Interpreting 2025 Tech Layoffs Without the Hype
Interpreting 2025 Tech Layoffs Without the Hype
This morning I read a popular article on Medium claiming that over 126,000 tech jobs were cut in 2025, and that the layoffs were not random but "precise", targeting specific roles while sparing others. The argument was compelling, confident, and neatly packaged. Maybe too neat.
So I did what any engineer should do when presented with a clean narrative built on big numbers. I checked the data.
The headline figure, around 126k layoffs, is not fabricated. Depending on the tracker and the cutoff date, it is broadly correct. One widely used dataset reports just under 124,000 tech employees laid off in 2025, another lands closer to 127,000. Already here, the first crack appears. This is not a single number. It is an approximation that depends on definitions, geography, and reporting methodology.
Then you look at other trackers and the picture widens fast. Some datasets report closer to 245,000 people impacted in the same year. Not because the situation suddenly doubled, but because "tech", "layoff", and even "employee" are not consistently defined. Public vs private companies, subsidiaries, contractors, international offices, WARN notices versus press releases. Different lenses, different totals.
So the first correction is simple: 126,000 is not the number. It is a number. Good enough to show scale, not good enough to support precise conclusions on its own.
The article then moves to roles. Frontend-only engineers, junior data analysts, manual QA, prompt engineers, middle managers. Crushed. Infrastructure engineers, security, data engineers, platform teams. Safe, or even growing.
This is where verification really matters, because this framing implies intent. As if companies sat down and decided that certain job titles had become obsolete.
The data does not support that level of precision.
When you look at actual layoff breakdowns reported company by company, a different first-order pattern shows up. One of the most heavily hit functions across 2024 and 2025 was not engineering at all, but recruiting and HR. This makes perfect sense if you think operationally. When hiring freezes, talent acquisition scales to zero almost overnight. It is the fastest cost lever available, and it does not touch production systems.
Engineering layoffs did happen, obviously. But they were rarely clean cuts along specialization lines. Many reductions affected engineering, product, recruiting, and adjacent roles at the same time. Backend, frontend, infra, even senior engineers all appear in the same notices. This is not a moral judgment on skill value. It is budget surgery.
So no, layoffs were not random. But they were not a surgical strike against "shallow frontend" either. They were constrained decisions made under pressure, with one dominant rule: cut what you can cut without breaking the business.
That distinction is critical, because it reframes the rest of the argument in a much more useful way.
The original article is strongest not when it talks about job titles, but when it talks about leverage. Once you strip away the rhetoric, what actually survives scrutiny is this: in a contracting market, skills that directly reduce cost, risk, or operational chaos are protected longer than skills that mainly produce optional output.
That is why cloud cost control matters. Not because it is trendy, but because cloud bills do not shrink by themselves during layoffs. Someone has to right-size, kill waste, and make spend predictable. When an engineer can turn an infrastructure change into a measurable financial outcome, that engineer suddenly speaks a language the board understands.
The same logic applies to security literacy. You do not need everyone to be a security specialist, but you cannot afford engineers who casually create exposure. Regulatory pressure, breach costs, and reputational damage all scale badly in downturns. Engineers who understand auth flows, privilege boundaries, and failure modes reduce existential risk, and that makes them harder to justify cutting.
Production ownership follows the same pattern. Being able to ship code is table stakes. Being able to stabilize a system at 2 a.m., isolate a failure, mitigate blast radius, and explain what happened afterward is operational leverage. This is why platform engineering, observability, and SRE-adjacent work keeps expanding in scope even when headcount shrinks. Not because companies love hero culture, but because downtime is expensive.
Data engineering fits here too, but again not for ideological reasons. Pipelines, ingestion, and data quality are production systems. Dashboards are consumers of those systems. In a tight budget environment, the work that keeps data flowing reliably survives longer than the work that rearranges it for presentation. You can see this indirectly by looking at demand signals. Data engineering roles remain consistently advertised even as overall job postings remain far below their 2022 peak.
That macro context is the last piece the original article largely ignores. By mid 2025, active job postings in the US tech market were roughly 45 percent below early 2022 levels. This matters because it explains why "skill positioning" suddenly feels decisive. The bar did not move because engineers became worse. The bar moved because slack disappeared.
Seen through this lens, the real mistake of the original article is not factual, but structural. It frames survival as a referendum on specific roles, when it is actually a referendum on operational usefulness under constraint.
Frontend is not dead. Shallow frontend is exposed. Data analysis is not dead. Analysis without ownership of the underlying system is fragile. AI did not kill jobs. Roles whose entire identity depended on a tool improving were never stable to begin with.
The uncomfortable but accurate conclusion is this: in 2025, employability in tech correlates less with how many tools you know, and more with whether your work measurably reduces cost, risk, or chaos. If it does not, your role is easier to cut, regardless of how smart or hardworking you are.
That is not cynicism. It is a system behaving exactly as designed.
Once you see it this way, the fog lifts. The layoffs stop looking like a mystery or a purge, and start looking like a constrained optimization problem. And that clarity is far more useful than another viral list of "safe roles".