capexScore 65/100Research
Zuckerberg: Llama 5 training cluster to scale beyond Llama 4's 100k+ GPUs; infra bet is 'hundreds of billions'
“Llama 3, you know, we could train on you know, 10 to 20,000 GPUs. Llama 4 we could train on you know, 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
First-party quantification of Meta's GPU cluster scaling trajectory (10-20k → 100k+ → larger) and an explicit 'hundreds of billions' infrastructure commitment tied to the scaling hypothesis.
Investment implication
Bullish for GPU/accelerator suppliers (NVIDIA, AMD) and data center infrastructure players if Meta follows through; also frames a risk — a scaling plateau would strand part of that capex.