Meta's multi-million dollar fight to stay ahead in generative AI development

On high alert, Mark Zuckerberg and Meta are spending billions to bolster their artificial intelligence (AI) team and get back in the race, a strategy that has drawn skepticism.

In mid-June, the American social media empire did not hesitate to spend more than $14 billion to acquire a 49% stake in Scale AI , which specializes in data processing used to develop AI models.

According to several US media outlets, the Menlo Park, California-based group had previously contacted OpenAI co-founder Ilya Sutskever, as well as Google's self-proclaimed rival Perplexity AI and video AI powerhouse Runway.

Open AI chief Sam Altman said Meta offered individual signing bonuses of more than $100 million to "many" OpenAI employees and roughly the same amount in annual salary.

On Monday, in an internal memo confirmed by Meta , CEO Mark Zuckerberg said at least seven of them had decided to leave, as well as Scale AI chief Alexandr Wang and several employees from rivals Anthropic and Google. He also said more were coming.

In an internal letter reported by Wired magazine, OpenAI's head of research, Mark Chen, said: "I have a visceral feeling right now, like someone broke into our house and stole something."

Chen noted that the company was working "tirelessly" to contact those with offers and find ways to retain them at OpenAI.

According to several media outlets, it was Zuckerberg himself who launched the initiative, concerned that Meta is lagging behind in generative AI despite having invested tens of billions.

Its last major AI model, Llama 4, released in early April, was a disappointment.

It ranks behind all the American, Chinese, and French heavyweights in the rankings compiled by the independent evaluation platform LMArena on code writing, and even behind its predecessor Llama 3 for its text interface.

Meta wants to integrate its new hires into a new team dedicated to developing "superintelligence," an AI that surpasses human understanding and thought capabilities.

"Uncontrolled spending"

"I think it will attract real talent, and I didn't have many options," blogger Zvi Mowshowitz told AFP. "But this mercenary aspect is very problematic, not to mention the fact that no one wants to work" for Meta and these products unless it's for a very high salary.

"So I don't expect it to work," he says.

On Wall Street, although the stock price is approaching an all-time high and market capitalization is approaching $2 trillion, some are beginning to have doubts.

"Institutional investors are primarily concerned about the company's liquidity (cash flow) and sound capital management," notes Baird analyst Ted Mortonson. "And, for now, there's no counterbalance" to Zuckerberg.

"Those who own shares are saving them for AI advertising, in which Meta is extremely well positioned," he says. "But they're also worried that this spending will get out of control."

In an interview for the Stratechery podcast, Zuckerberg explained that his group plans to completely replace marketing and advertising agencies with AI, and soon offer a turnkey solution directly to advertisers, creating a new source of revenue.

"This doesn't change the short-term profitability potential," says CFRA analyst Angelo Zino, who is also confident about the long term, "because it will create more opportunities, more ways to monetize AI, whether through advertising, connected devices (glasses and headsets), or even Llama."

As for "superintelligence," or general AI, the latter equaling humans and the former surpassing them, "we'll still have to wait at least three to five years," Zino predicts.

"But we need to recruit these people and invest heavily to be ready when we move into this phase."