The AI Paradox: Legacy Mainframes Gain Strategic Value as Data Moats

The rush to integrate artificial intelligence into the corporate stack is creating an unexpected lifeline for legacy hardware. While the industry long viewed the mainframe as a relic, the reality of AI deployment is forcing a rethink of the modernization playbook.

As first reported by The Register, enterprises are becoming increasingly selective about replacing legacy kit. Rather than pursuing total rip-and-replace strategies, many organizations are opting to "sweat" their existing assets. This trend is evidenced by comments from the HPE managing director for the UK, Middle East, and Africa, who noted that some customers are extending refresh cycles from five years to seven years.

This shift is driven by the realization that legacy systems are critical repositories of intellectual property. Brian Klingbeil, Chief Strategy Officer at Ensono, notes that these systems house decades of business logic and data. According to Ensono’s 2026 State of IT Modernization report—a survey of 500 IT decision-makers in the US and UK—78 percent of respondents now view legacy systems as more important than they did two years ago, specifically because they are key to making AI functional.

The report indicates that nearly half of organizations view mainframes as a critical foundation and data source for AI, while 47 percent use them selectively for specific applications. Consequently, more than half of surveyed organizations are choosing to optimize these systems while modernizing applications in place—a preference more pronounced in the UK (57 percent) than in the US (48 percent).

However, integration is not without friction. Ensono reports that 33 percent of respondents cite the difficulty of integrating AI into existing workflows as a top challenge, while 28 percent point to infrastructure limitations.

The mainframe's evolution into a strategic asset has been noted by others; Kyndryl observed two years ago that "big iron" was a prime candidate for AI workloads, and Gartner stated earlier this year that migrating workloads to a mainframe was more logical for VMware users than adopting new licenses from Broadcom. As Ensono reports that 45 percent of firms are scaling AI deployments, the most advanced future-state technologies are becoming increasingly dependent on the most antiquated hardware.