Zuckerberg's $15 Billion Gamble: Meta's 'Superintelligence Labs' Spark Most Costly AI Talent Competition in Silicon Valley

Mark Zuckerberg has a track record of making bold moves that have the potential to transform entire industries, though they sometimes end in failure. After spending $46 billion on the metaverse with limited success, the CEO of Meta is now diving into a new venture: superintelligent AI.

This time, the risks are higher, the rivals more formidable, and the possible outcomes more groundbreaking than anything Meta has tried before.

Offering massive compensation packages and making infrastructure investments far exceeding those of the metaverse, Zuckerberg's superintelligent AI project is set to become one of Silicon Valley's most expensive talent battles. This could either solidify Meta's dominance in the technology sector or serve as yet another example of big dreams clashing with reality.

The launch of Meta Superintelligence Labs represents a new direction for the social media titan. In an exclusive interview with The Information's TITV live-stream, Zuckerberg mentioned to founder Jessica Lessin that the highlight this year is witnessing early signs of self-improvement in models, indicating that superintelligence development is within reach.

This vision has driven Meta to overhaul its AI division entirely, aspiring to deliver what Zuckerberg terms 'personal super intelligence to everyone worldwide.' This lab was established following internal challenges within Meta's AI team, including management issues, staff turnover, and lackluster product launches.

Rather than making gradual changes to existing systems, Zuckerberg has opted for a complete overhaul, bringing in new leadership and rethinking the company's AI development strategy.

At the heart of Meta's superintelligent AI goals is a recruitment strategy that has created shockwaves in the industry. Zuckerberg has embarked on a spending spree to build the new lab, offering massive salaries to lure top researchers from firms such as OpenAI, Google, Apple, and Anthropic.

When The Information's Lessin questioned reports of $100-$200 million compensation packages, Zuckerberg acknowledged the competitive landscape, saying the reports weren't entirely accurate. Still, he noted, 'it is a very hot market, with a small number of top researchers in demand at every lab.'

Meta's AI talent acquisition approach is not only about money. Zuckerberg emphasized that having 'the most computing power per researcher is a strategic advantage, not only for completing work but also for attracting top talent.' This outlook shows an understanding that in the superintelligent AI race, talent density is more critical than team size.

A key part of Meta's talent strategy was acquiring leadership from Scale AI. In June, Meta invested $14.3 billion in the AI startup, taking a 49% ownership stake. Under the agreement, the company's founder, Alexandr Wang, and several key Scale employees assumed leadership roles at Meta.

At just 28, Alexandr Wang now serves as Meta's chief AI officer, heading the newly named 'Meta Superintelligence Labs.' Within this broader AI unit, Wang oversees a team of newly recruited researchers, a few of his former colleagues from Scale AI, and Nat Friedman, the previous CEO of GitHub.

The integration of Wang's team signifies a new approach to AI development at Meta. The group operates in an office space separated from the rest of the company and is situated near Zuckerberg, highlighting the initiatives importance.

One of the most significant changes from Meta Superintelligence Labs is the potential shift away from the company's longstanding open-source philosophy. Recently, key team members, including Wang, discussed moving away from Meta's potent open-source AI model, Behemoth, in favor of developing a closed model.

This marks a departure from Meta's traditional strategy. For many years, Meta has open-sourced its AI models, allowing other developers access to the code for further development. Company leaders have claimed that such transparency facilitates faster AI advancement and broader accessibility for developers.

The shift is a response to competitive pressures in the AI landscape. Meta had completed training its Behemoth model but postponed its release due to unsatisfactory internal performance, prompting a reevaluation of their strategy.

In addition to talent acquisition, Meta is investing in infrastructure to support its superintelligent ambitions. Zuckerberg disclosed that the company is 'constructing multiple, multi-gigawatt data centers' and developing new construction methods, such as 'weatherproof tents,' to speed up their deployment.

The magnitude of these investments is staggering. One of Meta's new data centers, Hyperion, will eventually scale up to five gigawatts in coming years, covering a substantial area comparable to the footprint of Manhattan.

These infrastructural investments are facilitated by Meta's robust financial situation, with Zuckerberg noting, 'we can fund this entirely through the company's cash flow.'

Meta's approach differentiates itself from competitors by focusing on 'personal superintelligence' instead of centralized AI systems. During his interview, Zuckerberg elaborated that while other companies aim 'to automate all economically productive work in society,' Meta's focus is on 'real-world concerns like relationships, culture, creativity, and enjoying life.'

The vision also extends to Meta's hardware goals, especially its AR glasses project. In the same TITV interview, Zuckerberg predicted that 'without AI glasses, individuals may be at a cognitive disadvantage,' describing future scenarios where AI companions could 'observe and follow up on your life's activities.'

The implications of Meta's superintelligence drive reach beyond the company. Meta's aggressive talent acquisition has led to wage inflation in the sector, pushing competitors to match or exceed Meta's offer to retain their skilled researchers.

When asked about industry interactions at Sun Valley, Zuckerberg acknowledged the competitive scene, affirming, 'we're not targeting anyone specifically. My goal is to familiarize myself with all leading researchers in the field.'

This diplomatic stance conceals an underlying zero-sum contest for a limited pool of elite superintelligent AI talent. Furthermore, shifting away from open source development hints at a broader industry trend toward proprietary AI approaches.

Meta's superintelligence initiative signifies a rethinking of the company's future. Following a costly metaverse attempt that didn't pay off, Zuckerberg now aims even higher with AI, with potential investments surpassing $100 billion in the coming years.

The future success or failure of Meta Superintelligence Labs is set to determine not only the company's future but also shape the broader AI landscape. With some employees anticipating an 'exodus of AI talent not chosen for Wang's team,' the stakes are considerable.

Whether Meta's talent acquisition strategy will produce the groundbreaking AI technologies Zuckerberg envisions remains uncertain. What is clear is that Silicon Valley's most expensive talent war has begun.