Cloud Strategy Evolves From 'Repatriation' to 'Replatforming'
The corporate cloud narrative is shifting from simple migration or cost-driven "repatriation" to a more mature focus on architectural efficiency, a recent analysis suggests. Companies are now focused on "replatforming" applications to best suit business needs, AI workloads, and risk profiles, whether on-premise or in a public cloud. This reflects an evolution toward tailored, hybrid strategies rather than a one-size-fits-all cloud adoption model.
- Replatforming, also known as "lift, tinker, and shift," strikes a middle ground between a simple "lift and shift" (rehosting) and a complete rewrite (refactoring). It involves moving an application to the cloud while making some modifications to take better advantage of cloud-native features. - A primary driver for replatforming is the need to optimize for AI and machine learning workloads, which often require specific hardware like GPUs and improved data access that legacy, on-premise architectures can't efficiently provide. Gartner predicts that by 2029, 50% of cloud compute resources will be dedicated to AI/ML, a significant increase from less than 10% today. - The trend away from simple repatriation is driven by high costs and operational risks. Repatriation can incur significant expenses, such as data egress fees, which for 1 petabyte of data from AWS S3 can range from $90,000 to $120,000, in addition to the cost of running parallel environments during the transition. - Replatforming supports the move towards multi-cloud and hybrid cloud strategies, which have become the norm for large enterprises. As of 2025, over 90% of large enterprises operate in a multi-cloud environment, seeking to avoid vendor lock-in and leverage the best services from different providers. - The shift to replatforming reflects a broader maturation of cloud strategy, moving beyond a "cloud-first" mandate to a more nuanced evaluation of where workloads run most effectively. This includes analyzing factors like cost predictability, data gravity, and regulatory compliance before deciding on placement. - The decision to replatform is often part of a larger effort to address technical debt. Up to 29% of an organization's resources can be spent on maintaining legacy systems, and replatforming allows for updating outdated components without a complete overhaul. - For steady-state, predictable workloads, repatriation can still offer a lower total cost of ownership compared to the public cloud's consumption-based pricing. However, replatforming offers a path to modernize these applications to be more efficient, whether they ultimately reside on-premise or in the cloud. - The focus on replatforming is also influencing hardware and software design, with CPUs evolving to better handle AI workloads and infrastructure software becoming a key investment area to bridge the gap between hardware and applications.
