McKinsey highlights that many companies carry an on-premises mindset into the cloud, which leads to six persistent mistakes that limit value and increase costs. Key issues include: 1. **Treating cloud as a simple “lift and shift”** Many organizations move applications to the cloud quickly to capture immediate savings in hosting, storage, and maintenance. While this can deliver short-term gains, they often keep the same technical and operational inefficiencies they had on-premises. As a result, they miss out on the cloud’s flexible infrastructure and advanced capabilities. McKinsey notes that “Year One” benefits (like speed to market and access to new capabilities) can exceed “Day One” benefits by **15% to 25%**, but only if companies plan beyond the initial migration. 2. **Clinging to a capital expenditure (CapEx) mindset** On-premises IT is typically treated as a capital investment. In the cloud, spending becomes operational expenditure (OpEx), where you pay for what you consume. Some companies fail to adopt a dynamic OpEx approach and don’t precisely measure demand. McKinsey stresses that efficient cloud economics now depend on being able to evaluate capacity demand and marginal costs at any moment—paying for capacity only when you need it, instead of paying for unused capacity. 3. **Forecasting cloud spend based on history, not priorities** When shifting to OpEx, many organizations still rely on historical patterns to budget for cloud, even though past usage is a weaker predictor in a highly elastic environment. McKinsey notes that estimates can miss actual spending by **more than 20%**. Instead, companies should tie cloud budgets to business priorities—such as a major Black Friday promotion or a new subscription model—because these initiatives directly affect cloud usage and cost. 4. **Not differentiating workloads by demand patterns** Cloud is especially valuable for workloads with highly variable consumption. McKinsey cites a video-streaming company that measured cloud costs per subscriber and aligned compute needs with demand patterns, achieving **over 95% accuracy** in predicting cloud consumption. Many companies miss similar savings because they don’t distinguish between workloads with short-term demand spikes and those with stable usage, like long-term subscriber data storage. 5. **Weak linkage between cloud architecture and cloud economics** Some organizations overestimate both their cloud usage and the value they’ll achieve because they don’t coordinate technology architecture planning with financial planning. While advanced cloud-native environments can reach resource utilization rates **above 60%**, most companies are **below 30%**, according to McKinsey. To improve this, businesses need to tightly connect their cloud business case with their cloud-architecture transformation. 6. **Assuming everything should move to the cloud** In some cases, keeping certain workloads on-premises can be more economical. For example, storage services or a small number of very large, homogeneous workloads may achieve cost structures on custom-designed on-premises infrastructure that match or beat cloud providers. McKinsey suggests that companies with a few massively scaled workloads should be selective and deliberate about what they move to the cloud. Overall, the core problem is not the technology itself but the failure to rethink financing, planning, and architecture for a consumption-based, flexible environment.

For CFOs, moving to the cloud is not just a technology shift—it’s a change in how IT is financed, managed, and forecast. Here are the key adjustments McKinsey recommends: 1. **Shift from CapEx to dynamic OpEx** On-premises IT is usually treated as a capital investment with long depreciation cycles. In the cloud, most spending becomes operational expenditure. Instead of buying capacity upfront, you pay for what you actually use. To make this work, finance teams need to: - Precisely measure demand and usage. - Understand incremental or marginal costs at any point in time. - Encourage teams to scale resources up and down based on real needs. McKinsey emphasizes that efficient cloud economics now depend on the ability to evaluate capacity demand and marginal costs continuously, so you avoid paying for idle capacity. 2. **Tie cloud budgets to business priorities, not just history** Traditional budgeting often leans heavily on historical spend. In the cloud, history is a weaker predictor because usage can scale rapidly with business activity. McKinsey notes that companies relying on past patterns can miss actual cloud spending by **more than 20%**. Instead, CFOs should: - Anchor cloud forecasts to specific business initiatives (e.g., a major promotion before Black Friday, launching a subscription model, or entering a new market). - Model how these initiatives will change traffic, transactions, and data usage—and therefore cloud costs. 3. **Build a strong FinOps capability** McKinsey recommends developing a FinOps (Financial Operations) function to bridge finance, technology, and the business. A FinOps team helps: - Application owners understand what drives their cloud spend. - Translate cloud usage into unit economics (e.g., cost per subscriber, cost per transaction). - Identify where elasticity can create savings and where workloads are better kept stable. One example: a video-streaming company measured cloud costs per subscriber and aligned compute capacity with demand patterns, achieving **over 95% accuracy** in predicting cloud consumption. 4. **Segment workloads by elasticity and economics** Not all workloads benefit equally from the cloud. Finance leaders should work with technology teams to: - Identify workloads with highly variable demand (ideal for cloud elasticity). - Separate stable, homogeneous workloads (such as long-term storage) that might be more cost-effective on-premises. McKinsey notes that some organizations with a small number of massively scaled, homogeneous workloads may achieve on-premises economics that are comparable to or better than cloud providers. 5. **Link financial models to architecture decisions** Cloud architecture choices (e.g., cloud-native design, autoscaling, right-sizing) directly affect utilization and cost. McKinsey points out that while advanced cloud-native environments can achieve utilization rates **above 60%**, most companies are **below 30%**. CFOs should ensure that: - Business cases for cloud are built together with architecture plans. - Financial models reflect how architecture changes (like refactoring or replatforming) will improve utilization and reduce waste. By reimagining IT spend as a flexible, consumption-based model and aligning it with business priorities, CFOs can move beyond simple cost-cutting and use the cloud to support growth, agility, and more precise unit economics.

McKinsey’s analysis suggests that a “cloud everything” approach is not always the most economical or practical. There are clear cases where keeping workloads on-premises can make sense. Situations where on-premises may be appropriate include: 1. **Massively scaled, homogeneous workloads** If your environment consists of a small number of very large, uniform workloads—such as certain storage or batch-processing systems—you may be able to design on-premises infrastructure that matches or even beats cloud economics. The scale and homogeneity of these workloads can justify custom-designed, highly optimized on-premises solutions. 2. **Stable workloads with limited demand variability** Cloud delivers the most value for workloads with significant swings in demand, where elasticity (scaling up and down) avoids paying for idle capacity. For workloads that are: - Predictable, - Steady over time, and - Not sensitive to rapid scaling needs, the economic advantage of the cloud may be smaller. In these cases, a well-managed on-premises environment can be cost-competitive. 3. **Storage-heavy use cases with consistent usage** McKinsey notes that some storage services may be better kept on-premises, especially when they involve large, consistent volumes of data. If you can design storage infrastructure tailored to your specific needs, the total cost of ownership may rival or improve on cloud storage pricing. 4. **When architecture and economics are misaligned** If your applications are not yet designed to take advantage of cloud-native capabilities (such as autoscaling, microservices, or serverless patterns), simply lifting and shifting them may not deliver the expected value. McKinsey points out that many companies that rush migration without rethinking architecture end up with low resource utilization—often **below 30%**, compared with **over 60%** in advanced cloud-native environments. In these cases, it can be more effective to: - Keep some workloads on-premises while you redesign or modernize them. - Move only those components that can immediately benefit from elasticity and cloud services. 5. **Strategic selectivity rather than all-or-nothing** McKinsey encourages companies, especially those with a small number of massively scaled workloads, to be selective about cloud adoption. This means: - Evaluating each workload individually based on elasticity, cost, and business criticality. - Comparing realistic cloud costs (including refactoring and operations) with optimized on-premises costs. - Building a hybrid strategy where some workloads remain on-premises while others move to the cloud. In practice, the most effective approach is often a hybrid model: use the cloud for variable, growth-oriented, and innovation-heavy workloads, while retaining certain stable, large-scale, or storage-heavy workloads on-premises when the economics and risk profile justify it.