The idea that moving to the cloud saves money has been repeated so consistently, by so many vendors, analysts, and technology consultants, that it has acquired the status of received wisdom in enterprise technology circles. Boards are briefed on cloud migration as a cost reduction strategy. CFOs are presented with business cases showing five-year TCO savings that justify substantial migration investments. And procurement teams are told that the flexible, consumption-based nature of cloud means the organisation will stop paying for infrastructure it does not need.
In 2026, with enough real-world cloud migration experience accumulated across enough enterprise organisations, it is worth examining this claim directly. Not to argue that cloud has no commercial merit, because it genuinely does for specific workload categories and specific organisational contexts. But to separate the cases where cloud genuinely reduces cost from the cases where it does not, and to be honest about the conditions under which the cost savings narrative reflects commercial reality and the conditions under which it does not.
The honest answer is that cloud does not automatically save money. It can save money under specific conditions. It can also cost significantly more than on-premises, under different conditions that are more common in practice than the vendor narratives acknowledge. Understanding which conditions you are in is the most commercially important thing an enterprise organisation can know about its cloud strategy.
Where Cloud Genuinely Does Reduce Cost
Cloud reduces cost most reliably in three categories. The first is variable, unpredictable workloads. Applications where resource demand fluctuates significantly based on business activity, such as batch processing environments, seasonal e-commerce platforms, and event-driven analytics workloads, benefit genuinely from cloud’s ability to provision capacity on demand and release it when not needed. In on-premises environments, these workloads require infrastructure sized for peak demand, which sits underutilised during the long periods of average or below-average load. Cloud allows the organisation to pay for peak capacity only when it is genuinely needed, and that match between resource consumption and resource payment is commercially real.
The second category is net new capability. Where cloud enables the organisation to access services that would require significant capital investment to replicate on-premises, such as managed AI services, global content delivery, or specialised analytics platforms, cloud consumption is genuinely cost-effective compared to the alternative of building the equivalent capability from scratch.
The third category is infrastructure elimination. Where cloud migration is combined with genuine on-premises infrastructure decommissioning, including physical server retirement, data centre space reduction, and the elimination of the staff and operational overhead that maintained the decommissioned infrastructure, the cloud cost can be lower than the total on-premises cost it replaces. The emphasis here is on genuine decommissioning. Organisations that migrate workloads to cloud while maintaining the on-premises infrastructure in a standby or partially active state do not realise the infrastructure elimination saving.
Forrester Research has tracked enterprise cloud migration financial outcomes across hundreds of organisations and has been among the most rigorous analysts in separating the conditions under which cloud produces cost savings from those under which it does not. Their Forrester enterprise cloud financial outcome and cost research provide the independent evidence base that distinguishes genuine cloud cost reduction from the vendor narrative, covering the workload types, organisational conditions, and governance practices that correlate with positive financial cloud migration outcomes.
Where Cloud Consistently Costs More Than Expected
The cases where cloud costs more than expected are, in 2026, well documented and consistently observed across enterprise organisations at scale.
Stable, predictable workloads are more expensive in cloud than on-premises in most configurations when fully costed. A database that runs at consistent load twenty-four hours a day, seven days a week, year-round, is a workload where cloud’s consumption-based pricing model offers no advantage and typically represents a significant premium over equivalent on-premises infrastructure that has been fully depreciated. The on-premises infrastructure cost for a stable workload is dominated by hardware depreciation, data centre operational cost, and staff time, all of which are fully visible and accounted for in the on-premises model. Cloud pricing for the equivalent service, billed continuously at consumption rates, often costs more over a three to five year horizon once all components are included.
The lift and shift migration is the most common source of cloud cost disappointment. When an organisation moves an existing on-premises workload to cloud without redesigning it for cloud-native consumption patterns, it replicates the resource consumption profile of the on-premises application in a cloud billing model where that consumption pattern is expensive. A legacy application designed for always-on, dedicated resources does not automatically become a variable consumption workload just because it moves to cloud. It continues to consume resources at the same rate, but now pays cloud pricing for the privilege.
Data egress, storage costs, and the licensing premium of cloud-delivered software all add to the cost picture in ways that are frequently absent from cloud migration business cases. As noted elsewhere, data egress from cloud environments is not free. Storage at cloud pricing for large data volumes carried from on-premises can be significantly more expensive than the equivalent on-premises storage cost. And the licence-included pricing of managed cloud database and application services embeds software licensing cost at rates that may exceed the BYOL equivalent for organisations with qualifying on-premises licence assets.
Accenture’s cloud economics research addresses the total cost of cloud migration and operation across enterprise workload categories, including detailed analysis of where cloud TCO exceeds on-premises TCO and the workload characteristics that most reliably predict each outcome. Their Accenture cloud economics and enterprise TCO research provide evidence-based frameworks for building cloud migration business cases that account for the full cost picture rather than the partial view that highlights cloud savings while underweighting cloud costs.
The Governance Gap That Makes Cloud More Expensive
Even where the technical case for cloud cost reduction is genuine, organisations frequently fail to realise the expected savings because the governance practices required to manage cloud costs effectively are not in place at the time of migration. Self-service cloud provisioning creates infrastructure without the approval controls that gate on-premises procurement. Development environments run continuously rather than being shut down outside working hours. Reserved Instance commitments are made on optimistic utilisation assumptions and are not actively managed as those assumptions prove incorrect. And the cost visibility that allows on-premises infrastructure costs to be understood and challenged does not automatically transfer to cloud, where costs are distributed across services, regions, and accounts in ways that require active FinOps capability to monitor and manage.
The FinOps discipline, which treats cloud as a financial management challenge as much as a technology management challenge, addresses this governance gap directly. Organisations with mature FinOps practices consistently achieve better financial outcomes from their cloud investments than those without them, because they have the visibility, accountability, and optimisation processes that prevent the cloud governance gap from accumulating waste. But FinOps maturity takes time to build, and most organisations begin their cloud migration journey before their FinOps capability is ready to govern it effectively.
The New Stack covers enterprise cloud-native architecture and the operational and financial governance practices that determine whether cloud migration programmes deliver their expected financial outcomes. Their The New Stack cloud-native operations and FinOps governance coverage address the governance and operational discipline requirements of managing cloud cost effectively, including the specific FinOps practices that close the visibility and accountability gap that consistently drives cloud costs above projection.
The Vendor Narrative vs the Customer Reality
It is worth being direct about the structural reason why the cloud saves money narrative has been so persistent despite the evidence that contradicts it in many circumstances. Every major cloud vendor, Microsoft, Oracle, SAP, IBM, and the hyperscalers, has a commercial interest in cloud migration. Cloud consumption generates recurring revenue at margins that perpetual licence sales and on-premises software support do not. The business case tools that these vendors provide for evaluating cloud migration, the TCO calculators, the migration assessment tools, and the financial modelling templates, are built to present cloud in its most favourable light. They are not designed to surface the scenarios where cloud costs more.
This does not mean cloud is a bad decision. It means the decision should be made on independently modelled analysis rather than vendor-provided financial tools. The independent TCO model should include full parallel running costs, licence interaction impacts, realistic utilisation assumptions for committed cloud capacity, egress and storage costs at realistic data volumes, and the governance investment needed to manage cloud costs effectively. When the analysis includes all of these elements, the decision to migrate specific workloads to cloud should be grounded in genuine financial evidence rather than a narrative.
PwC’s cloud economics and technology advisory practice has published extensive research on the financial realities of enterprise cloud migration, including the conditions under which cloud delivers genuine cost savings and those under which it does not. Their PwC cloud economics and enterprise migration financial outcomes research provide independently developed frameworks for cloud migration financial evaluation that account for the full cost picture, helping organisations build migration business cases that hold up to scrutiny rather than collapsing when the additional costs become visible during execution.
Conclusion
Moving to the cloud does not automatically save money. It can save money under the right conditions, with the right workloads, with genuine infrastructure decommissioning, and with the FinOps governance in place to manage cloud consumption effectively. Under different conditions, which describe a large proportion of real-world enterprise cloud migrations, cloud costs more than the equivalent on-premises deployment, sometimes significantly so. The organisations that make cloud decisions based on genuine, independently modelled financial analysis will achieve better outcomes than those that adopt the narrative. The technology is not the problem. The business case methodology is.
