AI Is Already Shrinking the Computer Science Graduate Job Market

Universities are back this month, and for final-year computer science students, that means the 2027 graduate job hunt effectively starts now. The data they’re walking into is uncomfortable: according to the UK’s official Graduate Outcomes survey, the share of computer science graduates landing an actual coding job has fallen from around 40% to 28% in a single year.

That’s not a rounding error or a one-off bad cohort. It’s the sharpest fall of any graduate occupation in the dataset, and it lines up with an independent global study published just weeks ago showing the same pattern in employment data, not just survey responses. Here’s what’s actually happening, and — more usefully — what it means if you’re one of the people this affects right now.

The headline number: coding jobs down from 40% to 28%

HESA’s Graduate Outcomes survey — a census-style study of more than 350,000 former students, run 15 months after they finished their courses in 2024 — is the most comprehensive picture available of what UK graduates actually go on to do. Compiled for the 2027 Guardian University Guide and reported by The Guardian, the data shows:

• The proportion of computer science graduates working as coders or programmers fell from roughly 40% to 28% in a year
• The proportion moving into any graduate-level occupation fell from more than 60% two years ago to 50%
• Coding and software development were the fastest-falling occupations for graduates of any subject, of any degree

The Guardian’s own reporting is careful here: the data shows the pattern, not the cause. AI is the leading explanation experts point to, but a survey of outcomes can’t on its own prove what’s driving them.

Why this isn’t just a UK story — or a coincidence

That’s where a second, independent piece of research matters. In August 2026, researchers at the Stanford Digital Economy Lab published a study using ADP payroll data covering 25 million workers — not survey responses, actual employment records. Their finding: employment for 22-to-25-year-olds in the most AI-exposed occupations is now 19% below where it would be had it kept pace with older or less-exposed peers, up from a 15% gap in July 2025 and 13% back in 2022.

The detail that matters most: this isn’t broad AI-driven job destruction. Employment for experienced workers, and for workers in less AI-exposed roles, has stayed steady or even grown over the same period. The effect is concentrated specifically on young workers in occupations built around what researchers call “codified knowledge” — formal, documented, textbook-teachable knowledge, exactly the kind of thing a language model can absorb and reproduce. Software development sits near the top of that list. Junior developer work, in particular, has historically been heavy on precisely this kind of routine, well-specified task.

Two independent datasets, two different countries, the same shape of result, in the same few months. That’s a considerably stronger basis for concern than either study would be alone.

It’s not just tech — the whole graduate market has tightened

Computer science is the sharpest edge of a broader trend. The Institute of Student Employers (ISE), the body that’s tracked UK graduate recruitment since 1991, reports:

• Graduate hiring fell 8% in 2024/25 — the weakest year since the pandemic
• A further 7% drop is forecast for 2025/26
• The average employer now receives 140 applications per vacancy — up 59% in a single year, and the highest figure ISE has recorded since it began tracking in 1991

Employers are also rebalancing: reducing formal graduate scheme intake while increasing investment in apprenticeships. Sectors still hiring at meaningful volume include the public sector and civil service (Fast Stream, NHS, Teach First), professional services (the likes of PwC, Deloitte and EY), infrastructure and construction, and law.

For anyone graduating this summer or next, most of the big branded graduate scheme intakes for 2026 closed between October 2025 and February 2026 — those seats are already filled. But the market hasn’t shut: well over 900 active early-career roles were still live UK-wide in mid-September, largely through direct-hire mid-market employers and rolling-intake positions that post and fill as headcount allows, rather than fixed annual cycles.

Where computer science graduates are actually landing

The HESA data points to a specific sideways move rather than a dead end: graduates are shifting from pure coding roles into cybersecurity and network engineering — fields that still draw directly on a CS degree, but sit further from the kind of routine, well-documented code-writing that AI tools now handle cheaply. It’s the same underlying skill set, redirected toward work that’s harder to fully codify and hand to a model.

This matches what employers are actually rewarding. Roles in software development haven’t disappeared — CareerMetrics’ occupation data still shows real demand and competitive pay for developers — but the entry point into that career has narrowed, and the roles immediately adjacent to it (security, infrastructure, systems) are absorbing more of the graduates who’d have gone straight into a coding job two years ago.

What this means if you’re heading into the 2027 cycle

1. Don’t treat “computer science degree → coding job” as automatic. It was closer to a 40% shot a year ago; it’s closer to 28% now. That doesn’t mean don’t study CS — it means plan for a wider set of outcomes than the degree marketing implies.

2. Look sideways, not just up. Cybersecurity, network engineering, and infrastructure roles are absorbing CS graduates who’d have gone straight into development roles previously, and they’re less exposed to the specific automation pattern driving this decline.

3. Big branded scheme season is largely over for this cycle — the market isn’t. With most 2026 intakes closed, focus on direct-hire employers, rolling-intake roles, and off-cycle positions rather than only the household-name graduate programmes.

4. Calibrate your salary and outcome expectations to real data, not scheme marketing. Graduate scheme headlines report the salaries of the schemes that still exist, not the median outcome for your subject. Compare your situation against what UK graduates in your field actually earn rather than the outliers.

5. Know where you actually stand. In a market this competitive — 140 applications per vacancy, on ISE’s numbers — understanding whether your CV, projects, and expectations are calibrated to what’s actually landing roles matters more than it used to. Where Do I Stand benchmarks you against real outcomes data rather than guesswork.

The bottom line

This isn’t a panic story — the Stanford data is explicit that AI isn’t wiping out employment broadly, and plenty of CS graduates are still finding good roles, often just not the exact ones the degree used to guarantee. But it is a real, well-evidenced shift, corroborated across two independent datasets in two different countries, landing at exactly the moment this year’s cohort is starting to plan their applications.

The honest read of the data is: the traditional path from CS degree to coding job has genuinely narrowed, the adjacent paths (security, infrastructure, systems) have genuinely widened, and the graduates who adjust their target roles accordingly are the ones the data suggests are still landing well.