AI & computing infrastructure
AI adoption drives investment in computing capacity, data centres and power. The next test is whether business use of models generates enough revenue and savings to justify that spending.
What the evidence says
IEA’s Key Questions on Energy and AI (16 April 2026) updates its analysis of data-centre electricity demand and grid and supply-chain constraints. Access to power remains an important condition for AI expansion.
IEA · Key Questions on Energy and AI · 2026 ↗Value chain — where revenue is generated
- Semiconductors and memory
- Networking, cooling and data centres
- Cloud and industry software
Analytical view: distinguish infrastructure suppliers from AI adopters. Suppliers benefit from capital spending; adopters must demonstrate productivity and retain the resulting economic benefits.
Key risks
Overinvestment, rapid hardware obsolescence, model pricing pressure, export restrictions and demanding valuations.
What to monitor
AI revenue versus capital spending; data-centre utilization; inference cost; margins and free cash flow.
What would undermine the thesis?
Persistently falling returns on capital despite rising AI spending would weaken the thesis.