Zuckerberg Handed Six Executives $921 Million — Then Fired 700 Employees That Same Afternoon

What does it tell you about the state of artificial intelligence when a company worth $1.5 trillion pays a single AI researcher $300 million — while simultaneously firing 700 of its own employees and handing six executives options that could pay out nearly a billion dollars each?

On Tuesday, Meta disclosed in SEC filings that it had granted stock options to six of its most senior executives: Andrew Bosworth, Chris Cox, Javier Olivan, Susan Li, Jennifer Newstead, and Naomi Gleit. It was the first executive option grant since the company's 2012 IPO. The options vest in tranches tied to share price thresholds. To reach full value, Meta's market capitalisation must hit $9 trillion by March 2031 — roughly six times its current valuation of $1.5 trillion, and more than double Apple's present worth. If every tranche vests for Bosworth, Cox, and Olivan, each of them stands to receive up to $921 million.

Hours after the SEC filing went public, Meta laid off approximately 700 employees across Reality Labs, recruiting, sales, and Facebook.

The juxtaposition was not subtle. It did not need to be. Mark Zuckerberg's message to the AI industry is written in the structure of these transactions, not in press releases. The company that eliminated 20,000 positions between 2022 and 2025, cut rank-and-file stock compensation twice in two years, and is now shedding another 700 roles is simultaneously offering individual AI researchers retention packages worth $300 million over four years. OpenAI, Google DeepMind, and Anthropic are all competing for the same narrow pool of people who can actually move the frontier of large language model training, inference optimisation, and GPU cluster design. Meta has decided that the price of losing any of them is higher than the price of paying them whatever it takes to stay.

This is what the AGI race actually looks like from the inside. Not robot demos. Not chatbot launches. A brutal, high-stakes competition for the forty or fifty researchers in the world capable of designing the next generation of foundation models — and a willingness to deploy compensation numbers that would have seemed like satire five years ago.

Zuckerberg has committed $115 billion to $135 billion in capital expenditure for 2026, almost entirely directed at AI infrastructure: data centres, custom silicon, and the compute required to train models at scales that rival anything OpenAI or Anthropic can field. Meta's supercomputer cluster now houses more than 350,000 NVIDIA H100 GPUs, and the company is actively testing its own custom AI training chips to reduce its dependence on NVIDIA hardware. The $9 trillion valuation target that unlocks the final tranche of executive options is not an arbitrary number — it is the number Zuckerberg believes AI will add to Meta's value if the bet lands correctly.

But the bet comes with visible costs. In 2025, Meta's stock-based compensation expenses reached approximately $42 billion, consuming roughly 96 per cent of the company's $43.6 billion in free cash flow. That is not a rounding error. When a company's equity awards absorb nearly all of its free cash generation, its margin for error on growth projections collapses. Meta's advertising business remains enormously profitable, but it is not growing fast enough to absorb both the infrastructure spending and the compensation structure unless AI meaningfully transforms revenue — which is precisely the thesis the executive options are designed to incentivise.

The tension between the executive grants and the employee layoffs is also a signal about where Meta believes value actually lives. The 700 workers cut on Tuesday were in Reality Labs, recruiting, sales, and Facebook — functions Zuckerberg has identified as overhead relative to the AI research and infrastructure investment he is making. The executives who received the options are the people responsible for executing the AI strategy. The AI researchers who received the $300 million retention packages are the people whose weights and training runs will determine whether that strategy works. Everyone else is, in the language of corporate restructuring, a variable cost.

This dynamic is not unique to Meta. Sam Altman's OpenAI recently disclosed that it has tripled its headcount plans, targeting 8,000 employees by end of 2026, and that it is drawing roughly a quarter of its engineering hires from Google alone. Dario Amodei's Anthropic, newly flush with investment and generating revenue at a pace that is closing in on OpenAI's, is hiring aggressively across LLM training, fine-tuning, and safety research. The eleven co-founders who left Elon Musk's xAI in the past few months — researchers from Google DeepMind, Microsoft, and OpenAI — will land somewhere in this market. The question of where is being answered in real time, and the answer will matter for every benchmark result published in the next eighteen months.

The $9 trillion target on Zuckerberg's executive options is a public statement that Meta's leadership believes the company will be the defining AI platform of the 2030s. It is not a promise. The first tranche does not vest until Meta's stock hits $1,116 per share — roughly double its current price. If the company never reaches that threshold, every one of those options expires worthless. Zuckerberg controls the company through supervoting shares and chose not to include himself in the option package, a detail that suggests either that he already has sufficient financial incentive or that he does not intend to share the downside risk he is asking his executives to absorb.

What is certain is that the race to hire, retain, and deploy the people who can build and run frontier AI systems has produced a compensation market that operates by entirely different rules than the rest of the technology industry. A researcher who can design training runs for hundred-billion-parameter models, or who understands the subtle interaction between data quality and loss curves at scale, or who can optimise inference throughput on custom silicon, is worth more to Meta, OpenAI, Google DeepMind, or Anthropic than most companies spend on their entire engineering departments. The $300 million package is not generosity. It is a calculation about what it would cost to let that person walk out the door.

The 700 people laid off on Tuesday were not making that calculation work in Meta's favour. The six executives now holding nearly a billion dollars in options are the people who have to make it pay off.

Deep Dive

For more context on the AI talent and infrastructure arms race, read our earlier coverage:

• Elon Musk Built a $250 Billion AI Lab — Then Every Single Founder Left: How all 11 xAI co-founders departed after Musk admitted the company was "not built right," and what their exit says about the limits of capital without culture.

• Zuckerberg Just Declared War on Intel and x86 — And His Weapon Is a Chip Called AGI: The hardware strategy underneath Meta's AI ambitions — why the company building its own silicon is the real story behind every model it ships.