AI Capex
The billions hyperscalers and AI labs spend each year on GPUs, datacenters, and training clusters · the #1 driver of AI spending narrative.
Text reviewed October 5, 2026
The billions hyperscalers and AI labs spend each year on GPUs, datacenters, and training clusters · the #1 driver of AI spending narrative.
Basic
This capex is the physical buildout behind every AI training run and serving capacity announcement. Current figures, with dates and sources, are on the BenchGecko economy pages.
Deep
Capex accounting: GPUs + data center shell + power infrastructure + networking + cooling. Data center shell + power can be another 30% of total spend. The capex-to-revenue ratio across AI hyperscalers exceeded 50% in 2025 · historically extreme. This gap drives the AI Bubble Index component on BenchGecko. Current figures, with dates and sources, are on the BenchGecko economy pages.
Expert
Capex depreciation: GPUs typically 5-6 year useful life, 2-3 year rapid obsolescence, driving aggressive writedowns. ROI on AI capex is uncertain · training runs deliver capability but revenue scales depend on downstream demand. AI capex has strong enterprise demand but monetization unclear past 2027.
AI capex as % of revenue is at dotcom-era levels. Every earnings call is scrutinized for capex discipline.
Depending on why you're here
- ·The massive amount AI companies spend on computers
- ·Hundreds of billions per year
- ·Why "AI is expensive" · it actually is, at scale
- ·Your pricing as a customer reflects capex amortization
- ·Capex decisions at hyperscalers drive your future serving costs
- ·Watch capex / revenue ratio as a leading indicator of pricing pressure
- ·Capex is the #1 question on every hyperscaler earnings call
- ·ROI uncertain past 2027 · telecom-era parallel real
- ·BenchGecko Bubble Index uses capex-to-revenue as 10% component
- ·GPU + datacenter + power + networking · full stack capex
- ·Depreciation 5-6 year, aggressive obsolescence of 2-3 years
Nobody knows yet · but BenchGecko tracks it daily.
Understanding capex cycles tells you which AI companies are about to cut pricing (high depreciation pressure) vs hold firm (delayed capex).