ARTICLE AD BOX
Meta CEO Mark Zuckerberg appears astatine the Meta Connect lawsuit successful Menlo Park, California, connected Sept. 25, 2024.
David Paul Morris | Bloomberg | Getty Images
Meta CEO Mark Zuckerberg connected Monday said helium plans to put "hundreds of billions of dollars" into artificial intelligence compute infrastructure, and that Meta plans to bring its archetypal supercluster online adjacent year.
A supercluster is simply a large, analyzable computing web that's designed to bid precocious AI models and grip their workloads.
"Meta Superintelligence Labs volition person industry-leading levels of compute and by acold the top compute per researcher," Zuckerberg wrote successful a Facebook post connected Monday. "I'm looking guardant to moving with the apical researchers to beforehand the frontier!"
Zuckerberg said Meta's archetypal supercluster is called Prometheus, and that the institution is gathering respective different multi-gigawatt clusters. One cluster, called Hyperion, volition beryllium capable to standard up to 5 gigawatts implicit respective years, helium said.
Zuckerberg has been connected a multibillion-dollar AI hiring spree successful caller weeks, highlighted by a $14 cardinal concern successful Scale AI. He announced a caller organization in June called Meta Superintelligence Labs that's made up of apical AI researchers and engineers.
Zuckerberg had grown frustrated with Meta's advancement successful AI, particularly aft the release of its Llama 4 AI models in April received a lukewarm effect from developers. He is revamping Meta's attack to amended vie with rivals similar OpenAI and Google.
"For our superintelligence effort, I'm focused connected gathering the astir elite and talent-dense squad successful the industry," Zuckerberg wrote Monday.
English (US) ·