Meta CEO Mark Zuckerberg criticized the centralization of artificial intelligence development during a media interview on Tuesday [1].

This shift in rhetoric signals a push for a more decentralized AI ecosystem where power is not concentrated within a small number of corporations. Zuckerberg said that the current trajectory limits competition and prevents the creation of AI tools tailored to individual user needs [2].

Zuckerberg advocated for a future defined by "personalized superintelligence for everyone," which would involve AI bots configured to the specific requirements of each person [3]. He said that the centralization of AI development is a problem that concentrates power and stifles innovation [4].

As part of this vision, Zuckerberg addressed geopolitical tensions regarding technology. He said he opposes a blanket U.S. ban on Chinese AI models because such a move would further concentrate AI power in a few hands [5].

To support these ambitions, Meta is investing heavily in infrastructure. Zuckerberg said he will provide "tens of gigawatts" of compute power to Meta’s AI data centers this decade [6]. The company has set a longer-term goal of reaching "hundreds of gigawatts or more" of power over time [7].

This infrastructure push is intended to ensure that the hardware capabilities exist to support widely accessible, high-performance AI models. By scaling compute power, Zuckerberg aims to move away from a model where only a few entities control the most capable systems [2].

“"We need personalized superintelligence for everyone."”

Zuckerberg's stance represents a strategic pivot toward open or decentralized AI, positioning Meta as an alternative to the 'closed garden' approach of other industry leaders. By opposing U.S. bans on foreign models and investing in massive compute power, Meta is attempting to lower the barrier to entry for personalized AI, which could potentially shift the industry standard from a few monolithic models to a multitude of specialized, individual agents.