Meta wants to build "superintelligence" for everyone, and CEO Mark Zuckerberg says the tech giant now has all the right ingredients to make that happen.
It all boils down to two key elements: compute and talent, Zuckerberg said on Wednesday during the company's blockbuster second-quarter earnings call.
Meta has made headlines for aggressively recruiting top AI researchers, with some reportedly receiving nine-figure offers.
One of the biggest draws for researchers joining Meta's new AI division is not just top tech salaries, though. It's access to the data center resources needed to train cutting-edge models — what's known as compute, Zuckerberg said on the call.
Compute is a key reason Zuckerberg believes Meta now has "all the ingredients" to build and deploy leading AI models, as it pivots toward superintelligence.
"The people who are joining us are going to have access to unparalleled compute as we build out several multi-gigawatt clusters," he said.
Meta is investing up to $72 billion this year in capital expenditures, much of it to build massive AI infrastructure. One of these clusters will be roughly the size of Manhattan, the company previously said.
Meta is also changing how it hires for AI roles, shifting toward "small, talent-dense teams," Zuckerberg said.
That also helps explain why individual researchers have been able to command sky-high offers in Silicon Valley's ongoing AI talent war.
"For the leading research on superintelligence, you really want the smallest group that can hold the whole thing in their head," Zuckerberg said.
Meta's pivot to superintelligence comes after it reshuffled its AI efforts in part due to the lukewarm reception to its latest model, Llama 4.
Correction: July 31, 2025 — An earlier version of this story misstated what Meta plans to spend $72 billion on. It's spending up to $72 billion in capex, not just on AI infrastructure.
