AI Capex Flow MapHyperscaler → Supplier Attribution
MODELEDForward-year AI-attributable capex · per-company guidance midpoint

Capex Flow Map

Two-stage Sankey of AI-attributable hyperscaler capex routed through infrastructure layers to specific suppliers. Layer mix uses an industry-research baseline; supplier shares within each layer reflect disclosed customer concentration. Every number on this page is a modeled estimate — see Methodology for the math and assumptions.

Hyperscaler → Layer → Supplier

Total AI-attributable capex in scope: $370.4B

MSFTMETAGOOGLAMZNORCLCompute SiliconNetworking & OpticalHBM & MemoryFoundry & EquipmentPower & ElectricalDatacenter BuildNeocloudsNVDAAVGOAMDMUANETCIENVRTETNGEVPWRTSMASMLAMATLRCXCRWV
Hyperscaler capexInfrastructure layerSupplier / beneficiary
Modeled total flow: $663B
Hover any node or path to highlight the connected flow. Width is proportional to dollars.
Layer mix assumption
LayerMix$BRationale
Compute Silicon45.0%$154.8BGPUs + custom AI ASICs (~45% of AI capex per SemiAnalysis composition)
HBM & Memory10.0%$11.1BHBM3E/HBM4 stacks (typically bundled in GPU BoM but allocated here)
Networking & Optical8.0%$29.6BSwitches + optical (back-end fabric ~7–9% of cluster cost)
Power & Electrical15.0%$55.6BSwitchgear, UPS, busway, transformers — rising with density
Datacenter Build12.0%$44.4BDC shell + electrical EPC + land prep
Foundry & Equipment7.0%$25.9BImputed via foundry/wafer cost back to AMAT/LRCX/ASML pass-through
Neoclouds3.0%$10.2BHyperscaler take-or-pay to GPU clouds (MSFT↔CRWV style deals)
Top supplier exposure

Modeled dollar flow from Big 5 AI capex

SupplierPrimary layer$BShare
NVDA
NVIDIA
Compute Silicon$111.8B30.2%
AVGO
Broadcom
Compute Silicon$39.0B10.5%
VRT
Vertiv Holdings
Power & Electrical$36.1B9.7%
ETN
Eaton
Power & Electrical$25.6B6.9%
PWR
Quanta Services
Datacenter Build$24.4B6.6%
ANET
Arista Networks
Networking & Optical$16.3B4.4%
GEV
GE Vernova
Power & Electrical$13.9B3.8%
AMD
Advanced Micro Devices
Compute Silicon$12.9B3.5%
TSM
TSMC
Foundry & Equipment$11.7B3.2%
MU
Micron Technology
HBM & Memory$11.1B3.0%
CRWV
CoreWeave
Neoclouds$10.2B2.7%
ASML
ASML Holding
Foundry & Equipment$5.2B1.4%
AMAT
Applied Materials
Foundry & Equipment$5.2B1.4%
CIEN
Ciena
Networking & Optical$4.4B1.2%
LRCX
Lam Research
Foundry & Equipment$3.9B1.1%
What this is not

It is not a revenue model. Hyperscaler capex is the spend; supplier revenue depends on shipment timing, mix, customer concentration, and competitive substitution. The flow map gives an upper-bound proxy for the addressable dollar base.

Internal accelerators are partially excluded. Google TPU is captured via AVGO (designer); Amazon Trainium and Microsoft MAIA are not in the supplier roster, so the "sum of shares" for AMZN and MSFT compute_silicon intentionally is <100% — the residual stays internal.

HBM is partial. Micron is the only HBM name in scope; SK Hynix and Samsung dominate. The shown $B for memory understates industry-wide HBM TAM by ~3×.