World & economy / Economics
AI’s Next Bottleneck: Why the Data-Centre Boom Is an Electricity Story
AI is discussed as software. Its expansion also depends on equipment, land, finance and electricity. That makes the current infrastructure race an economics story as much as a technology story.
By Farvenzo Explained · Published 22 October 2026
Why this topic is timely
NVIDIA’s GTC Berlin programme lists its AI conference for 21–22 October 2026, with the keynote on 21 October. Its agenda includes infrastructure, industrial applications and sovereign AI. Those are company event themes, rather than independent evidence that promised economic benefits have already arrived.
The useful question for readers is what has to happen between an announcement and a productive service. A large spending plan is an input. The resulting capacity, customer demand and commercial value need to be assessed separately.
The physical constraint behind the digital story
The IEA’s 2026 analysis projects global data-centre electricity consumption rising from 485 TWh in 2025 to about 950 TWh in 2030. This covers data centres overall, not AI alone, and the 2030 number is a projection.
Its discussion highlights constraints involving grids, equipment and financing. It also cautions that electricity-price effects depend on local conditions and policy choices. It would be misleading to claim that every new data centre automatically increases every household’s bill.
Growth opportunity, uncertain returns
The IMF’s October 2026 World Economic Outlook identifies stronger AI investment and productivity as possible sources of upside. It also treats a reassessment of AI expectations as a downside risk.
These are scenarios, not a declaration that a crash is certain or that every investment will succeed. The report’s estimates use information available through 1 October 2026.
Four questions worth following
- Is a project announced, financed, under construction or operating?
- Does it have an electricity connection and a credible delivery timetable?
- Is there evidence of demand for its services?
- Are claimed savings or productivity improvements measured, or still promised?
This checklist helps separate milestones that sound similar in a headline. A launch, a forecast and a measured result answer different questions.
From plans to evidence
AI infrastructure: four separate checks
- 01 / Input
Announcements
Spending plans and stated ambitions
- 02 / Milestone
Financing and grid connection
Funding secured; electricity access assessed
- 03 / Milestone
Operating capacity
Infrastructure delivered and running
- 04 / Result
Measured demand and productivity
Service use and improvements supported by evidence
For Farvenzo, the story to watch is whether physical infrastructure and useful applications develop together. Follow the evidence beyond the announcement, and keep projections labelled as projections.