Why the bot decided this
Microsoft walked away from a U.K. AI data-center project, while SemiAnalysis forecasts 50 GW of new AI data-center capacity in 2027.
The article
Microsoft Walks Away From AI Data Center Deal, but SemiAnalysis Predicts 50-GW Boom in 2027 More than 24 gigawatts of data center capacity is already in advanced construction across just 40 campuses, most of it expected to come online in 2027, according to new research from SemiAnalysis. The research firm forecasts 50 gigawatts of new data center capacity outside China in 2027, roughly as much as the industry had built in those markets over its entire history up to the end of 2024. The bullish outlook comes as Microsoft Corp. (NASDAQ:MSFT) walked away from a U.K. data center agreement because of electricity delays. The 50-megawatt facility was expected to house approximately 23,000 Nvidia Corp. (NASDAQ:NVDA) GB300 chips. Why SemiAnalysis Expects the Boom to Continue Most of the 40 campuses are more than halfway completed, with cooling and electrical equipment already installed. The researchers wrote that “gigantic sites are simply getting built, and many are on time” in Monday’s report. Their confidence rests on a surge in long-term leases. Developers typically spend only 5% to 10% of a project’s budget before securing a customer, using signed contracts to raise financing. SemiAnalysis estimates roughly $1 trillion in multiyear data center leases have been signed this year, helping fund construction scheduled for 2027 and 2028. The firm also argues that “the power generation problem has broadly been solved.” Rather than waiting five to 10 years for grid connections, developers are increasingly installing their own electricity generation, including fuel cells, refurbished turbines and engines originally designed for trucks. These alternatives could bypass the kind of grid delays that derailed Microsoft’s Essex deal, although they may not be practical at every location. Construction is accelerating, too. SemiAnalysis says Singapore-based DayOne has cut building timelines from more than 18 months to around nine. The firm estimates that modular construction can reduce certain on-site labor requirements by up to 63%. Could Chips Become the Next Bottleneck? Not everyone shares SemiAnalysis’s optimism. Altimeter Capital founder Brad Gerstner challenged an earlier 43-gigawatt projection from SemiAnalysis founder Dylan Patel on the All-In podcast. “I would suggest Dylan’s forecast of 43 gigawatts next year is too aggressive,” Gerstner said, estimating actual delivery closer to 25 gigawatts. His comments focused on U.S. infrastructure, while SemiAnalysis’s latest forecast covers a broader market. Meanwhile, traders on Polymarket put roughly 9% odds on an AI industry downturn by Dec. 31. The contract requires three qualifying events within 90 days, potentially including a 50% drop in Nvidia shares from their all-time high. SemiAnalysis expects memory and processor shortages to become a bigger constraint in 2027 and 2028. For Nvidia, that could mean demand for its chips continues to outstrip supply even as developers overcome the construction and electricity problems holding up projects today. Image: Shutterstock Read Also: Can Meta’s Fiber Partner Corning Capitalize On Rising AI Fiber Prices?