AI Data‑Center Delays Loom
- Social reporting flagged transformer shortages that are delaying AI data‑center builds and capacity expansions. - The post cited roughly $650 billion in commitments and warned of supply bottlenecks tied to a small number of suppliers. - The thread suggested transformer and critical‑component shortages are slowing deployment timelines for large AI projects. (x.com)
The bottleneck in the artificial-intelligence building boom is no longer just chips: utilities and developers are waiting on transformers, switchgear and other grid hardware. (bloomberg.com) A transformer is the heavy electrical equipment that steps power up or down so electricity can move from the grid into a data center safely. Large-transformer shortages are pushing up prices and delaying projects, the Financial Times reported in late 2025. (ft.com) The backlog is measured in years, not months. Utility Dive, citing the National Association of Electrical Manufacturers, reported that a new transformer ordered in late 2024 could take up to three years to arrive, versus four to six weeks five years earlier. (utilitydive.com) Those delays are colliding with a wave of artificial-intelligence infrastructure spending. OpenAI said on January 21, 2025 that Stargate planned to invest $500 billion over four years in U.S. artificial-intelligence infrastructure, and BlackRock’s AI Infrastructure Partnership said on March 19, 2025 that it aimed to mobilize up to $100 billion, including debt financing. (openai.com) (blackrock.com) The buildout already shows signs of slowing. Bloomberg reported on February 25, 2026 that U.S. data-center capacity under construction fell to 5.99 gigawatts at the end of 2025 from 6.35 gigawatts a year earlier, citing CBRE, even as demand for artificial-intelligence capacity kept rising. (bloomberg.com) CBRE’s own North America report said power availability and infrastructure delivery timelines were the decisive factors shaping site selection, leasing activity and pricing in major U.S. markets in the first half of 2025. The firm said 74.3% of under-construction capacity was already preleased. (cbre.com) Manufacturers say they are adding capacity, but not fast enough to erase the queue. Hitachi Energy said in March 2025 it would invest more than $250 million by 2027 to expand production of transformer components, after announcing more than $1.5 billion of transformer-capacity investment in April 2024. (hitachienergy.com) (markets.ft.com) Even component lead times remain long. Hitachi Energy’s delivery table updated April 21, 2026 listed some bushings, the insulated parts that let current pass through a transformer, at 85 weeks. (hitachienergy.com) The squeeze extends beyond one company or one part. Bloomberg reported on April 1, 2026 that U.S. artificial-intelligence data-center expansion has leaned on imported transformers, switchgear and batteries because domestic manufacturing has struggled to keep up. (bloomberg.com) Power demand is rising at the same time. The International Energy Agency said on April 16, 2026 that capital spending by five large technology companies topped $400 billion in 2025 and was set to rise another 75% in 2026, driven by data-center investment. (iea.org) So the race to build more artificial-intelligence capacity now runs through factories that make some of the grid’s least glamorous equipment. Until more transformers and related components arrive, money committed to new campuses will not automatically turn into megawatts in service. (bloomberg.com)
