Meta CEO Mark Zuckerberg is betting on a different future for artificial intelligence than many of his competitors. Instead of building centralized AI systems for business or research, Meta is focused on creating what Zuckerberg calls "personal super intelligence"- an AI tailored for everyday life and new hardware.
While most companies are working on centralized AI to automate business processes or tackle scientific challenges, Meta is taking another route. In an interview with The Information, Zuckerberg said Meta's mission is to "deliver personal superintelligence to everyone in the world... so that way you know we can put that power in every individual's hand."
The idea isn't entirely new, but Zuckerberg is pushing it further than most: AI shouldn't just be a tool for boosting productivity, but a personal assistant for daily and social needs. He explained that a lot of what people care about "is relationships and culture and creativity and having fun and enjoying life."
A future where not wearing AI glasses is a disadvantage
A key part of Meta's vision is integrating this AI into hardware, especially AR glasses. Zuckerberg outlined a future where people experience the world through intelligent eyewear that can see what you see, hear what you hear, and help guide you through your day. He believes that in the future, not having them will be a significant handicap. "If you need vision correction, and you don't have glasses... you're sort of at this cognitive disadvantage," Zuckerberg said. "And I think in the future, if you don't have AI glasses, you're basically going to be at a cognitive disadvantage, too."
Meta believes these glasses could become the main interface for AI applications. Early products from Meta's Reality Labs are already on the market, and Zuckerberg sees them as a bridge between hardware, AI, and mass-market adoption.
Data centers in tent form
The technical groundwork for this vision is being built at massive scale. Meta is investing billions in new data centers with power capacities of several gigawatts. The biggest clusters are named Hyperion and Prometheus - a nod to Greek mythology.
To speed up construction, Meta is skipping traditional concrete buildings. Instead, they're setting up weatherproof, hurricane-proof tent structures to house GPU clusters. Hyperion is set to reach up to five gigawatts in the coming years-an area that Zuckerberg says covers "a significant portion of the footprint of Manhattan in terms of space." The goal: build the largest compute fleet of any company.
Self-funded expansion
Unlike some competitors, Meta can finance these investments internally. Zuckerberg says the company's existing ad and platform business "throws off a lot of capital" and allows them to fund the superintelligence infrastructure from their own cash flow. "We have a very strong business model that supports this," he said. Other companies need to raise capital for similar projects.
Zuckerberg sees this as a strategic play: by investing heavily now, Meta can gain an edge in research and product development. Reinforcement learning also amplifies this effect, as user feedback helps refine future AI models.
Self-learning systems for product development
According to Zuckerberg, the first signs of this new AI strategy are already visible in Meta's products. He said they are "already seeing glimpses of models improving the Facebook algorithm and other AI systems across the company autonomously." This self-optimizing effect is a preview of what's coming: AI that improves itself and helps build new products faster and more efficiently.
Asked if Llama lags behind the competition, Zuckerberg admitted that "the field keeps on moving faster than we expect."
Llama 4's performance may be one reason for Meta's new strategy: the company is ramping up infrastructure, forming new teams, and recruiting top researchers. The goal is to catch up with - or even surpass - rivals with the next generation of models. The superintelligence initiative is meant as a fresh start.
Compute over hierarchy
Reports of nine-figure compensation packages for talent poached from Apple, Google Deepmind, and OpenAI were downplayed by Zuckerberg. The market is competitive, he said, but compared to the billions spent on infrastructure, talent costs are a small piece.
Zuckerberg argues that infrastructure is the real advantage. He's found that top researchers have a new priority. "Here people say I want the fewest number of people reporting to me and the most GPUs," Zuckerberg explained. With top AI researchers in high demand worldwide, that's a strong selling point.
No one knows exactly when superintelligence will be achieved. Zuckerberg thinks it could be soon, but the timeline is uncertain. The important thing, he says, is to start preparing now. "I just think that we should bet and act as if it's going to be ready in the next 2 to 3 years," he said. "I believe that there's a shot at that and I think if that's what you believe then you're going to invest hundreds of billions of dollars."
"Personal Super Intelligence" as a bet on the future
How Meta plans to make money from "personal super intelligence" is still unclear. Zuckerberg stressed that the core business could fund the necessary investment, but it's an open question whether the new AI products will pay for themselves. Unlike traditional software subscriptions or cloud services, there are no proven business models yet for selling personal AI companions.
Zuckerberg suggested users might pay directly for a premium, ad-free AI experience-similar to ChatGPT Plus - but whether billions of people worldwide are willing to do that remains to be seen. For now, Meta seems to be betting that a future AI platform, perhaps built around smart glasses, will create new ways for people to interact, leading to new business models over time. Until then, the race for superintelligence is being funded by today's profits.
