A single transmission line fault in Ashburn, Virginia—home to the world's largest data center cluster—took more than 3 gigawatts of load off the grid in seconds on July 22, 2026. The event was not unprecedented. Two years earlier, a failed surge arrester caused a drop of roughly 60 megawatts, underscoring a pattern of fragility.
The MIT Technology Review analysis frames these outages not as random equipment failures but as symptoms of an architecture problem. The electric grid was designed for distributed, predictable demand, while AI data centers concentrate enormous loads in specific regions and require near-constant uptime.
The piece argues that incremental fixes like adding backup power or upgrading individual substations will not be enough. What is needed is a rethinking of how power is transmitted, stored, and allocated around AI infrastructure—otherwise, the industry will keep building capacity that the grid cannot reliably feed.