ROOT / COMPUTE // HYPERSCALER CAPEX
// Compute · decoded · deep-research pass

Hyperscaler Capex, decoded.

Where the AI money actually goes, tracked per quarter from earnings.

LIVE · feed-ready◊ confidence: Medium
$100B+
quarterly total
~55%
avg YoY
~70%
AI share
5
hyperscalers
// the signal · what the data says
Combined hyperscaler capex is running well over $100B/quarter and climbing, with the majority now AI-directed. This is the clearest financial signal of the build-out and a leading indicator for every downstream layer (fabs, HBM, power). Figures should be refreshed each earnings season.
// comparison · every entry, benchmarked & annotated
NameMakerCapex/Q ($B)YoY (%)AI share (%)
AmazonAWS304060
◊ Largest absolute capex; Trainium + NVIDIA
MicrosoftMSFT255570
◊ Azure + OpenAI commitments
AlphabetGoogle244565
◊ TPU + NVIDIA fleet
MetaMeta226075
◊ Multi-GW Hyperion/Prometheus
OracleOracle109080
◊ OCI + Stargate backlog
// leaderboard · Capex/Q ($B)
01Amazon30
02Microsoft25
03Alphabet24
04Meta22
05Oracle10
// sources & where to verify
// full index, per-field sourcing & CSV export   [ AUTHENTICATE → ]
◊ Deep-research pass, mid-2026. Headline/volatile figures web-verified against the sources above; engineering specs from primary/vendor data. Green = best value in column · confidence: Medium · verify volatile figures (prices, counts, live feeds) before publishing.
← AI Training Clusters◊ indexFoundation Models →