Meta's AI ambitions under CEO Mark Zuckerberg are facing significant challenges as the company shifts its strategy from open-source models to developing proprietary systems. Last year, Zuckerberg touted the potential of the Llama AI models, but his focus has since shifted. In January, he emphasized Llama during Meta's earnings call, but by October, he barely mentioned it. The company's new direction includes a heavy investment in talent, aiming to compete with AI leaders like OpenAI and Google. Meta's recent $14.3 billion hiring spree, which brought in Scale AI founder Alexandr Wang, signals a shift in priorities. Meta is now working on a successor to Llama, codenamed Avocado, expected to launch in early 2026. Insiders suggest that this model could be proprietary, limiting access for outside developers. This change reflects a growing concern within Meta about the risks of open-source technology, especially after the lackluster reception of Llama 4. The company's leadership has undergone significant changes, with long-time executives being replaced to revamp its AI strategy. The pressure is mounting on new hires, like Wang and former GitHub CEO Nat Friedman, to deliver competitive products amid a rapidly evolving AI landscape. Despite robust revenue from its core advertising business, which exceeds $160 billion annually, Meta's stock performance has lagged behind competitors. Analysts have pointed out that Meta's previous success in AI is now overshadowed by doubts about its investment strategy and return on investment. The company's ability to adapt to the competitive nature of AI development, where significant resources are required to keep pace, will be critical. Meanwhile, rivals like Google and OpenAI continue to introduce advanced models, putting further pressure on Meta to demonstrate its capabilities in this space.