At a time when most CEOs are playing defense, Mark Zuckerberg is going all in. The Meta CEO just pulled off one of the boldest reorganizations in the company’s history, dismantling and rebuilding its entire AI division to chase one of tech’s most elusive goals: superintelligence.
As seen in Millionaire MNL, this multi-billion dollar move isn’t just about catching up to OpenAI or Anthropic. It’s about owning the next frontier of machine intelligence, regardless of the cost.
The Gamble of a Lifetime
Internally dubbed the “AI Phoenix” initiative, Zuckerberg’s restructuring consolidated more than a dozen disparate AI teams into one massive unit: the Meta AGI Team, focused solely on artificial general intelligence (AGI).
“We believe we can build superhuman-level AI in our lifetime,” Zuckerberg said in a recent staff meeting. “And we’re going to be the ones to do it.”
Meta is reportedly pouring up to $10 billion annually into the project, including:
A massive compute buildout with NVIDIA H100 clusters
Talent wars with six-figure signing bonuses for top AI researchers
Full integration of Meta’s Llama models into core products like WhatsApp and Instagram
From Social Media to Superintelligence
Meta’s pivot is the clearest sign yet that Zuckerberg is positioning the company not as a social network, but as an AI-first platform.
The shift was catalyzed by internal benchmarking that showed Meta’s Llama 3 model lagging behind GPT-4 and Claude in critical reasoning and multi-modal tasks. To close the gap, Zuckerberg authorized a full AI reboot, even as shareholders questioned the escalating costs.
“This is a generational opportunity,” said Meta CTO Andrew Bosworth. “If we hesitate, we’ll fall behind forever.”
Break Everything, Build New
The restructuring hasn’t been without friction. Multiple senior researchers exited amid disagreements over open-source strategies and safety protocols. Others were re-assigned as Meta adopted a “startup within a mega-cap” mindset, prioritizing speed over process.
Still, insiders say morale is high. Engineers now work under one mission: build an AI system that thinks, learns, and reasons like a human, and eventually, surpasses it.
Risks and Rewards
Superintelligence is a moving target. Critics argue that Meta is diverting from its core business. Others warn of ethical blind spots in its race to develop powerful models with minimal oversight.
But Zuckerberg appears undeterred.
“This isn’t about beating one company. It’s about shaping the architecture of intelligence itself,” he wrote in a company-wide memo.
Can Meta Pull It Off?
The coming year will test whether Meta’s bet pays off. If its models begin outperforming rivals, especially on open benchmarks, the payoff could be enormous. If not, the company risks burning billions and fracturing its product focus.
But as history has shown, Zuckerberg doesn’t mind being early, controversial, or misunderstood, if it means owning the future.
