AI and digital platforms: Chokepoints and risks
Model builders and apps selling AI as assistants, APIs and copilots, not hardware
Chokepoints and risks
- Chip concentration: Nvidia holds an estimated 80-90% of AI accelerator revenue and TSMC in Taiwan makes most leading-edge accelerators, while export controls limit who can buy them5.
- Power: data centres used about 415 TWh (1.5% of world electricity) in 2024 and about 17% more in 2025, and the IEA expects use to roughly double by 2030 (forecast), with gas turbines, transformers and grid connections already bottlenecks3,6.
- Capital intensity: infrastructure was about 55% of 2025 AI spending (calculated), so returns depend on sustained revenue growth1.
- Geographic concentration: the US took US$ 285.9 billion of 2025 AI investment, against US$ 20.9 billion in Europe and US$ 12.4 billion in China2.
- Copyright litigation and new AI regulation could restrict training data or raise compliance costs.