Gartner: 70% of 2026 mainframe revamps will fail

Gartner's forecast, published in June 2026, predicted that 70% of mainframe modernization projects started in 2026 will fail, and it pointed to overconfidence in generative AI as a factor. That prediction set the tone for the most-read system modernization articles of the first half of fiscal 2026 (April–September), which also saw Hitachi announce the end of sales for its VOS3 mainframe OS and a steady stream of news about the end of COBOL-era mainframe support. The articles suggest that the appeal of generative AI as a shortcut for legacy code conversion has run into a gap between vendor claims and real-world results. Behind this gap: mainframes run mission-critical operations, so migration failures have an outsized impact, and engineers familiar with mainframes are increasingly retiring. The body of coverage points to a recurring theme—that what humans must retain is not programming syntax but understanding of why a system was designed the way it was. The outcome hinges on whether AI tooling can meaningfully reduce the effort of converting legacy code, and on how organizations preserve the business-process knowledge that retiring engineers hold.