AI Datacenters Face Execution Risks in 75GW Behind-the-Meter Power Buildout

However, the rapid scaling of this infrastructure introduces significant execution risks. The report highlights challenges across six key areas: contract bankability, permitting delays, fuel supply, equipment procurement, labor shortages, and the electrical physics of operating without grid support. High-profile projects like Oracle’s Project Jupiter have faced delays, forcing emergency pivots to alternative power sources. While the number of engine manufacturers has grown from 12 to 22, the industry is still navigating uncharted territory in building GW-scale islanded datacenters.

The distinction between various BTM configurations—such as grid-parallel, export-only, and fully off-grid—is critical for regulatory and operational planning. As these projects move from announcement to operation, the focus shifts to managing the complex logistics of financing, permitting, and maintaining reliability in isolated power systems. The report suggests that while BTM is now mainstream, the path to successful delivery remains fraught with technical and financial hurdles that could impact timelines for major AI infrastructure rollouts.