What AI is really doing to tech jobs

The causes behind recent layoffs, what the evidence says about AI, and why the clearest effect so far is fewer junior hires.

Also called layoffs.fyi, layoff tracker, tech layoff numbers

What: Sites and reports that count announced layoffs, such as layoffs.fyi for tech companies and Challenger, Gray and Christmas for all US employers.
Why: They turn headlines into numbers you can compare by year, company and month. Each counts differently, so always name the source of a figure.
Example: By layoffs.fyi counts, tech layoffs were about 266,000 in 2023, about 153,000 in 2024 and about 123,000 in 2025, and 2026 passed the 2025 total by September.

Watch out

Comparing numbers from two trackers as if they measured the same thing.

Also called AI capex, AI spending

What: Reducing headcount in some areas to pay for heavy spending on AI data centres, chips and teams.
Why: It explains why profitable companies still cut: the money is being moved, not lost. People in older products and support functions are most exposed.
Example: Oracle said in a 2026 filing that adopting AI had reduced, and may continue to reduce, its workforce.

Also called delayering, fewer layers, removing managers

What: Removing layers of management so each manager has more reports, and some managers become individual contributors again.
Why: Managers of small teams and roles that mainly coordinate are exposed in a flattening, whatever their individual performance.
Example: Google said in 2025 it had about a third fewer managers of small teams than a year before.

Also called AI as the reason, AI-washing layoffs

What: Naming AI as the reason for layoffs that are mainly about cost, over-hiring or weak sales, because it sounds better to investors.
Why: Knowing it exists helps you read announcements calmly. Crunchbase counted AI as a stated reason in about a third of 2026 layoff events, against about 1% in 2024.

Watch out

Believing either that AI is behind every cut or that it is never the real reason.

Also called entry-level squeeze, canaries in the coal mine, fewer junior hires

What: The drop in hiring of early-career engineers since 2022, linked in research to AI doing more entry-level tasks.
Why: Stanford found employment of US developers aged 22 to 25 fell nearly 20% from its late-2022 peak, mainly through fewer hires rather than firing. Proof of real work matters more for a first job.

Watch out

Concluding juniors are no longer needed. Companies still need the seniors of 2030.

Also called AI exposure, exposure to AI

What: How much of a job is made of tasks that AI tools can already do well, such as boilerplate code, first drafts of tests and simple fixes.
Why: AI takes tasks before it takes whole jobs. Knowing which of your tasks are exposed tells you what to move away from and what to strengthen.
Example: An engineer who spends most of the week on small, well-defined tickets is more exposed than one who spends it debugging production and designing systems.

What: AI substitutes for work when it can do the task instead of a person; it complements work when it makes a person faster or better at a task that still needs them.
Why: Research found employment falling where AI substitutes and holding or growing where it complements, especially for experienced workers.

Also called adjacent areas, growing areas

What: Growing skills that sit next to what you already know, such as AI features for a backend developer or data pipelines for an analyst.
Why: Moving one step sideways keeps your existing value while adding demand. It is faster and safer than jumping to a completely new field.

Watch out

Chasing the hottest job title instead of the growing skill closest to your own work.

Also called news diet

What: Reading bad news for long periods, often on social media, without acting on any of it.
Why: It raises anxiety and lowers focus without improving your position. A weekly check of a few data-based sources is enough.