OpenAI is publicly warning that the U.S. electrical grid cannot support the pace of AI development. The company estimates the nation needs approximately 100 gigawatts of new generating capacity annually — a buildout rate not seen since post-war electrification.

Key topics covered:

• Data center power consumption projected to reach 123 GW by 2035 (up from 4 GW in 2024)
• Infrastructure bottlenecks: transformers, interconnection queues, transmission permitting
• Behind-the-meter solutions and on-site generation strategies
• Stargate, xAI Memphis, and Three Mile Island restart
• Grid-dependent vs. grid-independent development risk
• The premium for shovel-ready sites with secured power

Related keywords: grid capacity, AI power demand, behind-the-meter generation, interconnection queue, data center power, OpenAI infrastructure, Stargate project, utility constraints

The Power Allocation is brought to you by Spring Street Management Group — translating AI and data center hype into real infrastructure and assets.