Are Junior Developers Still Needed? What the Data Actually Shows

Every few weeks another headline says junior developers are going extinct. The numbers behind those headlines are real. But after two years of actually hiring for a team, I don’t think “AI took the jobs” is the right read on them — and the wrong read leads to the wrong decisions, whether you’re hiring or job-hunting.

The numbers that are real

Entry-level developer postings are down roughly 67% since 2022, and actual hiring into those roles has dropped even further, by around 73% (Hakia). Computer science graduates are carrying a 6.1% unemployment rate — high for a STEM field — and employment for engineers aged 22–25 has fallen close to 20% from its late-2022 peak (CodeToDeploy). Big Tech junior hiring specifically is down about 25% from 2023 levels, and time-to-hire across the board has stretched from roughly 38 days to 67 (Final Round AI).

None of that is spin. If you graduated in the last two years and the job search felt harder than it was supposed to be, it wasn’t just you.

The part that doesn’t fit the “AI did it” headline

Here’s the detail that gets dropped from most of these posts: when researchers try to isolate AI’s actual effect, it accounts for roughly 9–10 percentage points of the junior employment drop, measured six quarters after a company adopts generative AI tools (CodeToDeploy). That’s meaningful, but it’s a fraction of a 67–73% collapse. The rest is the much less exciting story of budget cuts and interest rates — the same forces that hit every cost center in a downturn, not something unique to people who write code for a living.

There’s also a genuine paradox in the postings themselves: job listings labeled “entry-level” actually grew about 47% between late 2023 and late 2024, while real hiring into those levels kept falling (Hakia). Some companies are posting junior reqs and quietly filling them with experienced engineers — for a title, for a pipeline, or because a recruiter’s dashboard looks better with the role open. That’s not a hiring freeze. It’s a hiring theater problem, and it’s worth knowing the difference if you’re the one applying.

It’s not one market — it’s two

The clearest signal in all of this is that “the market” isn’t a single thing right now. Consumer apps and SaaS, where a lot of the work really is CRUD screens and API glue that a capable AI assistant can draft a first pass of, are cutting junior headcount hard. Enterprise software, financial infrastructure, healthcare platforms, and anything with a decade of internal systems and compliance requirements are still hiring juniors — because someone has to become the senior engineer who understands that codebase in 2032, and that person doesn’t get hired in fully formed (danilchenko.dev). IBM is a public example: it reportedly tripled its junior developer intake in 2026, but restructured the role around reviewing and correcting AI output rather than writing everything from a blank file (danilchenko.dev).

That last part is the actual shift, and it’s the one I’d pay attention to.

We’ve had this conversation before

Every wave of “this technology ends the entry-level job” has looked a little different and rhymed a lot. Offshoring was supposed to end junior hiring in the US in the 2000s; it reshaped where junior roles lived more than it eliminated them. The 2008 downturn produced its own round of “nobody’s hiring juniors anymore” — which was true for about three years and then wasn’t. What made each of those cycles survivable for the people in them wasn’t ignoring the disruption. It was figuring out early which part of the job was actually becoming obsolete, and which part was just changing hands. Right now, “typing out code from a spec” is the part becoming obsolete. “Knowing whether the code is right” isn’t, and may never be — someone still has to hold that judgment, AI-assisted or not.

What’s changing on the team side

I lead a team of five to ten engineers, and the honest answer is that the shape of junior work has changed more than the need for it. A junior engineer who can prompt, review, and correct an AI-generated pull request — catching the subtle bug, questioning the unnecessary abstraction, knowing when the “clean” AI suggestion quietly breaks an edge case — is more useful to me today than one who can only produce code slowly by hand. That’s a different skill than the one bootcamps and CS programs were teaching three years ago, and it’s a fair amount of why the hiring bar feels higher even though the job title hasn’t changed.

The engineers I’ve hired recently who ramped up fastest weren’t the ones who memorized the most syntax. They were the ones who could read a system they didn’t write, form an opinion about it, and say so — with or without an AI tool in the loop.

If you’re the one hiring

The same shift changes what’s worth optimizing for on the other side of the table. A few things I’ve adjusted in how I hire and onboard junior engineers over the past year:

• The take-home test changed. Watching someone write a sorting function from scratch tells me less than it used to. Watching them review a PR — AI-written or not — and explain what they’d change and why tells me a lot more about whether they’ll be useful in six months.
• Mentorship time didn’t shrink, it moved. Less time explaining syntax, more time explaining why the codebase is shaped the way it is, and why the “obviously simpler” refactor an AI tool suggested would break three things it can’t see. That context is still entirely a human-to-human transfer.
• The ramp-up curve got steeper, not longer. Someone who’s genuinely good at working with AI tools gets productive faster than a junior did three years ago. Someone who treats the AI output as gospel gets slower, because now I’m debugging their code and their blind trust in a tool.

None of this means junior hiring got easier or cheaper. It means the signal I’m actually looking for moved from “can write code” to “can judge code” — and that’s a much harder thing to teach from zero, which is exactly why the engineers who already have it are worth the search.

What to actually look for

If you’re picking where to apply, a few signals are more useful than the company’s size or brand:

• How old is the core system? A codebase with real history needs people who’ll eventually own its context. A greenfield consumer app leans harder on AI-assisted output from day one, with less patience for a slow ramp.
• Does the job description mention reviewing or auditing anything? That’s often a tell that the role is already built around working with AI-generated output rather than pretending it doesn’t exist.
• Regulated or infrastructure-adjacent, even boringly so. Compliance, healthcare, fintech, and internal tooling at larger companies keep training juniors because the alternative — a codebase nobody understands in five years — is a real cost they can see coming.

None of these guarantee an offer. They do shift the odds meaningfully better than applying broadly into the segment of the market that’s actually shrinking.

The practical takeaway

If you’re early in your career, the useful reaction to “junior jobs are disappearing” isn’t panic, and it isn’t denial either. It’s picking your market on purpose. Consumer SaaS is genuinely brutal right now for entry-level roles. Infrastructure, security, data platforms, and regulated industries are still building the next generation of engineers who’ll own their systems — and they’re often less competitive precisely because fewer people are looking there.

The 67% headline is true. It’s just describing one half of the market, and treating it as the whole story is the actual mistake.