answers in a sea of fragmented signals.

And if you want to talk about wasted time, few things beat paying for data you don’t need. Charity joked that “a logline is not sacred,” despite years of industry messaging that insisted otherwise. You don’t need to store everything. Charity pointed out that “25% of your request volume is actually health checks,” which is both hilarious and a little depressing. Sampling exists for a reason.

AI gets us a step closer to The Dream

What really animated the conversation was AI. Not because AI replaces observability tools, but because AI finally makes it feasible for developers to interact with their systems through natural language instead of hopping across dashboards. As James put it, “natural language querying is super exciting.” Charity went further, stating that AI turns observability into a feedback loop that lives where developers already live: their editors, their PRs, their Slack channels. The dream is not having to leave your flow state to understand what your code did in production, and AI gets us a step closer to that.

The big theme was this: observability should be an investment, not a tax. Or in Charity’s words: “You should be happy about how much you're spending on it… how much more can I put into that account before I start to get diminishing returns?” If better visibility makes developers faster, users happier, and product decisions smarter, then you’re not managing a cost center, you’re investing in a growth engine.

Observability is for everyone

By the end of the session, the vibe was clear: observability is expensive because modern systems are expensive in complexity, in data, in cognitive overhead. But the teams who treat it as a source of power rather than pain are the ones who get to move fast with confidence. And to that point, Charity and James drove home that observability isn’t just for SREs or developers anymore—it's for everyone, including product managers and execs who all need different views of the same truth.

If you want your org to be more Intercom and less “someone found a single metric costing us $30,000,” the path is pretty straightforward: instrument widely, sample intelligently, collapse your data into something unified, and give developers feedback loops they don’t have to chase. Make it easy for people to know what’s happening in production without needing a PhD in Dashboard Archaeology.