Large and GSU Transformers

Think of electricity leaving a generator the way water leaves a garden hose at low pressure, perfectly usable nearby, but destined to lose most of its force if pushed over any real distance. A transformer is the pump that raises electricity's voltage before it travels, and a Generator Step-Up transformer, GSU, is the specific transformer sitting right beside a generator or power plant, whose entire job is stepping voltage up before that power reaches the wider grid or a data centre's own internal distribution network.

Large power transformers and GSUs have quietly become the single most cited reason new data centre capacity is slipping to later years than originally planned: lead times that ran a comfortable 24 to 30 months before the pandemic have stretched to three to five years for large power transformers, with GSU units specifically averaging around 144 weeks according to a mid-2025 industry survey, and demand for GSUs has grown roughly 274% since 2019, driven by data centres and grid modernisation hitting simultaneously.

India's transformer manufacturers have responded by expanding aggressively rather than simply raising prices, CG Power has taken its manufacturing capacity from roughly 17,000 MVA to around 65,000 MVA and won a transformer export order worth roughly ₹900 crore for United States data centre projects in 2026, while Hitachi Energy India reported orders up 73.7% year on year in the quarter ended December 2025, led by transformers, reactors and switchgear, covered elsewhere on this site, with data centres explicitly named as a major contributor, and is doubling capacity at its Vadodara plant through a ₹2,000 crore investment.

The structural causes behind the shortage go beyond simple demand, years of underinvestment in domestic transformer manufacturing worldwide, a sudden post-pandemic surge in construction and electrification, and volatility in specialised inputs like grain-oriented electrical steel have all compounded at once, which is part of why more than half of the United States data centre capacity originally targeted for 2026 has already faced delays pushing completion into 2027 or later, a pattern the industry now treats as the default rather than the exception whenever a new campus is announced, connecting directly to the broader power demand problem behind every AI data centre, covered elsewhere on this site.