For nearly a decade, Mark Zuckerberg championed the metaverse as the future of human connection—a vast, immersive universe where people would work, play, and socialize through avatars. That vision, which saw Meta pour tens of billions into virtual reality platforms and devices, has quietly faded into the background. Consumer indifference, sluggish adoption, and technological friction brought the metaverse down to earth faster than its creators anticipated.
Now, Zuckerberg has redirected his gaze—and capital—toward something he considers far more transformative: artificial intelligence. According to multiple internal sources and filings, Meta is committing over €62 billion ($65 billion) to a new AI infrastructure overhaul that aims to rival, and possibly outpace, today’s top players. Unlike the speculative push for digital realms, this time, the investment is deeply rooted in existing demand and rapidly evolving technological capabilities.
In August, Meta struck a multi-year Google Cloud deal worth $10 billion, gaining access to state-of-the-art GPU clusters powered by Nvidia hardware. It’s one of the largest cloud computing deals ever signed, signaling just how urgently Meta wants to scale its AI operations.
From Avatars to Algorithms
Meta’s original metaverse ambitions were sweeping, but they faced persistent friction. Despite marketing blitzes and flagship products like Quest headsets and Horizon Worlds, users failed to engage at scale. As reported by the Wall Street Journal in 2023, Horizon Worlds struggled to attract more than 200,000 monthly users—far below internal goals. Meanwhile, enterprise interest in virtual meetings through VR fizzled as companies returned to simpler remote work tools.
“The metaverse required people to change too much too quickly,” says Dr. Mary Gray, a senior researcher at Microsoft Research and co-author of Ghost Work. “AI, by contrast, integrates into existing behaviors. It doesn’t ask people to reinvent their digital lives.”
That’s a key reason why AI appears to offer more concrete returns—and why Meta is now positioning itself as a full-spectrum AI company. Internally, the program is known as The Goose—a tongue-in-cheek nod to the golden eggs it hopes to lay. The initiative includes expanding data infrastructure, recruiting top AI researchers, and exploring monetizable AI agents across its products.
Building the Pipes of Synthetic Intelligence
To support this pivot, Meta is not just training models—it’s rebuilding its technological backbone. Alongside the Google Cloud partnership, the company is expanding its own infrastructure. In September, it opened a new $1 billion data center in Kansas City, optimized for AI workloads and designed to handle the petabytes of data required to train large-scale models.
Meta’s approach embraces what analysts call a multi-cloud strategy. In addition to Google, it relies on Amazon Web Services, Microsoft Azure, and Oracle, creating redundancy and flexibility in how and where its AI is developed. As Gartner notes, this kind of strategy is gaining momentum as enterprises seek to reduce risk, improve service performance, and gain pricing leverage in a cloud-dominated world.
Meta is also reportedly exploring acquisitions of smaller AI startups across Europe and Asia, though no major deals have been announced publicly. This expansion strategy mirrors the aggressive playbooks seen during previous tech transformations—such as Google’s acquisition of DeepMind in 2014, which helped it dominate the early AI research landscape.
The New Logic of Digital Empires
What makes this shift particularly telling is how it redefines Meta’s role in the internet’s future. For years, the company was synonymous with social media, not infrastructure or foundational AI research. With Facebook and Instagram reaching maturity, AI offers a new pathway for relevance—turning Meta from a content-driven company into the backbone of digital intelligence.
But this transformation comes with scrutiny. Large AI models depend on vast datasets, many derived from user activity. Concerns around data privacy and algorithmic bias are mounting. In May 2025, the European Commission launched a preliminary investigation into how tech giants—including Meta—source and process user data for training advanced AI, under the framework of the EU AI Act.
Still, Meta’s position remains formidable. With billions of user interactions daily, it commands one of the richest data troves in the world—fuel for the next generation of machine learning systems. Combined with a war chest of over $60 billion and partnerships across the cloud industry, the company has the tools to shape what comes next.
