The First Batch of Major Companies Cut by AI: High Salaries, High Performance, High P

Source: 36Kr

Interview | Ren Cairu, Lan Jie, Peng Qian

Text | Ren Cairu

Editor | Qiao Qian, Yang Xuan

"630" Layoffs: Is AI the Culprit or the Scapegoat?

"There is a layoff list in the company now, and you are on it." One day in mid-May, Lin Yue was called into the meeting room by his team leader, who got straight to the point.

Lin Yue's first reaction was calm; he had anticipated this. As early as March and April of this year, rumors of layoffs had circulated within some internet companies. Since the beginning of the year, major Chinese internet companies have aggressively engaged in token competitions, training sessions, and hidden assessments centered around AI efficiency. When everyone was caught up in an "all in AI" movement, it was an unspoken consensus that "layoffs are bound to happen."

However, standing at the HR door, he still faced a moment of emotional breakdown: his hands began to shake, and he hesitated for a long time, thinking about how to start and how to adjust his demeanor. "I never want to go through this again."

Lin Yue earned a monthly salary of 25,000 yuan. He graduated with a bachelor's degree a year ago and joined Ctrip as a backend engineer—at that time, he felt extremely lucky. The recruitment boom in the internet sector was over; Ctrip received thousands of resumes but hired fewer than 500 people. However, he entered the company's most profitable hotel department, responsible for writing code for commercial products.

But now, looking back, who else would they cut if not a junior programmer with a monthly salary of 25,000 yuan and only one year of experience? Firstly, the compensation cost is low, and secondly, compared to older employees who are more familiar with the business, newcomers often have lower efficiency when using AI. "With business experience as a foundation, older employees are clearer about what they want to do with AI and what impact it may have," Lin Yue said.

Stanford University used the term "canaries" to describe young people just entering the workforce in a paper titled "Canaries in the Coal Mine?" The research showed that since the popularization of ChatGPT in 2022, employment for the youngest workers has significantly declined. By September 2025, employment for software developers aged 22-25 is expected to drop nearly 20% compared to its peak at the end of 2022.

In the past year, AI has intensified everything. Ctrip was once known as a "retirement factory" for internet companies: programmers started work at 10:30 AM, had a two-hour lunch break, and could leave work on time at 7 PM, with the main app iterating every two weeks. But shortly after Lin Yue joined, the explosion of AI coding capabilities led to a pace of one app iteration per week, with "working until 10:30 PM every day."

However, this accelerated pace was not due to explosive business growth, "but because if you don't find something to do, you will become a marginal department, and marginal departments will be cut," Lin Yue told 36Kr. Ultimately, he could not escape the fate of being "cut."

However, the "slashing" may also be indiscriminate.

Cang Shu never expected to be among the first batch of people on the layoff list.

On a Friday in May, half an hour before work, "the department suddenly called an all-hands meeting, and HR directly announced the results, informing everyone about it."

Before joining Meituan, Cang Shu was a high-salary campus recruit at ByteDance, and by the end, he was the highest-paid employee among his peers. After switching to Meituan, he was entrusted with almost all core projects in the group, and this year was supposed to be his promotion milestone.

In this wave of layoffs, the protective barriers of "high performance" and "high P" have failed. In the group next to Cang Shu, two employees who were laid off had received "exceeding expectations" performance ratings last year. By the end, almost everyone in Cang Shu's group was "cut off," with "the group nominally still existing, but in reality, no one was left."

When Lin Yue learned he was laid off, he realized that the two frontend engineers he often collaborated with had "their avatars turned gray without him noticing"; a large user growth group at Meituan, which originally had hundreds of members, now had only about half left; Alibaba's Gaode and Fliggy businesses were also in severe turmoil.

"630" has become a buzzword on social media. It marks the first quarter-end when AI truly entered the internet workforce on a large scale in China. From the end of June to mid-July, it is both a customary time for personnel changes in many companies and the commonly set "last day" in this wave of layoffs.

The trendsetter Silicon Valley has already begun layoffs, characterized by mass and large-scale cuts. In May, Meta announced it would lay off 8,000 employees, with 7,000 being transferred to AI departments, making it the most turbulent tech company in Silicon Valley, with executives admitting that "company morale is at its lowest in 20 years"; earlier, Amazon announced it would cut 16,000 white-collar jobs and invest the saved funds into AI.

Before the last round of layoffs in 2021, major domestic internet companies were rapidly expanding their boundaries, densely establishing new businesses, quickly recruiting a batch of people, and then quickly erasing them.

However, the underlying theme of this year's layoffs is not so singular. The intertwining of AI efficiency improvements, sluggish growth of large and heavy old businesses, or deep competition, and the cash pressure brought by investing in new AI businesses are all running parallel during this period. Many people notified to leave find it hard to clarify which factors weigh more heavily.

The author of "Hassabis: The Brain of Google AI" states that just as Oppenheimer created the atomic bomb but could not control its use, scientists pursuing truth are also "destroyers of all things": our work, ways of thinking, and even existence may be "destroyed." Ten years ago in Seoul, South Korea, AlphaGo brought the initial destruction to human player Lee Sedol. Ten years later, from Silicon Valley to Beijing, this destruction is spreading again.

For large companies, AI is a ticket that points to new businesses like large models or AI applications. But whether these new businesses can succeed and when they will succeed is uncertain. Faced with stagnant old businesses, large companies have to decisively improve efficiency and subsequently lay off employees in every certain and uncertain direction.

When Lin Yue confided his layoff experience to friends, he was comforted, "It's okay, we will all face this day; it's just that your day came a little earlier." But perhaps more important than self-comfort is how people should choose and act after being replaced by AI and laid off by major companies.

Anxious Executives, Intensified Middle Management, Frantic Grassroots

"Products that used to take two months to develop at ByteDance can now be done in two weeks." A former product manager at ByteDance, now an executive at an AI startup, told 36Kr that with tools like Claude Code and Codex, his team can now create a demo in three hours and complete idea validation within a week.

"A product manager is like a CEO." He said that the organizational structure can be significantly compressed, and the loss of information transmission is much less than in large companies, achieving a perfect "entropy reduction."

As startups leverage AI for rapid action, do internet giants look back at themselves and feel like slow-moving behemoths?

Statements from the highest levels of large companies often signal something.

In March of this year, Meituan CEO Wang Xing discussed his views on AI at an executive communication meeting, stating, "AI Agents impact me more than ChatGPT; AI is bound to create tremendous productivity and will certainly bring significant changes to organizations and work models."

Shortly after that communication meeting, Meituan held an online conference across the company, focusing on promoting the installation and use of "Lobster," encouraging every colleague to install "Lobster" and to write their daily work as reusable Skills as much as possible.

After the meeting, Chen Yujia, who is engaged in merchant operations in Meituan's core local business, received a notification that he needed to add a section to his weekly report detailing how he utilized AI for efficiency improvements and what Skills could be promoted for use across the group and department. "Then I felt like everyone was desperately trying to integrate AI into their work."

One day in April, an algorithm engineer at Alibaba unexpectedly received a token consumption leaderboard for the previous month, and he was publicly praised for ranking first with a consumption of 17 billion tokens. The department head stated that future annual KPIs and promotion assessments would reference this ranking. However, a month later, the new token consumption leaderboard did not arrive as scheduled, "perhaps the boss also realized that this ranking method was unreliable."

New rules followed one after another. Department leaders soon proposed that employees must upload their "time reports" every hour during working days from 11 AM to 6 PM, with plugins on the Agent automatically recording code and conversation content to generate work summaries—this means employees cannot modify their time report content. The next day, HR almost argued with the leader to dissuade him from this absurd system.

Such incidents are no longer surprising. The anxiety about AI from the top continues to be delegated downwards, with middle management intensifying efforts to subtly suggest to subordinates that this is an invisible competition of reporting, an arms race, and a competition for elimination.

Although there is no mandatory requirement for everyone to write Skills, Chen Yujia's department leader still closely monitors each subordinate's token usage, occasionally inquiring about specific situations. "He is also unclear about what AI can do specifically, but he said he does not allow anyone in our team to fall behind in this AI wave." Sometimes, during private dinners after work, everyone receives a sense of crisis subtly conveyed by the boss, "We must use AI; otherwise, when the time comes, I won't be able to help you."

An engineer from an AI coding product at Alibaba told 36Kr that some business heads in the group request their product teams to increase data tracking points to "clearly see the daily usage patterns of team members using AI."

Some middle managers at Meituan, after receiving layoff targets, even submit a more aggressive and higher proportion of layoff lists to upper management—fewer people and higher AI participation rates equate to a new form of "management performance" to some extent.

AI efficiency improvement has become something that any business or function can "play around with." However, regarding what AI can actually do and how to implement it, a long fissure remains between the grassroots and management—bosses at all levels have infinite expectations for AI, while the grassroots strive to realize them but can never reach that vision, ultimately only able to exhaustively "perform."

Jiang Ling works in customer operations at Alibaba's Taotian Group, where her job is to align consumer demand with merchant supply as much as possible. In her view, bosses always think of AI as very intelligent and simple.

Take the common anomaly scenario of "exploding orders" in e-commerce, for example; the higher-ups expect to find all "hot products" in advance through comprehensive inspections. However, the platform's daily product volume is in the tens of millions, far exceeding the capacity of existing manpower and tokens, so only small-scale tests can be conducted, selecting hundreds of thousands of products, which often results in a low hit rate due to the small sample size.

"As an employee, you can't refute the expectations of your boss, you know?" Jiang Ling said, both angry and helpless.

Many times, Jiang Ling feels like a donkey, being whipped from behind. "Exhaustion is not scary; having no direction and positive feedback is the scariest. You just keep grinding, not knowing where to go in the end."

"You can't treat AI as a wishing well." The CTO of an AI company summarized to 36Kr that AI efficiency improvement has many prerequisites, with data being the foundation, but many companies have not even done a good job of digitalization; moreover, many bottlenecks in processes lie with "people," which AI alone cannot solve.

"Every Generation Has Its Own Civil Engineering"

Positions in product, operations, and other large companies still feel uncertain anxiety, while programmers can only be the first to accept their declared fate.

Li Chuan, a frontend engineer at Baidu, was first shocked by AI capabilities when he used Claude Code earlier this year. "For the same complex requirements, using some domestic large models might require five to six rounds of dialogue, while Claude can do it in two or three rounds, and do it better."

He was amazed by AI for the second time in April when the Chinese large model company Zhipu released the GLM-5.1 model, "Firstly, it is cheap, and secondly, its capabilities can completely serve as a substitute for Claude Code."

Li Chuan realized that his job was at risk. By May, he indeed appeared on the "list".

Like two sides of a coin, on one side is May 2026, when Claude Code's parent company, Anthropic, has achieved an annual revenue (ARR) of around $47 billion, increasing four to five times in just six months; meanwhile, Zhipu has recently surged to a market value of over a trillion.

On the other side, the rapid maturation of AI coding capabilities has made programmers the hardest hit in this wave of layoffs. "The first to be affected are almost all production and research teams, especially positions like front-end development and testing development, which are often easily perceived by bosses as having diminished value," an HR representative from an internet company told 36Kr.

In 2025, Li Chuan entered Baidu as a campus recruit and became a front-end engineer. A year earlier, during campus recruitment interviews, AI merely played the role of a search engine, assisting programming through simple Q&A, and interviewers did not mention AI at all.

"Front-end" was Li Chuan's ideal profession because it is a job where results are visible; the quality of code is directly reflected in every detail of the product interface. Every New Year, telling his family, "Open the Baidu app, that thing on it is made by me," gives him a sense of achievement and "the meaning of work."

For many years, programmers in large companies have been distinctly categorized into roles such as algorithms, front-end, back-end, and testing, with front-end requiring higher soft skills in aesthetics and interaction, while back-end demands more rigorous technical abilities. The salary levels and "hierarchy of disdain" in this field are directly linked to "technical content"—front-end salaries are higher than testing but lower than algorithm engineers and back-end engineers.

In just one year, everything Li Chuan was familiar with has turned upside down. The work of writing and modifying code has been largely taken over by AI, and the boundaries of various programming roles have blurred. Even product managers can now step into programming.

In May, a development department at Alibaba received a notification from their boss, instructing everyone to pause all non-urgent demands and for each team to develop an Agent. From now on, any business requirements can only be directly interfaced with the Agent by product colleagues. Programmers can only modify the Agent and cannot touch the code. The boss hinted that by October this year, teams performing well will replace those underperforming teams to maintain the Agent.

Tencent's CSIG technical team has developed a pipeline for fixing bugs in the company's app—AI fixes the bugs, and programmers only need to check and click the "confirm" button after the bug is resolved, with an accuracy rate of 50%.

In May, Alibaba established several full-stack teams internally, converting front-end, back-end, and testing engineers into "full-stack engineers," becoming "super individuals." Starting in June, Meituan also began to fully implement the merger of front-end and back-end development.

The transition to "full-stack" is theoretically feasible, but in practice, it is a painful process of peeling off a layer of skin.

Suddenly turned into a full-stack engineer, Han Zhi had little time to learn and soon had to start her first "full-stack" project, handling front-end development, testing, and everything by herself. "Now all my demands are 'reverse scheduled,' with deadlines set for specific dates," she said, feeling overwhelmed with work, often still working at 9 PM, "I am really too tired."

But the trend cannot be resisted. From late last year to early this year, several leading companies in China have been trying to spend money to encourage programmers to consume tokens, gradually phasing out "traditional programming."

At its peak, members of Tencent's CSIG team enjoyed a token quota of $2,000/month. As long as the requests were reasonable and there was corresponding code output, they could apply for a doubling of the quota once used up. Token usage was also included in performance evaluations, "When your usage is very low, your leader will ask you why." Therefore, some people would lend their unused token quotas to others.

For many years, programmers in large companies meant high salaries and prestige. They are the cornerstone of internet companies, and the essence of "programmer spirit" is open source and sharing, the simplicity and elegance of code, a focus solely on results without noise, and the excitement of seeing characters jump on the screen.

But times have changed. Almost every interviewed programmer mentioned the same feeling to 36Kr: "Without AI, I cannot work. If AI 'crashes,' I would rather spend a lot of time looking for a new Coding plan than go back to looking at code and making changes"—discussing the so-called "programmer spirit" now seems out of place.

Li Chuan said that the cultivation of an excellent programmer in the past was about learning and iteration, as programming languages have been changing over the past few decades; if you don't learn, you can't keep up with the cutting edge of technology. It was common for him and his friends to go to cafes on weekends to study new technologies, "This group itself is quite competitive." But the terrifying iteration speed of AI has left people speechless.

"If AI Coding could be locked at the level of 25 years ago, it would level the technical skills of someone with one or two years of experience with those having seven or eight years, while still not truly replacing humans, as there are many things outside the 'dialog box' to do," Lin Yue lamented. But technology does not stop for anyone; he has no doubt that the extinction of programmers is already in progress, "just like textile workers after the invention of the sewing machine."

Old Growth is Gone, New Race Begins

When technology injects a company with a multiplicative efficiency lever, what happens next is usually one of two things—either the same people do more work, or a company no longer needs so many people.

"We are not laying off employees," a CEO of a software company told 36Kr. After all the effort to "train" these programmers who have a rich understanding of the industry and development methods, each one is a treasure for the company. When AI Coding increases programming efficiency by five times, his goal is not to lay off 4/5 of the staff but to expand the business fivefold.

This wish is certainly beautiful, but the problem is, is there enough incremental market left?

Before being laid off, Lin Yue briefly experienced the "liberation" of AI writing code, but soon he became busier. In the past, when the business had iterative demands on app details, they always had to wait for the schedule to slowly come around. Now, business demands pile up faster and faster, regardless of feasibility or importance, leading the development team to "try to make it first."

However, these demands seem somewhat "unnecessary" to Lin Yue—modifying the smallest "banner position" details or changing the floating window advertisement from "free cancellation" to "points deduction." "Product managers change this and that; we will do AB testing, but the cases where the modified effect can improve are really few."

"The departments with the least growth are all in on AI, always looking for new stories to tell," Cang Shu said. He has experience in both food delivery and drone businesses, and from his personal experience, the atmosphere of AI competition is much stronger in the former.

An Infra engineer who recently experienced massive layoffs at Meta told 36Kr that after learning to leverage AI, he and his colleagues now want to do things they previously had no time for. But now, with a large number of people leaving, the remaining colleagues are starting to cut unnecessary work.

The reality facing everyone is that the star products that emerged during the mobile internet era are now struggling to substantively boost growth by "doing more work." Some of these companies not only have no growth but are also suffering severe losses due to fierce external competition.

In 2025, several companies in the food delivery war burned 200 billion, dragging Meituan's profits and cash flow into a quagmire, leading Meituan, which already had low per capita profit contributions, to enter a layoff cycle first. However, from another perspective, Meituan's business heavily relies on offline fulfillment, and the potential for AI efficiency improvement is smaller compared to companies with higher online integration. "If even Meituan can reduce staff through AI efficiency improvements, other companies will definitely follow suit. It is a bellwether," said a Meituan employee.

Traditional cash cow businesses like advertising are continuously shrinking at Baidu, and similar situations are seen with Feizhu and Gaode, which have long been marginalized and contributed little within Alibaba.

Layoffs in old businesses are unavoidable, but are there opportunities for new growth?

Some management members, when discussing layoffs, tell employees, "The company is also working on AI; you can try to find projects you can work on." A Meituan employee told 36Kr. Recently, Meituan's core local business established an AI Transformation department, mainly tasked with exploring the use of AI to streamline internal business processes; in addition, many core mid-to-high-level managers are personally leading AI-related projects.

Wang Yue, a product manager at ByteDance, told 36Kr that he is starting an internal venture to create an AI efficiency product for B-end clients, "The company encourages everyone to explore such initiatives." At the project's inception, they actively removed the "design" and "testing" roles and emphasized to the review committee how much labor cost this product would save in the future. Another colleague of Wang Yue is developing an AI customer service Agent product, with the 2026 OKR being "to help the company lay off xx% of customer service staff."

Currently, such projects exist in every major company, with dozens of small teams working on them. "Sometimes several teams work on the same direction; whoever comes out first, the company will concentrate resources to promote them."—a new race has begun.

What is changing is not only the business focus but also the organizational structure, such as eliminating more middle management.

Starting this year, Tencent has begun implementing project-based systems, weakening management levels and restoring professional ranks for leaders; Meituan laid off some L9 (division director level) during its mid-year review and recently completely eliminated the X1 node (the lowest management level), reducing management layers.

Let Us Bid Farewell to the Past

Where the giant wave of AI will take people remains unclear, and most people have yet to have a "moment of enlightenment."

Before the end of the buffer period for resignations in mid-June, Lin Yue was intensively pursuing interviews with Taobao, Kuaishou, and ByteDance. Continuing his career as a "big factory programmer" is still the optimal route he hopes for. However, the olive branches from these companies have yet to come, "It's too difficult," Lin Yue said.

"Finding a job is easy, but once you leave a big company for a medium or small one, you can never return to a big company." In Lin Yue's mind, giving up a big company somewhat means a permanent fall, and he is unwilling to "settle for less."

Some have let go of their "big company obsession." Three days after leaving Baidu, Li Chuan seamlessly joined a startup. Naturally, his position changed from "front-end engineer" to "full-stack engineer." The main product of this company is an AI Agent for office use, and they even raised his salary.

Although everyone says the times have changed and programmers' skills are no longer reliable, Li Chuan still has some "technical aspirations," hoping to participate in a user-favorite product as a technical person, which may not necessarily have to be achieved in a big company.

After leaving Alibaba, Jiang Ling joined an old car company. Her current work does not have to be forcibly associated with AI, and she no longer has to worry daily about whether "the boss's AI tasks can be completed," nor does she have to "perform desperately." A project Jiang Ling is currently responsible for will go live on September 30, "These tasks fall within my comfort zone, and with ample time, I feel much more relaxed and happy."

Recently, her department has frequently released recruitment positions, "there are always a bunch of people from Alibaba coming to interview, frantically rushing into the manufacturing industry."

Perhaps the programmer community will ultimately leave behind 10%, but Cang Shu no longer wants to seek work in big companies, "to compete for this desperate 10%."

After being laid off by Meituan in May, he decisively embarked on the entrepreneurial path. Before the AI wave, he had already tried to do something on the side. At that time, just building a community and selling some skills allowed him to experience the taste of earning 100,000 a month.

In March or April of this year, some "students" in Cang Shu's community have already jumped into AI entrepreneurship, "starting their own companies and hiring many people, while I am still struggling at this job, is this right?" he questioned himself.

Today, Cang Shu's entrepreneurial project is focused on overseas markets, developing systems based on the needs of users with rare diseases and creating independent products. He also shares progress with netizens on his Xiaohongshu account "Cang Shu (Quit the Monthly Salary Version)" and on overseas social media. In addition to his main product, he is also working on several smaller products in parallel to maintain his skills. "A small tool can be completed in three to four days at most, while a complex system may take half a month." This is far quicker than the conventional scheduling rhythm of large companies.

AI may be the most powerful intellectual leverage in human history; it can amplify individual capabilities by N times, support the implementation of most startup products, and allow every good idea to be quickly recognized and priced.

Cang Shu, born in 2000, says he is destined to be an entrepreneur, but if it weren't for this round of layoffs, he might not have acted at this moment. "The company made the decision for me."

"No looking back, move forward with passion," is the last sentence in the farewell message from Meituan to every departing employee, and it is also a phrase many employees from large companies mention when they leave. In this complex transformation brought about by AI, whether leaving or staying at a large company, one can no longer continue on the past path.

After a brief "fragmentation," it is not about lying down. Whether changing careers or starting a business, those who accept change first may be able to see a different world sooner.

(Zhou Xinyu also contributed to this article; at the request of the interviewees, Lin Yue, Jiang Ling, Li Chuan, and Wang Yue are pseudonyms.)

Disclaimer: This content is provided for general branding and informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online events, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets or to use any services. Crypto assets are highly volatile and may result in loss. WEEX services and online events may not be available in all regions and are subject to applicable laws, regulations, and eligibility requirements. You are responsible for ensuring that your use of WEEX services complies with local laws and for carefully assessing the risks before participating in any crypto-related activities.