March 2026 was anything but calm. In just one month, 13 major Chinese tech companies were reported to have cut staff or shifted business operations toward AI. Overseas markets were just as unsettled.

Since last year, Microsoft has cut 15,000 jobs; Meta plans to lay off 20% of its workforce; Amazon has cut nearly 30,000 jobs since last year. According to public data, global tech giants are expected to have publicly laid off 230,000 people as they replace human labor with AI.

But here is the interesting part: many of the people who were laid off are being invited back to work. Gartner has even predicted that by 2027, half of the companies that cut jobs because of AI will rehire employees for similar roles.

The industry has a name for this pattern of “laying off and then rehiring”: the “boomerang.”

Is this sweeping wave of AI replacement a real revolution, or a farce?

March Goes Viral

On March 10, reports said Dewu had disbanded its independent front-end development department, moving all staff into the server-side team as it shifted toward AI full-stack development.

People familiar with the matter said pure front-end roles may last another one to two years. Once the workflow is in place, the company will likely keep only a few people for maintenance and cut the rest.

The company responded through employees on social media, calling it a “normal organizational restructuring.”

Dewu was not the only company making cuts. On March 18, reports emerged that NetEase would cut 30% of its outsourced editorial staff on April 1 and clear out all of them by May 1, starting with programmers.

The person who broke the news even posted a screenshot of Ding Lei. The outsourced team for the Identity V project was reportedly already cleared out, and users in Xiaohongshu’s comment section claimed the cuts were not limited to contractors and that even full-time employees might not be safe.

NetEase later said claims that it was “using AI to clear out all outsourced staff” were false, and that recent personnel changes were “only normal business adjustments and staff replacements for some projects.” At the same time, it made clear that it was gradually phasing out some outsourced workers in basic-skill roles.

ByteDance faced similar reports.

Since March 12, circulating claims have said that “ByteDance’s Wuhan team has been completely laid off, with no one left.” The reports said ByteDance’s task this year is “AI for everyone, doing three months of work in two months,” and that front-end roles may be hard to protect.

The company quickly denied the rumors, saying on March 14 that only about 50 employees were affected by office-location changes tied to business adjustments, and that ByteDance still had more than 2,000 employees in Wuhan.

Around the same time, Alibaba was reported to have reduced its workforce by about 34% in 2025 while doubling down on artificial intelligence, with Alibaba Cloud focusing on public cloud and AI infrastructure investment.

At the same time, major platforms including Bilibili, Tencent, Baidu, Hongguo Short Drama, Kuaishou, JD.com, Meituan and Weibo were all reported to have cut staff or shifted business toward AI.

One internet user wrote: “The whole of March was extremely anxious. Every day, my feed was flooded with news about layoffs at major tech companies.”

Project Dawn

While layoff news from China’s tech giants remained hard to verify, overseas giants began confronting the impact of AI directly.

Meta plans to cut 20% of its workforce in the second and third quarters of 2026, or about 16,000 people, in what has been called the largest round of layoffs in Meta’s history.

Mark Zuckerberg said: “Projects that once required large teams can now be done by one talented person plus AI.”

He required all Meta employees to undergo mandatory AI assessments. If you cannot use AI, it means you may not be far from leaving.

Amazon has cut a total of 30,000 jobs from last October to this year, equal to 10% of its white-collar workforce and the largest reduction in the company’s 31-year history.

The company’s internal codename for the layoffs is “Project Dawn.” Its CEO was even more blunt: “AI will reshape the labor structure, and the same business no longer requires the same amount of labor.”

Since last year, Microsoft has cut 15,000 jobs, with software engineers accounting for about 40% of one round of layoffs.

After cutting 10,000 jobs last year, Oracle plans to cut another 20,000 to 30,000 jobs starting this March, equal to about 18% of its workforce.

Among them, 47 database administrators were replaced by AI, with only three senior engineers retained for oversight.

So far, foreign tech giants have publicly announced 170,000 layoffs due to AI.

Including AI-related layoffs at Chinese, European and other global tech giants, direct AI-driven layoffs across the global technology industry are expected to exceed 230,000.

Goldman Sachs estimates that about 300 million full-time jobs worldwide face automation risks, and that 210 million roles could disappear permanently.

Boomerang

This frenzied wave of AI replacing people appears unstoppable.

Many people now feel insecure, believing they could be replaced by AI at any moment.

But many of the companies that cut aggressively have begun to “lay off and then rehire,” a pattern the industry calls the “boomerang” phenomenon.

A 2025 report by workforce analytics firm Visier found that 5% of laid-off workers were later called back by their original employers. The pattern is even more pronounced among people laid off because of AI, because companies often overestimate AI’s capabilities and cut the wrong people.

Chen Dingding, a customer service worker, said that two months after she was laid off, her supervisor invited her back to work.

“When I returned to my post, I found that the customer service department originally had 100 people, and half had come back.” That number is still rising.

Gartner has even predicted that by 2027, half of the companies that cut jobs because of AI will rehire employees for similar roles. Forrester’s forecast is even higher, at 55%.

Its survey found that 55% of employers regretted AI-driven layoffs. Why does the boomerang effect happen?

In August 2025, MIT released a sobering set of data: 95% of companies pouring money into AI had not yet made money from it. Several reasons sit behind that finding.

First, AI’s capabilities are still not good enough, and many technologies remain immature. Bosses imagine AI can handle everything. In reality, it can do only one-third of the work. In March 2026, Anthropic, the maker of the Claude model, released a report arguing that in theory AI can do 94% of computer-related work, but in real business settings it can handle only 33%.

(Image source: Source: The AI Corner, Anthropic 2026 AI Jobs Report) After cutting staff, many companies discovered that AI still could not solve many problems, so they swallowed their pride and hired people back.

In 2024, Swedish fintech giant Klarna carried out large-scale layoffs, with its CEO loudly claiming that AI customer service could replace 700 full-time employees.

But he soon changed his mind. AI can indeed answer routine questions instantly, but when it encounters situations that require warmth and complex judgment, such as an elderly person who simply wants to talk to someone about a pension, AI just coldly sends a link. Customers ended up furious.

In 2025, Klarna had to quietly start hiring again, humbly saying it wanted to “become the best at providing human interaction.”

It is not just customer service. Technical workers have not escaped either.

On February 26, 2026, payments company Block cut 4,000 people in one move, equal to 40% of its workforce. Less than a month later, some employees received return-to-work notices. Some were told they had been cut because of a “clerical error.”

Some stayed only because managers fought hard for them, while some technical leads protested directly, saying they could not maintain systems after the layoffs and would leave too unless people were brought back.

Second, maintaining AI carries a large amount of hidden cost.

AI is not something you can just plug in and use. It needs data, prompt tuning, error handling, human fallback, review and cleanup.

Zhou Ran, a programmer at one company, knows this well: “Our group was cut from eight people to three. Objectively speaking, the code AI writes can run, and it is efficient. But once there is a bug, troubleshooting it is even more painful than debugging human-written code.

”

Many users found after installing Big Lobster that debugging Big Lobster took longer than doing the work themselves.

In fact, AI operations and maintenance can sometimes cost more than hiring people.

Third, human society is complex. Amy, an HR executive at a major tech company, believes AI still lacks the ability to deal with human society.

She gave an extreme example: AI has no understanding of social norms at all. It is like a young woman fresh out of school who is highly capable at work but does not understand a major company’s “rules,” and gets pushed out all the same.

“Sometimes, the most efficient solution is not necessarily the best solution for the group,” Amy said. She believes the industry has seriously overestimated AI’s real capabilities and underestimated the complexity of real business, and that many companies are now cutting too aggressively.

Their internal strategy is to use AI over the next one to two years to replace 10% to 15% of basic roles, while also creating 5% to 10% new roles.

She believes this trend is not about AI replacing people, but about “people who know how to use AI” replacing “people who do not know how to use AI.”

This AI wave is not bringing a “wave of unemployment,” but a “restructuring of employment.”

The first step for everyone facing the AI wave is not to “fight AI,” but to “coexist with AI.”