AI infrastructure shifts from chip shortages to power constraints

TrendForce has reported that the next major bottleneck in AI development is shifting from a shortage of chips to a shortage of electricity and grid infrastructure. The firm predicts that AI servers will account for 33.4% of data center power demand in 2026, rising to over 40% by 2027. This indicates a growing need for cloud infrastructure to be restructured around AI. TrendForce estimates that global data center power demand will reach 161 GW in 2026, a 31% increase from the previous year. However, grid infrastructure and power distribution may not keep pace, potentially delaying new data center projects. Amin Vahdat, Google's Chief Technology Officer, noted that the core bottleneck for AI expansion has shifted from chip shortages to power supply, with the ceiling for AI computing power now being defined by the availability of electricity. TrendForce also forecasts that by 2030, data center demand could reach 490.7 GW, while the grid's capacity to supply power to data centers would only reach 222.6 GW, creating a significant gap of 268 GW. This highlights the growing importance of power infrastructure in the expansion of AI technologies.