capexScore 50/100Ignore
LeCun confirms Meta's ~350,000 NVIDIA H100 fleet, implying $100B+ GPU plus training infrastructure spend
“Mark made a video, Mark Zuckerberg, uh, very sexy video talking about 350,000 Nvidia H100s... the math of that is just for the GPUs, that's a 100 billion um, plus the infrastructure for training everything.”
Why it matters
Meta's chief AI scientist quantifies an enormous, ongoing GPU commitment, anchoring the scale of AI capex the company has committed to for Llama training and research.
Investment implication
Confirms Meta as one of the largest buyers of NVIDIA H100s, supporting the hyperscaler-capex-driven demand thesis for NVIDIA and AI data center infrastructure suppliers.