AI's next bottleneck isn't just chips — it's America's power grid, Goldman Sachs says

A recent analysis by Goldman Sachs reveals that while the United States leads the world in artificial intelligence infrastructure, it faces a critical challenge beyond just hardware and talent: its power grid. As the demand for electricity surges due to the increasing number of data centers supporting AI models, the current US power grid is becoming strained. Presently, these data centers account for approximately 6% of the nation's total electricity consumption, a figure that is projected to nearly double to 11% by 2030. Goldman Sachs analysts warn that this escalation could push portions of the grid beyond critical limits. They emphasize that a stable and sufficient power supply will be essential in determining the future of the AI race, particularly since resolving power infrastructure bottlenecks can be a protracted process. While the US holds a significant portion of the global AI infrastructure, with 44% of the world's data center capacity—equivalent to the combined total of China, the EU, Japan, Korea, and India—the tightening conditions in the US power market are alarming. Over the past five years, the peak summer spare power generation capacity has dropped from 26% to just 19%, driven by the rising demand from data centers. If this trend continues, analysts predict that by the end of the decade, spare capacity may fall below the critical threshold of 15%. In stark contrast, China is reportedly preparing for the future by amassing energy resources. Goldman Sachs anticipates that by 2030, China's spare power capacity could reach 400 gigawatts, more than three times the expected global demand from data centers. This strategic buildup follows an energy crisis in 2021, prompting Beijing to enhance its power generation capabilities across various sources, including renewables, natural gas, nuclear, and coal. The US, meanwhile, is moving away from coal at a faster rate than it can replace it with new natural gas or renewable energy sources. Other challenges hindering US data center expansion include prolonged project timelines and a global shortage of gas-powered turbines. As the race for AI technology intensifies, Goldman Sachs highlights that a dependable power supply will be critical to sustaining America's lead, especially given the slow pace at which power infrastructure issues can be resolved. Recently, Nvidia CEO Jensen Huang echoed these concerns, stating that energy could become a significant barrier to the US's AI growth and warning against excessive regulatory measures that could stifle innovation. In contrast, he noted that China's energy subsidies make it easier for local companies to power their AI developments, with energy costs being significantly lower.