Meta is secretly making big moves.

On the surface, OpenAI and Google are in the limelight, but Meta plans to invest $60-65 billion this year to build AI infrastructure, deploying 1.3 million GPUs by the end of the year—this scale is already on par with tech giants. What's even bolder is the self-developed AI chips that break the dependency on Nvidia, as they globally deploy liquid-cooled data centers to train the LLaMA model.

There has been no hesitation in this area of talent. They poached top researchers from DeepMind, OpenAI, and Anthropic, invested 14 billion in Scale AI, and then directly brought over the founder, Alexandr Wang, to lead the new superintelligence department. This guy dropped out of MIT to start a business, and Zuck's logic is very straightforward: world-class talent grows exponentially like capital.

The real killer feature is integration. LLaMA model → AI features of Facebook/Instagram/WhatsApp → Real-time translation and image recognition of Ray-Ban smart glasses → Feedback loop from 3 billion users. Each deployment generates new data, and each piece of data strengthens the next generation of models.

Some say that Meta is building the operating system of the next era, rather than pursuing the most glamorous models. Execution is the real test, but this company does have the patience to engage in a long-term battle in a large-scale, transparent environment.

Worth paying attention to.