Meta has a reputation for copying its rivals. Still, its attempt to copy a popular Chinese AI model backfired badly, forcing the company into a multi-billion-dollar overhaul of its entire AI strategy.
In a rush to compete with OpenAI and mimic the success of a Chinese startup called DeepSeek, Meta released a new version of its Llama AI model that flopped with developers. According to people familiar with the matter, the reaction to “Llama 4” was so negative that CEO Mark Zuckerberg decided to spend billions to revamp the company’s AI unit, a process that is still ongoing.
The problem started in January when the sudden popularity of an AI model from DeepSeek caught Meta off guard. DeepSeek’s model used a different, more complex architecture that seemed cheaper and more efficient. Meta scrambled to copy this approach for Llama 4. But while its predecessor, Llama 3, was a hit with developers, Llama 4 proved too difficult for many to customize and use, leading them to stick with the older version.
Zuckerberg reportedly lost confidence in his AI team and is now on a massive hiring spree to regain the company’s footing. In June, Meta invested $14.3 billion in the AI firm Scale AI, bringing its CEO and a handful of employees over to lead a new “Meta Superintelligence Labs.” The company has also hired top researchers from rivals like OpenAI and Google. This all comes with a hefty price tag, and it’s happening at a time when Meta’s revenue growth is slowing.
Zuckerberg is betting that this expensive overhaul is worth it if Meta can get back in the race and create the next generation of powerful AI.
