Meta's $135 Billion Gamble: Sunsetting Reality Labs to Build the Foundations of AGI
Dillip Chowdary
Strategy Correspondent • March 25, 2026
Mark Zuckerberg has officially pivoted Meta once again. In a bombshell announcement today, the company confirmed it is reallocating massive portions of its Reality Labs budget to fund a $135 billion AI infrastructure spree.
The pivot includes several hundred layoffs across the VR and Metaverse divisions, marking a symbolic end to the aggressive "Metaverse First" era. Instead, Meta is doubling down on Superintelligence, with plans to invest $600 billion in U.S.-based data centers by 2028.
The $9 Trillion Valuation Goal
Meta's internal strategy, codenamed "Project North Star," aims to reach a $9 trillion market capitalization by 2028. To incentivize this, top executives have been offered stock options that only vest if this valuation is achieved. This would place Meta significantly ahead of today's tech leaders, assuming its AI infrastructure pays off.
The investment focuses on building "AI Factories"—massive data centers powered by custom MTIA silicon and hundreds of thousands of NVIDIA B200 GPUs. These factories will be used to train Llama 5 and beyond, aiming for true Artificial General Intelligence (AGI).
Why Now? The Reality Labs Sunset
While Quest headsets remain popular, the capital requirements for high-end AI have forced Zuckerberg's hand. The Reality Labs division has burned through tens of billions with limited returns. By cutting "recreational" metaverse projects, Meta can redirect high-performance compute to its core AI mission.
Zuckerberg noted in an internal memo that "the metaverse is the interface, but AI is the brain." By prioritizing the brain, Meta hopes to own the foundational intelligence that will eventually power the AR glasses and spatial computing devices of the 2030s.
Conclusion
Meta's $135 billion annual CAPEX is a staggering figure, dwarfing the GDP of many nations. It is a winner-take-all bet on AGI. If Meta succeeds in becoming the primary provider of open and closed-source intelligence, its $9 trillion goal may not be as far-fetched as it sounds.