Tech companies are cutting jobs in the name of AI. Evidence of returns is still lacking

From January to August 2026, U.S. tech companies laid off at least 94,000 people, 16.8 percent more than a year earlier. They are increasingly citing artificial intelligence as the reason, although there is still little hard evidence of returns from it.

Money in large tech companies is flowing into artificial intelligence, while jobs are disappearing from other departments. We checked whether these cuts are based on measured returns from AI or rather on belief in them. For a manager planning a team for the next year, that distinction matters. It determines whether the reduction will bring savings or merely push the problem into the future.

Layoffs are increasing, but they come in waves

According to the Crunchbase News tracker (opens in a new tab), which collects information on layoffs at U.S. tech companies, at least 94,046 people lost their jobs from January to August 2026. In the same period of 2025, the figure was 80,486. In December 2025, 5,151 layoffs were recorded, and in January already more than 20,000. May was the worst month: 31,513 layoffs, the most in a single month since March 2023. That result included, among other things, a reduction of 8,000 jobs at Meta, or 10 percent of its workforce.

Since June, the numbers have been falling. From June to August, companies laid off a total of 19,331 people, 16.2 percent fewer than a year earlier, and 2,347 in August alone. Crunchbase cautions that it is too early to conclude that the trend has permanently reversed. It also adds that the actual figures are probably higher, because many companies do not disclose the scale of their cuts.

The biggest layoffs were carried out by the giants. By the end of August, Amazon had cut 17,388 jobs, including 16,000 as part of the reduction announced in January. Meta laid off 10,400 people, Microsoft 4,800, and PayPal 4,760. According to Roger Lee, founder of Layoffs.fyi, large companies account for about 87 percent of all laid-off workers this year, similar to 2025 (85 percent). At the beginning of September, Uber was expected to join that list with 3,300 layoffs.

AI as a justification, not necessarily as a cause

What has changed above all is how companies explain the cuts. According to Layoffs.fyi, in 2026 AI appeared as a reason in 33 percent of layoff announcements in the industry. In 2024, it was 1 percent. The service attributes 92,913 layoffs to AI, or 72 percent of all layoffs in the industry worldwide that it recorded. Layoffs.fyi counts globally, while Crunchbase covers only U.S. companies, so these figures should not be compared directly.

Roger Lee himself says that in the largest layoffs explained by AI, there is little evidence that the technology actually took over the work of the people who were laid off. In his view, large companies are spending huge sums on AI and cutting costs elsewhere. At the same time, they are counting on smaller teams being more productive. Andrew Challenger of consulting firm Challenger, Gray & Christmas describes two mechanisms. Some work, such as programming, can indeed now be done by fewer people. At the same time, companies are shifting budgets: they cut jobs in some departments and recruit in those dealing with AI. That is why the same company can announce layoffs and post job openings at the same time.

There are also early signs of doubt. According to Business Insider reports, Amazon is contacting some former employees about open positions, including in cloud and AI divisions.

What studies say about returns from AI

McKinsey’s 2025 report “The State of AI” (opens in a new tab) shows that 88 percent of organizations use AI in at least one area of their operations. However, only about 39 percent see any impact of AI on operating profit (EBIT). Just about 6 percent of companies, 109 out of 1,933 surveyed, achieve at least 5 percent of that profit thanks to AI.

Even more critical is the July 2025 report by the NANDA team at MIT Media Lab. According to it, 95 percent of generative AI pilots did not produce measurable financial returns, despite company investments estimated at $30-40 billion. However, this figure should be treated with caution. Critics point out that the report considered only an impact on financial performance within six months as success. Some of the conclusions are based on 52 interviews, which the authors themselves describe as indicative.

A similar picture emerges from measurements of work with specific tools. A Microsoft Research study showed that users of Copilot, Microsoft’s AI assistant, spent about half an hour less per week on email and finished documents 12 percent faster. However, fewer than 40 percent of the people who were given access used it regularly. In a study by research organization METR, experienced programmers working with AI on code they already knew needed on average 19 percent more time, although they themselves believed they were working faster.

CFO plans and the cost of fear in teams

In a March 2026 Duke University survey conducted together with Federal Reserve (Fed) branches in Atlanta and Richmond, 44 percent of surveyed companies (a sample of about 750 chief financial officers) declared plans for AI-related cuts. Across the U.S. economy, this corresponds to about 0.4 percent of employment, or about 502,000 jobs. That is about nine times more than the roughly 55,000 layoffs attributed to AI a year earlier. Roughly half of those positions are office jobs. Study co-author John Graham emphasizes that this is not the catastrophic scenario seen in headlines. Companies with fewer than 500 employees are even more likely to plan to hire technical specialists. The results come from a working paper that has not yet undergone scientific peer review.

The most telling evidence on the costs of cuts comes from a study by the team of Prof. Mark Ma from the University of Pittsburgh, described in August 2026. The researchers analyzed millions of Glassdoor reviews (a site with employee opinions about employers), companies’ financial reports, and about 10,000 earnings call transcripts. Company performance was linked to employee attitudes toward AI, not to management optimism. After layoffs explained by AI, team engagement declined. According to the authors, it was precisely this decline, more than skills gaps, that ate into the expected gains from implementations. The stock market reacted to such announcements with near indifference.

Dominik Libicki, co-founder of AI Investments and former CEO of Cyfrowy Polsat, says in an interview with Business Insider Polska (opens in a new tab) that most companies use AI only to improve existing business models.

What this means for a manager planning a team

In our view, the current wave of layoffs is largely the result of budget shifts and belief in future returns, rather than documented replacement of people by AI. Before the line item “fewer people thanks to AI” appears in the hiring plan, it is worth checking a few things:

• whether the time savings in a given process have been measured rather than merely declared,
• what share of the team actually uses the tool rather than just having a license for it,
• who will take over supervision of AI’s work and verification of its results,
• how the reduction will affect the engagement of the people who remain.

Amazon’s case suggests that it is easier to make a reduction decision than to reverse it. It is safer to start with a measured pilot in a single process. It is worth making decisions about jobs only when the numbers show an effect.