Meta PlatformsMETA
Meta is spending heavily on the largest AI training clusters in the world to power Llama and ranking — almost entirely Nvidia today.
Where the cash goes~$146B / yr · FY2025
Capex$68B
R&D$43B
Buybacks$30B
Dividends$5B
Money flows left → right · hover a node · click to open the stock.AI AcceleratorsCustom SiliconNetworkingFoundryPower & Cooling
Most exposed to META's spend
- NvidiaNVDACurated
Hundreds of thousands of H100/B200 GPUs for training clusters.
AI Accelerators - BroadcomAVGOFiling
Co-develops Meta's MTIA inference silicon and networking.
Custom Silicon - Arista NetworksANETCurated
Ethernet fabric for AI superclusters.
Networking - TSMCTSMCurated
Fabricates GPUs and Meta's custom MTIA chips.
Foundry - VertivVRTCurated
Power and liquid cooling for high-density racks.
Power & Cooling
Relationships are hand-curated and sourced from filings, announcements, and public reporting — a map for generating ideas, not a complete supply chain, and not investment advice. Exposure weights are relative, for sizing only. Spot something missing? Contributions welcome on GitHub.