← Depreciating CollateralEren Kahraman · November 2026

Figure 1.1. Data centre investment needs, 2025-2030

Lower scenario (AI, low end of range), USD trillion (cumulative, 2025-2030): ≈ 3.0. Central scenario (continued momentum), USD trillion (cumulative, 2025-2030): 6.7 (of which USD 5.2 trillion for ai workloads). Upper scenario (accelerated demand, AI), USD trillion (cumulative, 2025-2030): 7.9.

0123456789USD trillion (cumulative, 2025-2030)AI workloadsTraditional IT workloadsabout USD 3 trillion to USD 8 trillion≈ 3.05.26.77.9USD 3.1 trillionLower scenario(AI, low end of range)Central scenario(continued momentum)Upper scenario(accelerated demand, AI)
AI workloadsTraditional IT workloadsabout USD 3 trillion to USD 8 trillionLower scenario (AI, low end of range)≈ 3.0Central scenario (continued momentum)5.26.7USD 3.1 trillionUpper scenario (accelerated demand, AI)7.90123456789USD trillion (cumulative, 2025-2030)

Note: The lower and upper scenarios cover AI workloads only. The value for the lower scenario is the approximate low end of the range reported by McKinsey.

Source: McKinsey (2025), The cost of compute: A $7 trillion race to scale data centers.

View data table
Figure 1.1. Data centre investment needs, 2025-2030: data
ScenarioUSD trillion (cumulative, 2025-2030)
Lower scenario (AI, low end of range)≈ 3.0
Central scenario (continued momentum)6.7 (of which USD 5.2 trillion for ai workloads)
Upper scenario (accelerated demand, AI)7.9
Open standalone

From Depreciating Collateral: Technology risk in AI infrastructure debt and who holds it, Eren Kahraman, November 2026. Data as of 7 November 2026. Download the paper (PDF)