Meta is pouring $229 billion into artificial intelligence infrastructure over two years, with up to $135 billion allocated for 2026 alone, while simultaneously cutting roughly 16,000 jobs to offset the costs. The strategy raises serious questions about how the company plans to fund this massive bet and what it means for the workers caught in the middle.

How Is Meta Funding Its AI Expansion?

Meta's approach to funding its AI ambitions involves a controversial cost-cutting strategy that goes beyond typical workforce reductions. The company is offsetting its infrastructure spending by eliminating jobs and then using the remaining employees to train the very AI models meant to replace them. In 2026, Meta moved to cut roughly 16,000 jobs, with about 8,000 positions eliminated in the first wave by May, with additional cuts signaled for the future.

The most striking element of this strategy emerged in April 2026, when Meta rolled out the "Model Capability Initiative." This program includes software installed on employees' work computers that records their mouse movements and keystrokes. The stated rationale is that new AI models "learn from watching really smart people do things." Over 1,000 employees signed a petition to stop the surveillance program, but Meta declined to halt it. Some workers even papered the offices with flyers reading "Don't want to work at the Employee Data Extraction Factory?".

• Infrastructure Investment: Meta is committing $229 billion over two years, with $135 billion in 2026 alone, to build the computing infrastructure needed for frontier AI development.
• Workforce Reduction: Approximately 16,000 jobs are being cut to offset AI spending, with 8,000 positions eliminated by May 2026 and more reductions planned.
• Employee Data Harvesting: The "Model Capability Initiative" uses keystroke and mouse-movement tracking software to train AI models on employee work patterns without explicit consent.
• Compensation Cuts: While the company invests billions in AI, median employee pay dropped nearly $30,000 from 2024 to 2025, and rank-and-file workers saw raises slashed.

Why Is Zuckerberg Able to Make These Decisions Alone?

A critical factor enabling Meta's aggressive AI strategy is Mark Zuckerberg's unusual control over the company. While Zuckerberg owns only 13.6% of Meta's shares, he controls 99.7% of the Class B shares, which carry ten votes each. This structure gives him 61.2% of the voting power, making him effectively unaccountable to institutional investors.

The major institutional investors, including BlackRock, Vanguard, and State Street, collectively own roughly 20% of Meta but cannot fire Zuckerberg, change his board, or even exit their positions. Their index funds are required to hold Meta stock, making them trapped shareholders. This concentration of power means Zuckerberg answers to no one but himself when making strategic decisions.

What Is Meta's Track Record With Unchecked Spending?

Zuckerberg's history with Meta's metaverse division offers a cautionary tale about what happens when a CEO has absolute control and no accountability. From 2020 through 2025, Meta's metaverse division bled a cumulative $83.6 billion. The flagship product, Horizon Worlds, was downloaded 45 million times but generated only $1.1 million in lifetime consumer revenue against that division's $83.6 billion in losses.

Even when the company attempted to shut down Horizon Worlds in March 2026, Zuckerberg reversed the decision within 48 hours after facing backlash. The investors could only watch as billions were spent on a project that failed to generate meaningful returns. This pattern suggests that Meta's current $229 billion AI investment carries significant risk, particularly given that the company is funding it through workforce cuts rather than organic growth or profitability.

What Does This Mean for Meta's AI Strategy?

Meta's approach to AI development differs markedly from its earlier strategy. When ChatGPT launched in late 2022, Zuckerberg realized Meta had fallen behind in consumer AI products. In response, Meta open-weighted its Llama AI model in 2023, giving it away free to companies including Goldman Sachs, AT&T, and DoorDash. Zuckerberg called this "democratizing AI," but the move had a strategic purpose: flooding the market with a free standard to bury the competition.

The strategy backfired. Chinese AI labs downloaded Meta's freely released model weights and improved upon them. DeepSeek and other competitors used Llama to train smaller, cheaper models that now compete with top US frontier systems for a fraction of the cost. Zuckerberg had armed a rival power and watched it lap him.

Now that Meta is competing at the frontier instead of chasing it, the "democratization" rhetoric has quietly disappeared. Meta's newest model, Muse Spark, is closed-source, with the weights locked up tight. The moment openness stopped serving Zuckerberg's interests, he abandoned it. This shift reveals what has always been Meta's guiding principle: self-interest.

The $229 billion AI bet represents Zuckerberg's attempt to regain competitive ground in a rapidly evolving field. However, the strategy of funding it through workforce reductions and employee surveillance raises ethical questions about whether the company is prioritizing shareholder value and technological dominance over worker welfare and privacy. With Zuckerberg's unchecked control and Meta's history of aggressive spending on failed projects, the long-term viability of this approach remains uncertain.