ChatGPT didn't kill the developer job market: the number of programmers has grown noticeably since 2022

Talk of the imminent disappearance of the developer profession is not yet borne out by the numbers. The market has been changing since ChatGPT launched, but instead of a collapse in employment, the statistics show a growing number of programmers and a shift toward new kinds of work.

Numbers versus panic

The latest estimates are fairly blunt. According to the US Bureau of Labor Statistics, about 2.5 million software developers were employed in the country in February 2026. That is a record high, and roughly 400 thousand, or 19%, more than in 2022, when ChatGPT had just appeared.

Looking more broadly, JetBrains estimates the global developer community at 20.8 million people, versus 17.3 million in 2022, an increase of almost 20%. SlashData offers an even higher estimate: in early 2025 the global developer population exceeded 47 million, while in the first quarter of 2022 it was about 31 million.

Even on the most cautious reading, the conclusion is the same: the profession has not been hollowed out en masse. On the contrary, the market kept growing in the era of generative AI, although the structure of demand for specialists has begun to change noticeably.

How the role is changing

Boston University professor James Bessen does not dispute that AI affects software development. He disputes the conclusion that this automatically means fewer people. His thesis is simple: AI takes over some tasks, raises output and speeds up code writing, but that does not equal the disappearance of jobs. Developers spend less and less time on routine pieces and more on oversight, architecture and supervising swarms of AI agents that can handle individual tasks autonomously.

"AI takes on development tasks and raises productivity, but this does

not turn into job losses."

Against this backdrop, employers are no longer looking only for "pure coders." In Pluralsight's analysis, the in-demand skills include fundamentals such as Python and SQL, as well as the new layer around AI development: cloud, security, networking knowledge and work with agentic tools. In other words, engineers are now expected to do more than write functions. They must also build a reliable working system around a model, in production and in a team.

• working with AWS, Azure and Google Cloud
• Python, SQL and network engineering
• AI and cloud security
• agentic AI and MCP servers
• critical thinking and business-level communication

Anxiety has not gone away, though. In an Anthropic study, professionals said outright that they think almost constantly about the risk of losing their office job to AI, and some complained that managers had started handing out more complex tickets, counting on help from models. So the workload does not disappear; it is redistributed. AI removes part of the mechanics, but at the same time raises the bar for quality, speed and the ability to understand generated code.

Why demand isn't falling

Bessen's key argument concerns not only productivity but also demand. By his estimate, developer productivity grew by roughly 3.9% a year from 2003 to 2022, and by 6% a year from 2022 to 2025. Figures like these usually suggest that companies will need fewer people. But for software, it doesn't work so linearly.

If development gets cheaper and faster, companies don't simply cut headcount. They ship more products, try more ideas, improve quality and automate new processes. Bessen puts real growth in software output at about 9.3% a year, meaning demand for software is growing faster than the productivity of a single developer. In this model the total number of specialists need not fall: to cover an expanding market, more people are still needed.

This logic explains well why technological shifts rarely boil down to the formula "the machine will replace the human one to one." Automation can remove specific operations while expanding the market itself and the number of scenarios where software is needed. For the software industry, this means a move from manual production to the role of operator, architect and editor of AI output, along with greater responsibility for the quality, verification and integration of machine-written code into real products.

What this means

The developer apocalypse still looks more like a media storyline than a statistical fact. AI is already changing the entry into the profession and the set of required skills, but in the near term what matters is not "replacing the programmer" but growing demand for those who can build, verify and direct AI-driven development.