META
$72 billion, two superclusters, and a CEO on a mission
Meta is going all-in on AI, with plans to spend hundreds of billions over the next few years.
Mark Zuckerberg says this will fund the build of two massive data centres, Prometheus, launching in 2026, and Hyperion, which could eventually scale up to 5 gigawatts.
These aren’t your average server farms; they’ll be almost the size of Manhattan.
So, HUGE.
To back it all up, Meta’s been on a hiring spree, offering eye-watering pay packages (some hitting $100m) to top AI talent.
Here’s what’s happening:
Meta’s AI infrastructure plans include gigawatt-scale data centres like Prometheus and Hyperion.
Capital spending is set to hit up to $72bn in 2025 to stay competitive.
Superintelligence Labs will focus on turning cutting-edge research into real-world tools.
The company recently pulled in former Scale AI CEO Alexandr Wang and ex-GitHub boss Nat Friedman to lead its newly formed Superintelligence Labs, a fresh division focused on taking AI from the lab into real-life products like Meta AI, smart glasses and image-to-video ad tools.
Some investors are still cautious, wondering if this scale of spending will pay off.
But Zuckerberg says Meta’s $165bn revenue last year and strong ad business give it the headroom to bet big.
Analysts agree that AI is already helping them push more ads at better margins.
I wanna go into Zuck’s brain just to understand what his final plan is…