Meta CEO Mark Zuckerberg announced on July 14, 2025 that the company will invest hundreds of billions of dollars to build multiple massive AI data centers as part of its push toward superintelligence.
He emphasised that the strength of Meta’s core advertising business provides the necessary capital to underwrite this unprecedented expansion in computing infrastructure.
Meta reorganised its AI efforts under a new division named Superintelligence Labs following setbacks with its Llama 4 model and key staff departures earlier this year.
To jump-start capabilities beyond human-level intelligence, Meta invested $14.3 billion in Scale AI, hired its CEO Alexandr Wang, and brought on former GitHub chief Nat Friedman and Safe Superintelligence founder Daniel Gross.
Zuckerberg outlined plans for several multi-gigawatt clusters designed to supply industry-leading compute per researcher:
| Cluster Name | Location | Power Capacity | Expected Online | Notes |
|---|---|---|---|---|
| Prometheus | New Albany, Ohio | 1 GW | 2026 | Multiregion “supercluster” |
| Hyperion | Richland Parish, LA | Up to 5 GW | Phased through 2030 | Campus spanning Manhattan–sized real estate |
| Titan Series | Multiple US sites | Multi-GW each | Ongoing | Additional clusters beyond Prometheus/Hyperion |
Prometheus will be Meta’s first 1 gigawatt build, using natural gas turbines for rapid deployment. Hyperion is slated to scale to 5 gigawatts over several years, rivaling the output of five nuclear reactors and powering roughly 3.5 million homes for a year.
Meta also plans multiple additional “Titan” clusters, each covering substantial real-estate footprints comparable to Manhattan blocks.
In April, Meta raised its 2025 capital expenditure outlook to $64 – $72 billion, aiming to keep pace with rivals OpenAI and Google DeepMind.
Beyond that, Zuckerberg’s pledge extends total spending into the “hundreds of billions,” reflecting multi-year commitments to data center construction and operational costs.
Semianalysis estimates Meta is on track to be the first lab to bring a 1 GW-plus supercluster online, underlining its ambition to lead the AI infrastructure race.
To fuel its superintelligence ambitions, Meta has aggressively poached elite AI talent:
- Alexandr Wang, former Scale AI CEO
- Nat Friedman, ex-GitHub chief executive
- Daniel Gross, co-founder of Safe Superintelligence
- Ruoming Pang, Apple’s head of AI foundation models
These hires form the core of Superintelligence Labs, which Zuckerberg personally oversees to ensure an “elite and talent-dense team” driving frontier research.
Meta’s mega-data center push places it squarely against deep-pocketed rivals including:
- Microsoft-backed OpenAI
- Alphabet’s Google DeepMind
- Amazon-backed Anthropic
- Emerging players such as Perplexity
OpenAI alone is investing $500 billion over four years on its Stargate data center, with a planned 5 GW facility in Texas backed by Oracle and SoftBank.
Meta’s $1.82 trillion market cap dwarfs OpenAI’s $300 billion valuation, but the ultimate payoff remains uncertain as the AI arms race intensifies.
Data centers of this magnitude carry significant environmental footprints:
- Prometheus’ natural gas turbines may boost local emissions, raising air quality concerns.
- Fast Company researchers estimate AI center pollution could cause up to 1,300 premature deaths yearly by 2030 and incur $20 billion in annual public health costs today.
- Water shortages in Georgia communities adjacent to Meta sites highlight strain on local resources.
The U.S. Department of Energy is fast-tracking energy projects – coal, nuclear, geothermal, natural gas – to meet surging AI power demands, framing data centers as critical infrastructure for national competitiveness and security.
Meta’s investment bet hinges on translating raw compute into AI breakthroughs that unlock new revenue streams – image-to-video ad tools, advanced AI assistants, and smart glasses applications.
Share prices rose modestly after the announcement, reflecting cautious optimism among investors even as concerns linger over the scale, cost, and environmental trade-offs of the superintelligence pursuit.
