The secret team Mark Zuckerberg is creating to master AI: it cost him 15 billion

Mark Zuckerberg doesn't want Meta to be relegated to a secondary role in the global race for artificial intelligence . Although its LLaMA models have received attention in the world of generative AI, the company has been overtaken in the media by rivals such as OpenAI, Google DeepMind, and Anthropic. Now, Meta is striking back with an ambitious, secret, and multi-million-dollar plan: to create an artificial superintelligence capable of matching or surpassing humans .

At the heart of this endeavor is so-called artificial general intelligence (AGI) . Unlike current models like ChatGPT, which are designed for specific tasks (writing, translating, summarizing, programming, etc.), AGI would be an intelligence capable of reasoning, learning, and adapting autonomously to any type of task , just as a person would.

No such system exists yet , but tech giants consider it the "holy grail" of AI . And it's also the most controversial , due to the ethical, political, and economic implications of a machine with human-like or superior cognitive abilities.

Zuckerberg wants to lead that revolution. And he's paid $15 billion to do it.

As reported by Bloomberg and The Information , Meta has agreed to acquire a 49% stake in Scale AI for approximately $14.8 billion . This US startup is a leading provider of labeled data for training AI models, which is essential for building complex and reliable systems.

But it's not just the money that matters most, it's the talent: Scale AI founder and CEO Alexandr Wang will join Meta as head of a new internal "superintelligence" lab . His signing is strategic: Wang is considered one of the most influential young people in the sector, with experience in both technology and business leadership.

Meta already has a dedicated AI team, but this is a different story. According to Bloomberg , Zuckerberg is forming a separate unit, with around 50 experts, dedicated exclusively to developing AGI . It's a discrete project, separate from the rest of the company, with independent leadership and its own resources.

Zuckerberg is said to be personally involved in the recruitment, in part due to his frustration with the performance of LLaMA 4 , its latest language model, launched in April. The reception was not as expected , and Meta even decided to delay the launch of the more powerful Behemoth due to internal concerns about its actual capabilities.

With this new team and the arrival of Wang, the goal is to overcome these failures and develop a more advanced, autonomous, and competitive intelligence , capable of competing with future models from OpenAI or Google.

Why is Scale AI key?

Founded in 2016, Scale AI focuses on what many consider the least visible, yet most critical, part of artificial intelligence: preparing the data used to train it . Without quality data, models don't work .

The company works with governments and large technology companies and generated $870 million in revenue in 2024. This year, it expects to exceed $2 billion. According to The Information , it had more than $900 million in cash at the end of last year, indicating a very strong financial position.

By not acquiring it outright, Meta avoids potential regulatory intervention from antitrust authorities , who have already investigated previous acquisitions like those of Instagram and WhatsApp. But with a 49% stake, and in exchange for nearly $15 billion, Zuckerberg is guaranteed priority access to talent, data, and technology.

A war of titans for the next era of AI

Meta isn't alone in this race. OpenAI has already publicly stated that its mission is to develop AGI. Google-owned DeepMind was founded with the same purpose . Companies like Anthropic, Inflection AI, and Elon Musk's startup xAI are also exploring this field with billions in investments.

But this transaction marks a turning point. Meta isn't just buying technology: it's reorganizing its internal structure to focus primarily on AGI , reflecting the extent to which it sees this as the next major technological frontier.

If Zuckerberg succeeds in his bet, Meta could go from being a company known for Facebook and Instagram to becoming the first giant to create autonomous and universal artificial intelligence. This would radically change the relationship between humans and machines, with applications in medicine, science, education… and also considerable risks .

For now, AGI remains a promise. But with $15 billion on the table and a secret laboratory underway, the future we imagined for decades to come could be closer than we think.