· Charlie Holland · DevOps  · 4 min read

The DevOps Bubble Is Bursting — Good

We never really needed ten thousand DevOps engineers. The market's not dying — it's clearing out the noise. When the dust settles, the builders will still be here.

We never really needed ten thousand DevOps engineers. The market's not dying — it's clearing out the noise. When the dust settles, the builders will still be here.

“If AI takes your DevOps job, maybe it wasn’t real work anyway.”

That’s not me saying it — that’s Sam Altman. And while I don’t agree with everything that comes out of Silicon Valley, he’s not entirely wrong on this one.

DevOps was never a job title

Let’s go back to basics. DevOps isn’t a role. It never was. It’s a philosophy — shared ownership of the whole delivery lifecycle, faster feedback loops, less theatre between the people who write the code and the people who run it. That’s what Patrick Debois had in mind when he organised the first DevOpsDays conference in Ghent in 2009. Done properly, it’s just solid engineering.

But then came the gold rush.

Certifications appeared for things that used to be called “knowing how to deploy your own code.” “DevOps Engineer” became a job title, then a career path, then an entire industry of training courses, bootcamps, and conference circuits. “DevOps as a Service” became a thing you could buy from a consultancy. Dashboards appeared with more dials than a 747 cockpit, monitoring things nobody looked at.

Everyone got a fancy title and a Terraform logo on their hoodie.

The market’s flooded — and the cracks are showing

Now the bill’s coming due. The market is saturated with people who can write a Terraform module but can’t debug a networking issue. Who can set up a CI pipeline from a template but can’t design a deployment strategy for a system that actually matters. Who have five cloud certifications and have never been paged at 3am.

Every “cloud expert” is chasing the same ten jobs. Recruiters can’t see the wood for the trees. Companies think they can pay 2010 rates for 2025 skills because there are so many CVs in the pile.

The skill requirements have become absurd. I regularly see job specs asking for deep expertise in ten tools, three cloud platforms, Kubernetes, service mesh, GitOps, infrastructure as code, observability, security, and — oh — “AI/ML experience preferred.” For a mid-level salary. The market has lost its mind.

Meanwhile, the work still needs done

Here’s the thing that gets lost in the noise: real systems still break. Real infrastructure still needs designing, building, and operating. Real engineers still fix things when they go wrong.

I’ve been contracting for nearly 30 years. Some of the rates I’m hearing now are lower than when I started. That’s not because the work is easier — it’s because the market can’t distinguish between someone who’s operated production Kubernetes clusters at HSBC and someone who completed a Udemy course last month.

That’s a signal problem, not a demand problem.

AI will accelerate the shakeout

Here’s where Altman’s comment lands. AI is genuinely good at the mechanical parts of DevOps — writing boilerplate Terraform, generating pipeline configurations, suggesting monitoring queries, even debugging straightforward deployment failures. If that’s the entirety of your job, then yes, AI is coming for it. Probably soon.

But AI is terrible at the things that actually matter in production:

  • Understanding context. Why does this system exist? What are the business constraints? Why was this particular architectural decision made three years ago, and what breaks if you change it?
  • Making trade-offs under pressure. The database is failing, two services are degraded, and the business needs to process payroll in an hour. What do you triage first? AI can suggest options. It can’t own the decision.
  • Designing for the team you have. The technically optimal solution is meaningless if the team can’t maintain it. That judgement — what’s appropriate for this organisation, at this stage of maturity, with these people — is deeply human.
  • Saying “we shouldn’t do this.” AI is an eager assistant. It’ll help you build anything you ask for. It won’t tell you that the thing you’re building is a bad idea, that you’re over-engineering a simple problem, or that the real issue is organisational, not technical.

The engineers who can do these things — who can think, not just execute — will be more valuable, not less. The ones who were essentially human script runners? They’ll need to adapt.

The philosophy holds

Strip away the certifications, the tooling fetishism, and the inflated job titles, and what’s left is the original idea: engineers who own the full lifecycle of their software, from commit to production and beyond. People who care about reliability, who automate the boring stuff so they can focus on the hard stuff, who treat operations as a first-class engineering discipline.

That philosophy hasn’t changed. It doesn’t need a certification. It doesn’t care what cloud you’re using. And it’s not going away just because AI can write a Dockerfile.

When the noise settles and the bubble finishes bursting, the builders will still be here. The question is whether the market will get better at finding them — or whether we’ll just start the next hype cycle and do it all again.

I know which one I’d bet on.

Back to Blog

Related Posts

View All Posts »
Ops in a Frock: Why Most SRE is Still Theatre

Ops in a Frock: Why Most SRE is Still Theatre

Pipeline jockeys cranking YAML. Firefighting teams on endless rota. Developers throwing features over the wall. Call it SRE, call it Platform Engineering — if nobody owns production, it's just ops in a frock.

Iteration and Agility

Iteration and Agility

Since the 1950s we've been looking for better ways to build software. Iterative development, continuous integration, and SaaS delivery have fundamentally changed the feedback loop — and the winners are those who embrace it.

Claude Code Is Brilliant. It's Also Forrest Gump.

Claude Code Is Brilliant. It's Also Forrest Gump.

AI agents will get you to dev-done at record speed — right up until you realise you're miles past the actual problem. 80% of projects still spend 80% of their time at 80% complete. Agents don't fix that. They just help you reach it sooner.

Scrum Just Trained Your Replacement

Scrum Just Trained Your Replacement

Tiny Jira tickets. Daily check-ins. Hand-holding. Scrum built the perfect assembly line — and AI loves assembly lines. The question isn't whether AI can do the tickets. It's whether your team is more than the sum of its tickets.