IBM’s problem is not that the mainframe disappears tomorrow. It is that most new workloads are being built elsewhere.
AWS, Azure, Google Cloud and Linux are now the defaults for new applications, while IBM Z, AIX and IBM i increasingly depend on existing customers, transaction growth and hardware-refresh cycles.
AI is weakening the mainframe’s strongest defense: the difficulty of understanding decades of undocumented COBOL and business logic. Toyota’s use of Amazon Q to analyze legacy COBOL and support migration planning is a warning. AI does not need to replace the mainframe directly; it only needs to make leaving cheaper, faster and less risky.
IBM’s traditional software is also exposed. DataStage faces cloud-native ELT, zero-ETL, Redshift, S3 and Iceberg. Red Hat remains useful, but many customers can choose EKS, AKS or GKE instead of OpenShift—and Amazon Linux instead of RHEL.
IBM Cloud is not a credible fourth hyperscaler, and IBM is not investing in AI infrastructure at anything close to Amazon, Microsoft or Google. Its AI strategy increasingly depends on competitors owning the compute, models, distribution and economics.
Quantum offers no immediate rescue. IBM is a serious contender, not the undisputed leader, and commercial returns remain years away.
IBM has repeatedly acquired companies—Lotus, Informix, Rational and many others—without consistently creating lasting platform leadership. Continuing to buy software companies while distributing billions through dividends does not address the underlying problem.
IBM should make an honest choice: cut the dividend and reinvest aggressively, or break up and sell its valuable businesses while they still command premium value.
Harvesting legacy cash, buying another company and protecting the dividend is not a growth strategy. It is managed decline.