Author: Claude, Deep Tide TechFlow

Deep Tide Introduction: A survey by the National Bureau of Economic Research (NBER) of 600 executives from four countries shows that nearly 90% of companies believe AI has "no impact" on employment and productivity in the past three years, but by Q1 2026, the technology industry has laid off 78,557 people, of which 47.9% are attributed to AI. While productivity data remains blank, a wave of layoffs surges in the name of AI, leading economists to liken this contradiction to the AI version of the "computer paradox" proposed by 1987 Nobel Prize winner Solow.

With $250 billion invested, nearly 90% of enterprises say AI has not brought any productivity improvements. Meanwhile, tech companies are laying off employees on a large scale in the name of AI.

This is the most absurd scene in the current AI industry.

According to a report by Fortune magazine on April 19, a study released by NBER in February this year, covering 6000 corporate executives in the US, UK, Germany, and Australia, found that nearly 90% of the surveyed companies reported that AI had no measurable impact on their employment and productivity over the past three years. While two-thirds of the executives are using AI, the average weekly usage time is only 1.5 hours, with 25% of respondents stating they do not use AI in their work at all.

On the other hand, according to data from RationalFX quoted by Nikkei Asia, from January 1 to early April 2026, the tech industry has laid off 78,557 people, of which 37,638 (47.9%) are explicitly attributed to AI and workflow automation. Over 76% of the layoffs occurred in the United States.

Apollo's chief economist Torsten Slok directly quoted the classic statement of 1987 Nobel Prize winner Robert Solow to summarize the current situation as the AI version of "Solow's Paradox". Solow's original words were: "The computer age is everywhere except in the productivity statistics."

Slok's judgment reflects today's situation almost exactly. There is no trace of AI in employment, productivity, and inflation data.

90% of companies do not see AI effects, $250 billion investment return in doubt

The data from the NBER study is quite solid. Among the four countries, 69% of enterprises use AI to some extent, with the US being the highest (78%) and Germany the lowest (65%). But usage is one thing, effectiveness is another: over 90% of managers indicate that AI has no impact on their company's employment size, and 89% state it has no impact on labor productivity (measured by per capita sales).

According to the Stanford University 2025 AI Index report, global AI investments exceeded $250 billion in 2024. PwC's 2026 global CEO survey shows that only 12% of CEOs reported that AI brought both cost reductions and revenue growth, while 56% stated they have not seen any significant financial benefits.

Slok pointed out in his blog post that, aside from the "big seven," AI has no visible impact on profit margins and earnings expectations.

This is not an isolated opinion. A study by MIT in 2024 predicted that AI can only enhance productivity by 0.5% in the next decade. The study's author, Nobel Prize winner Daron Acemoglu, candidly stated at the time: "0.5% is better than zero. But relative to the promises made by the industry and tech media, it is indeed disappointing."

A study released by the Boston Consulting Group (BCG) in March this year revealed an counterintuitive phenomenon: when employees use fewer than three AI tools, productivity improves; however, after using four or more tools, self-reported productivity significantly declines, with employees reporting "brain fog" and more minor errors. BCG termed this "AI brain overload."

The 2026 Global Talent Outlook by ManpowerGroup shows that among nearly 14,000 employees from 19 countries, the regular usage rate of AI rose by 13% in 2025, but confidence in AI's practicality plummeted by 18%.

Almost 80,000 layoffs in Q1, is AI the biggest "scapegoat" or the real culprit?

As productivity data remains absent, the wave of layoffs is advancing at an astonishing speed.

According to Nikkei Asia, the tech industry laid off 78,557 people in Q1 2026, with 47.9% attributed to the implementation of AI and workflow automation. Oracle recently quietly laid off over 10,000 employees, and the savings were invested in data center construction. Both Anthropic CEO Dario Amodei and Ford CEO Jim Farley publicly stated that AI will eliminate half of the entry-level white-collar jobs in the United States within the next five years. Research from Stanford University also shows that junior programming and customer service jobs have been impacted, with related hiring positions dropping by 13% in three years.

A simulation study by MIT provided alarming figures: AI could replace 11.7% of the US workforce, involving about $1.2 trillion in total wages.

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

Cognizant's Chief AI Officer Babak Hodjat bluntly told Nikkei Asia: "I’m not sure whether these layoffs are directly related to actual productivity improvements. Sometimes, AI is just a scapegoat in financial terms—companies overhired and want to downsize, and then blame it on AI."

OpenAI CEO Sam Altman also acknowledged the existence of "AI washing" at the AI Impact Summit in India, stating, "There’s a certain proportion of 'AI washing' where people blame layoffs on AI that were already planned, but indeed some jobs are genuinely being replaced by AI."

Analysts at Deutsche Bank directly labeled this phenomenon as "AI redundancy washing," believing that companies attribute layoffs to AI because "it sends a more positive signal to investors than admitting weak demand or prior overhiring."

IBM increases entry-level hiring against the trend, Cognizant refuses layoffs

Not all companies are going with the flow.

IBM has doubled its entry-level hiring in 2026, with the company's chief human resources officer Nickle LaMoreaux reasoning that while AI can accomplish many entry-level tasks, cutting these positions would destroy the talent pipeline that nurtures future middle management, jeopardizing the company's long-term leadership reserves.

Cognizant—a process outsourcing giant that heavily relies on human labor—also stated it would not lay off employees due to AI. The company has set up AI labs in San Francisco and Bangalore to develop customized AI agents for clients (because off-the-shelf generic AI products have not performed well in enterprise environments and have performance and security issues), but its employees will be trained to work alongside AI, rather than being replaced by it.

Hodjat emphasized, "Many recent graduates will find it difficult to get jobs and lack domain expertise. You must hire them and let them learn how to use AI across various fields on the job."

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

J curve or mirage: When will the AI productivity inflection point arrive?

Historical experience offers some hope.

IT investments from the 1970s to 1980s also seemed ineffective, but between 1995 and 2005, IT-driven productivity growth reached 1.5%. Erik Brynjolfsson, director of the Stanford Digital Economy Lab, pointed out in the Financial Times that the inflection point for AI productivity may already be showing: the US saw a 2.7% productivity growth last year, with Q4 GDP tracking growth at 3.7%, but the new job creation during the same period was only 181,000—this decoupling of employment growth from GDP growth could signal the beginning of AI's impact. Former Pimco CEO Mohamed El-Erian also noted the same decoupling phenomenon.

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

Apollo's Slok described the future impact of AI as a "J curve": first experiencing a period of decline in performance, then ushering in an exponential leap. But he also pointed out that unlike the IT era in the 1980s, where innovators had monopoly pricing power, today’s AI tools are continuously dropping in price due to fierce competition. Therefore, the value creation of AI does not lie in the product itself, but in how "generative AI is used and deployed across economic sectors."

Hodjat's judgment may be the most pragmatic: in another 6 to 12 months, companies will start to see real productivity improvements brought by AI, and "this transitional period will be painful for all of us."