Meta’s internal “K-shaped split” is intensifying: elites get ten-million-year salaries, while ordinary employees become “AI fodder”

Meta is staging a Silicon Valley-style experiment in class division. As AI reshapes the organizational structures of major tech companies, a new fault line is deepening inside enterprises: an AI elite class enjoys sky-high pay and limitless resources, while other employees face the looming shadow of layoffs, collapsing morale, and even start to hope they will be included on the layoff list.

This phenomenon is called a “K-shaped company”—mirroring the logic of income inequality in a K-shaped economy. Inside the company, two clearly different tracks are forming: at the top, AI elites’ compensation and status keep rising, while at the bottom, ordinary employees are increasingly treated as replaceable parts.

Meta employees have already publicly criticized management, launched joint petitions, and in the UK even attempted to form a union.

Management has openly acknowledged that such practices are problematic and is rolling out a series of soothing measures—but analysts say the structural imbalance is far from being something that can be bridged by improving employee welfare alone.

CEO Mark Zuckerberg recently admitted that the company’s progress in developing AI agents has not met expectations. Meta’s latest Muse Spark AI system also failed to match the leading models from OpenAI and Anthropic on key benchmark tests.

The AI elites’ feast: nine-figure compensation and special treatment

Within Meta, the perks enjoyed by top AI talent are vastly different from those of ordinary employees.

Compensation packages for the company’s top AI executives and researchers can reach nine figures (tens of millions of dollars), and some people are offered additional equity incentives of up to $500k, with the condition being only that they stay—while at the same time, a cloud of large-scale layoffs hangs over other parts of the company.

One Meta employee said plainly that “the only ones who are really doing well inside the company are the group with the highest pay and the closest ties to AI development.” Another employee put it even more directly: “The only people who don’t feel dissatisfied are those executives.”

This internal split is not unique to Meta; it reflects the two-track employment structure that has long existed in Silicon Valley—differences in treatment between full-time employees and contractors have been around for years. But the new variable brought by this AI wave is that even full-time employees face the risk of marginalization if they are not inside the AI core circle.

The plight of ordinary employees: monitoring, layoffs, and morale collapse

For employees in the lower tier of a K-shaped structure, the situation is quite different. Some employees are forcibly placed into a team described by insiders as “suffocating”; headcount continues to be cut, and while company profits grow, the compensation median actually declines.

More controversial is that the company previously tracked employees’ keyboard keystrokes and mouse clicks to train AI agents—before the project was shut down after it encountered security vulnerabilities that employees had warned about earlier.

CTO Andrew Bosworth summarized management’s overall stance toward employees with unusual directness: “You can leave, or you can keep dissent and still do the job.”

Low morale has evolved into a management crisis Meta now has to confront directly. In expectations for the latest round of layoffs, some employees even hoped they would be among those laid off—an outcome that carries deep symbolic meaning for Meta, which had long been viewed as one of the most desirable employers in US corporate circles.

Amy Edmondson, a professor at Harvard Business School, said in an interview: “They’ve mixed up Silicon Valley’s ‘fail fast’ mantra with reckless execution. This is not the chaos that should necessarily accompany AI innovation—it’s a failure of basic management.”

AI progress lags behind expectations, trust in management is depleted

Internal turmoil has begun to negatively affect Meta’s core strategic goals. Zuckerberg recently publicly admitted that Meta’s AI agent development acceleration is slower than executives expected. The company’s newly launched Muse Spark AI system also lags behind OpenAI and Anthropic’s main models on performance metrics, and it has similarly faced setbacks in building open-source AI models.

Management internally has also realized where the problem lies. Executives privately described the rollout of the company’s new AI organizational structure as “absolutely terrible,” and acknowledged that the work environment has become “brutal.”

When Zuckerberg recently announced a hackathon to boost morale, reports said employees’ reactions were lukewarm—many not only had their energy drained, but some even said directly: “I’m not sure this company supports the hackathon culture anymore. I don’t feel enough security to spend time on innovation that isn’t tied to performance metrics.”

Bosworth, Meta’s CTO, made a rare and candid diagnosis in an internal memo: “We’ve damaged your trust—you believed your professional capabilities and contributions would be valued, that you could grow and be promoted, and that this is a place where you can truly make an impact.”

The cost of fragmentation and an uncertain fix

Meta has currently rolled out a series of measures to soothe employee sentiment, including promises to provide more detailed management support, higher stability, no large-scale layoffs this year, increased team budgets, and improvements to office infrastructure. A widely criticized non-fixed seating policy is also being narrowed.

But Laszlo Bock believes this may signal a deeper turning point: in the post-pandemic era, employees’ continuously compressed voice may be entering a stage where the “green shoots” of resistance are rebounding again. Meta’s case shows that treating employees as replaceable parts rather than partners ultimately backfires in ways that harm the business.

The core paradox of this K-shaped split is that AI was supposed to flatten organizational structures, yet in reality it has produced an even more rigid power arrangement. How to drive technological leaps while maintaining basic trust within the organization has become a management question that large tech companies cannot avoid.

Risk warning and disclaimer

        The market involves risks; investment is需谨慎. This article does not constitute personal investment advice, and it does not take into account any individual users’ special investment objectives, financial situations, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article are consistent with their specific circumstances. Invest at your own risk based on this.