Goldman Sachs warns of AI revenue needs and power grid risks

Goldman Sachs analysts have issued warnings regarding the economic and infrastructural challenges facing the artificial intelligence sector. To reach a break-even point on massive capital expenditures, hyperscalers such as Amazon, Microsoft, and Oracle may need to generate approximately $300 billion in AI-related revenue within the next few years.

Beyond financial requirements, the physical infrastructure poses significant risks. Analysts warn that next-generation AI server racks, which could consume 50 to 60 times more power than older designs, may trigger localized blackouts in the United States. US data center power demand is projected to reach roughly 108 GW by 2030.

The risk is intensified by geographic concentration, as 72% of US data center capacity is located in just 1% of counties. If capacity additions do not accelerate, reserve margins could fall below the 15% threshold, leaving grids vulnerable to outages during peak demand. Challenges to expanding energy infrastructure include long interconnection backlogs and shortages of specialized equipment like transformers.