Violent layoffs of 16,000!

Source: TuChong

For those still struggling with the trivial daily grind, probably the least favorite topic is “AI replacing humans.”

But whether you accept it or not, this trend is becoming increasingly obvious.

According to Reuters, Meta is planning an unprecedented round of massive layoffs, with the layoff rate possibly reaching 20% or more of the total company workforce.

What does 20% mean?

Based on Meta’s announced total of about 79,000 employees by the end of 2025, this means approximately 16,000 workers will soon receive their “ticket to the job market.”

Putting this data into the current environment, it’s truly absurd.

In the first three quarters of 2025, Meta’s total revenue was $141.073 billion, net profit $37.690 billion, and operating cash flow reached $79.586 billion.

Their earning capacity is quite strong.

The outlook isn’t bad, so why such drastic measures?

01. Burning Money Frenziedly

The reason is obvious: although profitable, they are spending even more.

Mainly due to huge setbacks in AI research and development.

Last summer, they planned to launch the largest version of Llama 4, codenamed “Behemoth,” but due to misleading benchmark results and unmet expectations, this project was shelved.

Subsequently, the “Super Intelligence Team” pinned hopes on two new models codenamed “Avocado” and “Mango.”

However, these models failed again in internal testing, and their release has been pushed back to May 2026.

Research setbacks mean hundreds of billions of dollars in invested compute power might be wasted.

To salvage the situation, Meta has had to spend wildly.

With heavy investments, there are brave souls.

According to Business Insider, Meta has offered top researchers in the “Super Intelligence Team” super salaries worth hundreds of millions of dollars (vesting over four years); even ordinary AI engineers’ salaries and equity incentives are about twice as high as those in traditional business lines.

Meanwhile, tech media MLQ.ai reports that Meta is aggressively acquiring companies, such as the social platform Moltbook, and spending $2 billion to acquire Chinese AI startup Manus, even recruiting AI big shot Alexandr Wang into the Super Intelligence Lab.

In February this year, Meta also signed a five-year strategic partnership with AMD, purchasing up to 6 GW of GPU compute capacity, totaling over $60 billion; and expanded cooperation with Nvidia, procuring millions of Blackwell architecture chips.

…

On one side, huge money wasted on R&D failures; on the other, aggressive acquisitions and sky-high salaries for top talent.

In 2026 alone, Meta’s capital expenditure on AI will reach $135 billion.

And to stay ahead in the AI arms race, Meta has set an outrageous goal: to invest $600 billion before 2028 to build 30 super-large data centers.

What if the company’s resources are exhausted?

It seems the only option left is layoffs.

According to cybersecurity media Cybernews, Meta is expanding the scope of “below expectations” performance ratings.

This means the company is using strict performance evaluations to find “legitimate” reasons for the upcoming large-scale optimization.

But even if 16,000 people are cut, it won’t fill such a huge financial gap:

Goldman Sachs estimates that if Meta’s layoffs reach 20%, it could save about $8-10 billion annually in labor costs (including salaries and stock options).

This also gives Wall Street analysts something to explain.

The question is: how to cut? Who to cut?

02. Who Are the “Excess”?

First, look at the departments.

First is Reality Labs, the metaverse division.

In January, Meta already laid off 1,500 people in this department, and more cuts are likely.

The metaverse vision is becoming more abstract under the impact of AI’s dimensionality reduction, and long-term projects are clearly giving way to short-term AI compute needs.

Next are non-AI priority teams, such as Facebook and other social projects; as well as routine administrative and operational roles, as Meta is increasing the use of AI-assisted operations.

These departments can’t be fully cut, so who exactly will be cut?

Actually, as early as January’s earnings call, Zuckerberg warned: “I see some projects that used to require large teams now only need a very talented individual to handle.”

Less than two months later, just last week, Meta established a new “AI Engineering Organization.”

The biggest feature of this department: a manager-to-employee ratio of 1:50.

Traditionally, a frontline manager leads 5-8 people, directors oversee several managers, VPs manage multiple directors, creating a layered “matryoshka” management structure, resulting in a large middle management layer.

Now, with AI code assistants, AI project management agents, automated testing and deployment tools, one manager can directly oversee 50 employees, eliminating the need for intermediaries.

These massive middle management roles are the targets for “optimization.”

Let’s not even consider the ethical issues of replacing humans with AI.

From Meta’s perspective, this is not only cost reduction and efficiency improvement but also a strong signal to the market: management has high efficiency in controlling costs.

In today’s context of AI infrastructure costs soaring into hundreds of billions of dollars, investors fear that companies are buying GPUs while still maintaining thousands of “idle” staff.

Zuckerberg’s approach, though ruthless and lacking compassion, greatly reassures investors.

More importantly, it’s a deep genetic optimization.

By shedding 20% of the workforce and channeling more funds into AI R&D, Meta’s business model will shift from a “labor-intensive brain workshop” to a “capital-intensive AI compute factory.”

If the “Avocado” and “Mango” models can be successfully released in May this year and reverse the decline of Llama 4, Meta has the capacity to deeply embed AI agents into the daily lives of billions of users through social platforms.

The risk is that if the $600 billion capital expenditure cannot be monetized through AI (such as enterprise AI services, AI advertising, smart hardware), Meta will face huge asset depreciation pressures, eroding its net profit.

This is a gamble where failure means total loss.

03. New Rules

Whether Meta succeeds or not, its corporate culture will be fundamentally reshaped.

Top AI experts with contracts worth hundreds of millions are treated like ancestors, while ordinary programmers, product managers, and operations staff become “replaceable cogs.”

This is a “whistle” for the entire tech industry, Silicon Valley, and even future global enterprises.

As Business Insider comments: this marks a fundamental shift in tech strategy: as AI becomes more powerful, big tech companies are betting—they can build with fewer people, faster, cheaper.

In recent months, although not as extravagantly as Meta, many similar cases have emerged:

Amazon confirmed in January it cut about 16,000 jobs, nearly 10% of its workforce, citing AI development as the reason;

Atlassian announced in February layoffs of about 1,600 employees, 10% of staff, citing “AI and efficiency pursuits”;

Verizon cut 13,000 middle management roles, reducing quarterly operating costs by $920 million;

Fintech company Block eliminated over 4,000 jobs, citing “new AI tools enabling smaller teams and higher efficiency”;

Even Wall Street investment banks are not immune—Goldman Sachs laid off 2,500, as the trend of replacing traditional financial analysis and copywriting with AI becomes more evident…

In the past, large-scale layoffs were seen as signs of poor management.

But now, “layoffs for AI development” have become a perfect cover.

More critically, cutting “redundant” staff doesn’t harm business; instead, it speeds up growth by reducing internal communication costs and increasing AI tool integration.

In the past, analysts looked at metrics like “per-store sales efficiency” or “per-person output.”

But in the future, these terms may be outdated when analyzing company financials.

What will we look at?

• RPGPU (Revenue Per GPU): revenue generated per GPU.

• AI-to-Human Ratio: the ratio of AI agents to human employees.

• Cost of Compute vs. Cost of Labor: how companies optimize between these two.

…

This aligns with the new mantra: revenue per employee.

No matter how much you earn, if your employee count keeps rising, you’re a bad company. If profits grow while employee numbers decline, that’s a good company.

Thus, layoffs are no longer just crisis management but a normalized management tool.

Eventually, a tacit “herd effect” forms.

From a company perspective, if everyone is doing it, not doing so is foolish; only waiting for death.

The once-proud “corporate welfare” culture in Silicon Valley—free massages, laundry, paid leave—will become history.

Remaining employees must prove their value exceeds what AI compute power can generate at the same cost.

And this trend, sooner or later, or perhaps already, has spread to every corner of the world.

04. Epilogue

Looking back at these events, the core is one word: competition.

Tech giants are competing fiercely over compute power, models, and talent.

They can, like Liu Bang ruthlessly kicking out his children, do anything to win.

Ordinary people are also competing.

What about those middle managers laid off? Most will probably find similar roles in smaller companies.

But small and medium enterprises are also shrinking their workforce, making such positions highly competitive.

…

Future AI might indeed make the world better.

But for now, it’s just making everyone more internally competitive.