AI is sucking up so much electricity that it’s literally making it too expensive for physical factories to stay in business.
Data centers for AI models are so profitable they can pay massive premiums for power, driving up capacity prices. This 'thermodynamic crowding out' is making it too expensive for traditional industrial plants to stay on the grid.
Thermodynamic Crowding Out: Empirical Evidence of Energy Cannibalisation in US Capacity Markets (2024-2026)
SSRN · 6263699
<p><span>The rapid expansion of Artificial Intelligence (AI) and hyperscale data centre infrastructure is precipitating a novel form of resource scarcity in United States electricity markets. This paper analyses load forecast data from the PJM Interconnection (2024–2026) to test the hypothesis of "Thermodynamic Crowding Out" – the structural displacement of physical industrial capacity by virtual computational load within a physically constrained grid. We identify a statistically significant div