Meta Platforms CEO Mark Zuckerberg announced plans to invest hundreds of billions of dollars in AI data centers to boost superintelligence capabilities.

The initiative includes multi-gigawatt facilities, with the first center, Prometheus, launching in 2026. Another center, Hyperion, could reach up to 5 gigawatts.

Meta is also developing titan clusters, one expected to rival Manhattan in size.

Its AI operations have been reorganized under a new division, Superintelligence Labs, to unify R&D efforts.

Meta raised its 2025 capital expenditure forecast to US$64 billion to US$72 billion.

Despite investor concerns, Zuckerberg pointed to strong advertising revenue, which hit nearly US$165 billion last year.

Superintelligence Labs will be led by former Scale AI CEO Alexandr Wang and ex-GitHub chief Nat Friedman.

🔗 Source: Reuters

🧠 Food for thought

1️⃣ The economics of AI infrastructure is becoming a critical competitive battleground

The scale of Meta’s planned investment reflects the emergence of AI infrastructure as a new battleground for tech dominance.

While Meta plans to spend significant amounts on multi-gigawatt data centers, other tech giants are taking different approaches to gain cost advantages in AI compute infrastructure.

Google has developed custom Tensor Processing Units (TPUs) that provide an estimated 80% cost advantage over companies relying on Nvidia GPUs 1.

This infrastructure competition directly impacts operational economics. For companies like OpenAI, compute costs are projected to exceed 80% of total expenses by 2025 1.

Meta’s ability to fund this massive investment from its advertising business ($165 billion revenue last year) demonstrates how established revenue streams are being leveraged to build future AI capabilities.

Companies without such revenue streams face significant challenges competing at this scale, potentially widening the gap between established players and startups in the race for advanced AI capabilities.

Recent Meta developments