Grid Delays and Energy Bottlenecks Threaten the Rapid Expansion of Artificial Intelligence
The explosive growth of artificial intelligence is colliding with the slow pace of building electrical infrastructure, as energy bottlenecks threaten to stall the ongoing technological race. A new report from financial institution Goldman Sachs warns that power limitations represent the single greatest risk to trillions of dollars in projected spending, as physical grid connections fail to keep pace with the rapid speed of software development. Global data center power demand is projected to surge by 170% between 2025 and 2030, with the United States accounting for more than 60% of that growth. Jason Tofsky, Goldman’s global head of digital infrastructure investment banking, characterized the situation bluntly in the report, noting that power is likely to remain a permanent constraint on the sector. To support the burgeoning demand across computing, data centers, and power generation, the bank estimates approximately $7.6 trillion in capital spending will occur between 2026 and 2031.
The physical infrastructure required to sustain modern artificial intelligence operations moves significantly slower than the technology itself. Across the United States, approximately 2,600 gigawatts of proposed power generation and storage projects are currently awaiting connection to the electrical grid. This backlog is more than double the country’s existing installed capacity, though industry analysts note that a substantial portion of these proposed projects may ultimately fail to materialize. The median waiting time from an initial application for a grid connection to achieving commercial operation stands at roughly five years. This regulatory and logistical lag means the race for artificial intelligence dominance is no longer defined solely by advances in frontier models, but rather by which companies can successfully secure adequate electricity.
This structural mismatch introduces considerable financial risk, particularly because the massive infrastructure expansion is heavily financed through debt. Goldman projects that global hyperscaler debt issuance will reach $420 billion in 2027, marking an increase of more than 60% compared to estimates for 2026. Debt is expected to finance roughly 35% of hyperscaler capital expenditures next year. Furthermore, hyperscalers account for only 40% of overall AI ecosystem credit fundraising, demonstrating that borrowing extends widely across the broader supply chain and supporting industries. If construction timelines slip due to grid delays, companies investing heavily before securing power face extended timelines for financial returns and heightened credit risks.
Local opposition and permitting hurdles are compounding these physical delays. Since January, 86 projects have experienced significant delays or cancellations, with 76 of those directly attributed to local resistance, according to the report. Despite the friction, some market observers suggest that slower deployment could provide a beneficial side effect by preventing over-investment and allowing actual consumer demand time to catch up with capital spending. Meanwhile, the power boom is fundamentally transforming traditional utilities from slow-growing enterprises into primary beneficiaries of artificial intelligence investment. Utility capital plans are projected to reach $1.3 trillion between 2026 and 2030. To circumvent traditional utility bottlenecks, major technology companies are increasingly exploring arrangements to directly fund the power generation, transmission lines, and electrical infrastructure necessary to run their data centers.
