Mark Zuckerberg's planned datacentre spanning Manhattan may just be the social network founder's most aggressive infrastructure gamble yet, pledging Meta to up to US$65 billion in AI spend this year.
The multi-gigawatt facility, dubbed Prometheus, launches in 2026, marking a dramatic escalation in the computing arms race where Meta, Amazon, Alphabet and Microsoft collectively have earmarked ~$320 billion for tech buildouts for 2025 so far.
Zuckerberg's commitment of hundreds of billions toward computing infrastructure dwarfs typical corporate expenditures and signals Meta's determination to avoid trailing rivals such as OpenAI and Google.
Prometheus alone would consume over two gigawatts - equivalent to two nuclear power stations.
And Hyperion, another planned site, is scaling to 5 gigawatts across several years.
The buildout will incorporate 1.3 million NVIDIA graphics processors by year-end - ~50% growth on 2024's outlay and over 2X 2023 figures.
Tech titans flood sector with capital
Amazon leads with $100 billion targeted spending, while Microsoft allocates $80 billion for fiscal 2025 workload centres and Alphabet targets $75 billion in capex.
This surge follows Project Stargate's unveiling - a $500 billion collaboration to build sprawling U.S. data centres.
McKinsey projects chip hardware will command the largest allocation, estimating infrastructure investments reaching $6.7 trillion by 2030.
Meta recently acquired 49% of Scale AI for $14 billion, bringing its founder Alexandr Wang aboard.
