AI Deflation in IT: Klarna and IBM Cases Show Why Vacancies Are Up But Salary Growth Is Weak

AI does not directly kill demand for developers, but quickly reduces the price of typical engineering work: more vacancies appear, but salary growth weakens, and for some seniors the market has already turned negative. The author calls this AI-deflation — a situation where businesses get more results with AI tools but don't share this gain with employees through higher salaries. According to U.

S. data, IT compensation growth in 2025 was only 1.6% — the worst figure in roughly 15 years.

Meanwhile, senior developers in certain segments lost about 10% year-over-year, and mid-level SQL developers lost about 7%. At the other end of the spectrum are AI specialists: they see 9–12% growth, and average compensation for LLM developers reaches 209 thousand dollars a year. It turns out the market is splitting into two layers: those whose skills AI makes more expensive, and those whose work AI makes cheaper.

The most interesting paradox is that formal demand hasn't fallen. According to data from Citadel Securities and Indeed, the number of vacancies for software engineers grew by roughly 11% year-over-year. A couple of years ago, this would have almost automatically meant another round of salary escalation.

Now the logic is different. If development becomes cheaper and faster thanks to code generation, auto-testing, documentation, and legacy system analysis, a company doesn't necessarily hire fewer people. It can hire the same number or even more, but on more modest terms.

In other words, the market gets more positions, but the average value of one typical position declines. This is exactly what's described as deflation — not a collapse in employment, but a reduction in the price of standard work. This model is already visible among major players.

Block conducted large-scale layoffs, and the market took it positively: the company's stock rose. IBM reported billion-dollar savings in operating expenses thanks to AI and automation, including significant cuts in HR and IT functions. BCG records an even more telling signal: 93% of executives view AI primarily as a tool for cutting costs, not as a source of new salary raises.

The business logic here is transparent. If AI makes a specific activity 15% more efficient, a company either reduces its team or demands it process 15% more tasks. The option of "paying 15% more" barely factors into such models.

The consequences differ for different developer levels. It's hardest for juniors: AI eats away at exactly those entry-level tasks where experience was once gained — template code, simple integrations, basic documentation, fundamental debugging. Therefore, market entry becomes narrower and more demanding.

Mid-level developers are under pressure if their value comes down to quickly writing typical CRUD and maintaining predictable services. But strong mid-level developers have a real chance to win more than others: with a good set of AI tools and product understanding, they can handle the volume of work that previously required a wider team. Seniors are holding up better for now, but even for them status is no longer protective in itself.

The market increasingly pays not for years in the profession, but for the ability to make architectural decisions, reduce risks, lead teams, and turn AI into measurable business results. The Russian context is particularly important. The picture here is still softer than in the U.

S., but signs of cooling are already visible. According to data cited by the author, the median salary for an IT specialist in the second half of 2025 was 183 thousand rubles and barely changed year-over-year.

In Moscow growth was around 4%, while inflation according to Rosstat reached 5.6%, meaning real incomes actually declined. At the same time, the market across Russia looks uneven: in some places salaries are still growing, while in others stagnation is already visible.

Therefore, it's premature to speak of proven AI-deflation in Russian IT, but the trend of weakening salary dynamics is already apparent. The main conclusion is simple: AI doesn't cancel the developer profession, but changes the value structure within it. What becomes cheaper is not "development in general," but typical and easily scalable work.

What becomes more expensive are roles requiring domain expertise, architecture, responsibility for results, and the ability to embed AI into a product or process, rather than simply using editor prompts. If this logic takes hold, the market could show growth in vacancy numbers for a long time simultaneously with weak salaries. For specialists this means one thing: winners will not be those who simply write code faster, but those who, with the help of AI, deliver significantly more to the business than their salary would have cost under the old rules.