跪拜 Guibai
← All articles
Developers · Artificial Intelligence · Interviews

Big Tech's 2025 Layoffs Are Not About Survival — They're a Strategic Bet on AI

By 狂师 ·
Read original on juejin.cn ↗ Google Translate ↗ Alt translation

The same companies laying off tens of thousands are spending hundreds of billions on AI. That disconnect tells every working developer something urgent: the job you hold today is being re-priced in real time against what a model can do, and the safe middle of the skill curve is disappearing fastest.

Summary

Layoffs across the global tech industry surpassed 110,000 in the US alone during 2025, with Microsoft, Google, Meta, and Amazon each cutting five-figure headcounts. In China, Alibaba shed 80,000 employees in a single fiscal year, Baidu executed its largest-ever round of cuts, and ByteDance saw profits crater as it redirected earnings into AI. What distinguishes this wave is that companies are slashing jobs inside profitable business units — not just shuttering money-losing experiments — while simultaneously committing $350–400 billion in combined capex, overwhelmingly directed at AI.

AI provides a convenient public rationale, and roughly 55,000 US layoffs were explicitly attributed to automation. But the deeper drivers are an industry hitting the ceiling on user and advertising growth after a decade of aggressive headcount expansion. The sector is shifting from land-grab mode to operational efficiency, and AI is accelerating a headcount correction that was already overdue. Junior and mid-level programmers face the most immediate pressure, with global developer roles down 8% and Anthropic reporting that 75% of coding tasks are now covered by AI.

The psychological ripple effects extend beyond those laid off. Young engineers are reconsidering whether big-tech careers still offer a viable long-term path, while mid-career staff face a job market that has flipped from talent scarcity to role scarcity. Yet high-end architects, AI specialists, and business-savvy engineers command 120–210% salary premiums, and some laid-off workers are using severance to launch independent products. The industry isn't dying — it's maturing from an expansion play into an operations play, and that transition rewards depth over speed.

Takeaways
US tech layoffs exceeded 110,000 in 2025, with Microsoft, Google, Meta, and Amazon each cutting between 3,600 and 15,000 roles.
Alibaba's headcount dropped from 205,000 to 124,000 in one fiscal year — a loss of 80,000 employees, partly due to asset sales but substantially from core business cuts.
Baidu executed its largest-ever layoff in late 2025, with some MEG divisions cutting 20–30% of staff, including ten-year veterans and entire teams.
ByteDance's net profit fell over 70% year-on-year as earnings were redirected into AI investment.
Tencent was a rare counter-example, growing headcount from 82,000 to 87,000.
Companies are cutting jobs inside profitable business units, not just failing ones — a deliberate restructuring, not a defensive cost-cutting measure.
Meta, Alphabet, Amazon, and Microsoft are projected to spend $350–400 billion in combined 2025 capex, mostly on AI.
Roughly 54,000–55,000 US layoffs were explicitly labeled by employers as caused by AI and automation.
Anthropic reports that 75% of programmer tasks are now covered by AI, and global developer jobs fell 8% in 2025.
One in four junior programmers faces unemployment risk.
China's mobile internet user growth and advertising revenue growth have both plateaued, removing the expansion imperative that previously justified large headcounts.
Senior system architects, AI algorithm engineers, and business-savvy technical staff still see salary increases of 120–210%.
JD.com entered food delivery in 2025 at a Q3 operating loss of 1.05 billion yuan, prioritizing future market share over near-term profit.
Pinduoduo distributed 7.9 billion yuan in stock-based compensation in 2025, averaging 310,000 yuan per employee, to retain talent.
Some laid-off engineers are using severance packages to fund independent SaaS products, technical content creation, and solo development careers.
Conclusions

Layoffs during profitable quarters signal a structural shift, not a cyclical downturn. Companies are reallocating capital from labor to compute, and that trade-off is now explicit in earnings calls.

AI is a politically convenient scapegoat that lets executives avoid admitting they over-hired during the zero-interest-rate era. The technology is real, but it's also being used to justify headcount decisions that were already overdue.

The disappearance of junior and mid-level roles creates a pipeline problem: if entry points vanish, the senior architects and AI specialists commanding premiums today have no replacement cohort in training.

An industry that scares off young talent and makes mid-career workers risk-averse loses the very people who would build its next generation of companies. The long-term damage from this confidence collapse may exceed the short-term savings from layoffs.

ByteDance's 70% profit drop reveals a strategic gamble: accept ugly financials now to secure an AI position that may determine survival later. This is not cost-cutting — it's a bet-the-company reallocation.

The shift from 'expansion-type' to 'operation-type' industry means the skills that got people hired in 2015 — speed, generalist coding, willingness to grind — are being devalued in favor of depth, domain expertise, and business judgment.

Independent development and personal-brand building are no longer side projects for the eccentric; they are becoming rational hedges against single-employer dependency in a sector where loyalty offers no protection.

Concepts & terms
N+1 severance
A common severance formula in Chinese tech layoffs: N months of salary (where N is years of service) plus one additional month, as required or exceeded under Chinese labor law.
MEG (Baidu)
Baidu's Mobile Ecosystem Group, the business unit responsible for Baidu's core search, feed, and mobile app ecosystem, which underwent deep cuts in 2025.
Capital expenditure (capex) in AI context
Spending on physical assets like data centers, GPUs, and networking infrastructure. The $350–400 billion figure for 2025 represents an unprecedented buildout of AI training and inference capacity by the four largest US tech firms.
Anthropic's 75% task coverage claim
A finding from Anthropic's economic analysis indicating that large language models can now perform roughly three-quarters of the discrete tasks that make up a programmer's job, though task coverage does not equal full job replacement.
Expansion-type vs. operation-type industry
A framework distinguishing industries that grow by acquiring new users and territory (rewarding speed and boldness) from those that grow by extracting more value from existing assets (rewarding patience, efficiency, and depth).
Source: juejin.cn ↗ Google Translate ↗ Backup ↗