6,000 executives say AI has no impact, yet 40,000 were laid off in Q1 2026

Author: Claude, DeepChain TechFlow

Shenchao Summary: A survey by the National Bureau of Economic Research (NBER) of 6,000 executives across four countries found that nearly 90% of companies believe AI has had "no impact" on employment and productivity over the past three years. Yet, in Q1 2026, 78,557 tech workers were laid off, with 47.9% of these layoffs attributed to AI. Amidst a vacuum in productivity data, layoffs are surging under the banner of AI—economists liken this paradox to the AI-era version of the "computer paradox" proposed by 1987 Nobel laureate Robert Solow.

$250 billion invested, yet nearly 90% of companies say AI has not improved productivity. Meanwhile, tech companies are laying off large numbers of employees under the guise of AI.

This is the most absurd scene in today's AI industry.

According to Fortune magazine on April 19, a February study by the NBER, which surveyed 6,000 business executives across the United States, the United Kingdom, Germany, and Australia, found that nearly 90% of respondents reported no measurable impact of AI on employment or productivity over the past three years. Although two-thirds of executives use AI, they average only 1.5 hours of usage per week, and 25% of respondents said they do not use AI at work at all.

On the other hand, according to RationalFX data cited by Nikkei Asia, between January 1, 2026, and early April, the technology sector laid off 78,557 people, with 37,638 (47.9%) explicitly attributed to AI and workflow automation. Over 76% of these layoffs occurred in the United States.

Apollo's chief economist, Torsten Slok, directly cited the classic observation by 1987 Nobel laureate Robert Solow, summarizing the current situation as an AI version of the "Solow Paradox." Solow's original statement was: "You can see the computer age everywhere but in the productivity statistics."

Slok's assessment maps almost exactly to today. AI is nowhere to be seen in employment data, productivity data, or inflation data.

The data from this NBER study is quite robust. Among the four countries, 69% of firms use AI to some extent, with the highest rate in the United States (78%) and the lowest in Germany (65%). However, usage does not necessarily translate into impact: over 90% of managers reported that AI had no effect on their firm’s employment levels, and 89% said it had no effect on labor productivity (measured by sales per employee).

According to the 2025 AI Index Report by Stanford University, global AI investment exceeded $250 billion in 2024. The PwC 2026 Global CEO Survey found that only 12% of CEOs reported that AI has simultaneously reduced costs and increased revenue, while 56% reported no significant financial gains.

Slok noted in his blog post that, aside from the "Seven Giants," AI has no visible impact on profit margins or earnings expectations.

This is not just one opinion. A 2024 MIT study predicted that AI will increase productivity by only 0.5% over the next decade. The study’s author, Nobel laureate Daron Acemoglu, admitted at the time: “0.5% is better than zero, but it is indeed disappointing compared to the promises made by industry and tech media.”

A study by the Boston Consulting Group (BCG) published in March this year revealed a counterintuitive phenomenon: productivity increases when employees use fewer than three AI tools, but self-reported productivity drops significantly when using four or more tools, with employees reporting "brain fog" and an increase in minor errors. BCG refers to this as "AI cognitive overload."

ManpowerGroup’s 2026 Global Talent Outlook shows that among nearly 14,000 employees across 19 countries, the regular use of AI increased by 13% in 2025, but confidence in AI’s practicality dropped by 18%.

While productivity data remains absent, layoffs are advancing at an alarming pace.

According to Nikkei Asia, 78,557 tech industry layoffs occurred in Q1 2026, with 47.9% attributed to AI implementation and workflow automation. Oracle recently quietly laid off more than 10,000 employees, redirecting the saved funds toward data center construction. Anthropic CEO Dario Amodei and Ford CEO Jim Farley have both publicly stated that AI will eliminate half of all entry-level white-collar jobs in the U.S. within the next five years. Stanford University research also shows that entry-level programming and customer service roles are already being impacted, with related job postings declining by 13% over the past three years.

A simulation study by MIT provided a startling figure: AI could replace 11.7% of the U.S. workforce, involving a total salary volume of approximately $1.2 trillion.

But how many of these layoffs were truly driven by AI?

Cognizant’s Chief AI Officer, Babak Hodjat, told Nikkei Asia bluntly: “I’m not sure these layoffs are directly tied to actual productivity gains. Sometimes, AI becomes a scapegoat financially—companies have hired too many people and want to downsize, so they blame it on AI.”

OpenAI CEO Sam Altman also acknowledged the existence of the "AI washing" phenomenon at the India AI Impact Summit, stating, "There is a certain degree of 'AI washing'—people blame AI for layoffs that were already planned—but there are also genuinely some jobs being replaced by AI."

Deutsche Bank analysts have more directly labeled this phenomenon as "AI redundancy washing," arguing that companies attribute layoffs to AI because "it sends a more positive signal to investors than admitting to weak demand or previously over-hiring."

Not all companies are following the trend.

In 2026, IBM doubled its entry-level hiring, with Chief Human Resources Officer Nickle LaMoreaux reasoning that although AI can perform many entry-level tasks, eliminating these roles would destroy the talent pipeline for developing future mid-level managers, jeopardizing the company’s long-term leadership pipeline.

Cognizant, a process outsourcing giant whose business heavily relies on human labor, has also stated it will not lay off employees due to AI. The company has established AI labs in San Francisco and Bangalore to develop customized AI agents for clients, as off-the-shelf general-purpose AI products often underperform in enterprise environments due to performance and security issues. However, its employees will be trained to collaborate with AI rather than be replaced by it.

Hodjat emphasized: "There will be a large number of recent graduates who can't find jobs and lack domain-specific expertise. You must hire them and let them learn on the job how to use AI across various fields."

Data from the European Central Bank also supports this perspective from another angle: companies that extensively deploy and invest in AI are more likely to expand their hiring.

Historical experience offers some hope.

IT investments in the 1970s and 1980s appeared equally ineffective, but between 1995 and 2005, IT-driven productivity growth reached 1.5%. Erik Brynjolfsson, Director of the Digital Economy Lab at Stanford University, wrote in the Financial Times that the productivity inflection point for AI may already be emerging: U.S. productivity grew by 2.7% last year, with GDP tracking growth of 3.7% in the fourth quarter, yet only 181,000 new jobs were added during the same period—the decoupling of job growth from GDP growth may be a sign that AI is beginning to take effect. Former Pimco CEO Mohamed El-Erian has also observed this same decoupling phenomenon.

A study by the Stanford Institute for Economic Policy Research, using web browsing data from 200,000 U.S. households, found that AI improved efficiency by 76% to 176% on online tasks such as job searching, travel planning, and shopping. However, researchers found that users spent the time saved on socializing and watching TV, rather than on work or learning new skills.

Apollo’s Slok describes the future impact of AI as a “J-curve”: an initial period of performance decline followed by exponential growth. However, he notes that, unlike the IT era of the 1980s, when innovators held monopolistic pricing power, today’s AI tools face intense competition, driving prices continuously downward. Therefore, the value creation from AI lies not in the products themselves, but in “how generative AI is used and deployed across various sectors of the economy.”

Hodjat’s assessment may be the most pragmatic: it will take another 6 to 12 months before companies begin to see genuine productivity gains from AI, and “this transition period will be painful for all of us.”