A Bloomberg story found that Meta has cut roughly 8,000 employees — about 10 per cent of its workforce — while redirecting 7,000 others into newly formed AI teams. Software engineers and machine learning specialists make up the group moving up the new AI hierarchy, while traditional product management, marketing, and mid-tier engineering roles account for most of the departing positions.
The pattern fits what analysts are calling a K-shaped company — a term borrowed from economics. The original concept of a "K-shaped economy" emerged during the COVID‑19 pandemic to describe a diverging recovery, where some segments of society rebounded quickly while others lagged. Picture the letter K: one line slopes upward, the other slopes downward. After a structural shift, different groups move in opposite directions — one side experiences growth and compounding advantage, while the other faces stagnation and long-term erosion. Applied inside a single company, the same logic holds. Those on the upper leg of the K are the AI elite, ascending and expanding their pay and prestige. Those on the lower leg are treated as disposable — their futures more uncertain, their power waning.
Inside Meta, the gap is hard to miss. Top AI executives and researchers are courted with nine-figure compensation packages, and some have reportedly been offered $500,000 in additional equity simply to stay, even as mass layoffs loomed over much of the company. Meanwhile, some non-AI staff have seen budgets slashed even as Meta's profits grow. Workers' keystrokes and mouse clicks have been tracked to train the company's AI agents.
Executives have called the rollout of the company's new AI organisation "atrocious" and acknowledged the workplace has become "brutal." Meta employees are signing petitions, openly criticising management and, in the UK, even attempting to form a union.
For working professionals, the salary data spells out which skills now command the most. AI engineer salaries jumped to an average of $206,000 in 2025, a $50,000 increase from the previous year, and specialised roles in areas like deep learning, LLM fine-tuning, and MLOps are commanding premiums of 30 to 50 per cent above generalist engineering salaries. Large language model expertise leads demand, with nearly 57,000 active job listings and average annual pay approaching $199,000. LLM fine-tuning — a more specialised subset — pays even more, averaging $208,000 per year.
Meta's workforce split is not just a morale story — it is a structural preview of what companies become when AI infrastructure is the only function that justifies permanent headcount. As one Bloomberg columnist warned: "In a K-shaped economy, distrust, fear and cynicism festers among those at the bottom. The same goes for a K-shaped company."
