AI-powered mainframe exits are a bubble set to pop: Gartner

The verdict

Analyst firm Gartner has delivered a cold splash of water on the AI-for-mainframe-exit party. It has been reported that Gartner’s paper warns “more than 70 percent of mainframe exit projects initiated in 2026 will fail to produce the intended benefits due to an overestimation of generative AI tooling capabilities.” Ouch. The firm also projects that “by 2030, 75 percent of vendors operating in the ‘mainframe exit’ market will either pivot their business models or cease to exist,” it has been reported that the analysts say.

Why they’re sceptical

The reasoning is blunt: mainframes house mission‑critical apps and decades of interlinked data. Generative AI, useful for spotting and describing technical debt, allegedly trips up when asked to convert and migrate that tangled code while preserving performance and throughput. Gartner flags a dangerous gap between marketing claims and reality — the kind of gap that can threaten business continuity, not just budgets.

Who wins and who loses

Gartner also blames aggressive investor demand that pushes vendors to slap “AI” on every silver bullet; the firm alleges this is forcing vendors into ill‑fitting AI bets. Is this just another AI gold rush where hype outpaces engineering? Possibly. The note lands at an awkward time: it has been reported that IBM’s shares slid after rival vendors hyped COBOL-conversion demos, even as Big Blue’s mainframe revenues remain robust. In short: expect winners, losers, and a lot of pivots.

What to do instead

Gartner’s takeaway is practical if unspectacular: don’t believe the marketing promise. Most customers should focus on modernizing and optimizing their iron in place, evaluate workloads platform‑by‑platform, and treat migration as a high‑stakes engineering project, not a quick AI trick. Want drama or a disaster? Choose the shiny demo. Want continuity? Take the slower, smarter road.

Sources: The Register