capexScore 65/100Watch
Meta planning 100,000+ GPU clusters for Llama 4, scaling further for Llama 5; betting 'hundreds of billions' on continued AI scaling infrastructure
“Llama 4 we could train on more than 100,000 GPUs. Llama 5 we can plan to scale even further... we're basically making these bets on how much infrastructure to build out for the future and this is like hundreds of billions of dollars of infrastructure.”
Why it matters
Meta is explicitly committing to massive GPU cluster expansion across multiple model generations. This is a concrete capex signal tied to specific compute scales and multi-year infrastructure buildout, directly impacting demand for GPUs, data center infrastructure, and power/cooling systems.
Investment implication
NVIDIA GPU demand from Meta alone will remain elevated through Llama 5 training cycles. Indirect beneficiaries include data center operators, power infrastructure suppliers, and advanced cooling vendors. Zuckerberg's caveat about possible scaling limits suggests he's hedging, but the 'hundreds of billions' figure underscores committed spending regardless.