Your Cloud Program Is Probably Solving the Wrong Problem
The issue isn’t whether to invest in cloud and infrastructure. Most US banks already have. The issue is that lift-and-shift migrations, ungoverned multi-cloud sprawl, and FinOps dashboards with no executive ownership are producing spend without returns. McKinsey’s Global Banking Annual Review 2025 shows the sector’s price-to-book sits 67% below other industries. This guide explains why – and what to do differently.
US banks spend roughly $600 billion annually on technology. Yet, as McKinsey’s Global Banking Annual Review 2025 reveals, the sector’s price-to-book ratio sits at 1 – 67% below the average for all other industries. The spend is real. The returns are not.
The gap isn’t a budget problem. It’s an architecture problem. Deloitte’s Tech Trends 2026 report found that the infrastructure built for cloud-first strategies simply cannot handle AI economics – pushing organizations toward a more deliberate hybrid model.
This guide breaks down what US BFSI executives need to reckon with: where cloud investments actually pay off, where they quietly erode value, and the strategic missteps that are hardest to see from the inside.
What’s Changed: AI Economics and the New BFSI Infrastructure Reality
The cloud-first era made sense when workloads were predictable and AI was experimental. That calculus has shifted.
According to Deloitte’s Tech Trends 2026, AI token costs dropped 280× in two years – yet some enterprises still face monthly AI bills in the tens of millions because usage scaled faster than cost management did. The economics of running AI at scale demand a different infrastructure model.
The emerging default for BFSI is strategic hybrid: cloud for elasticity, on-premises for consistency, edge for immediacy. This isn’t a compromise – it’s a design choice. Real-time fraud detection, personalization engines, and regulatory reporting each have different latency, data residency, and cost profiles. A single-cloud approach rarely handles all three well.
Explore how AI-ready cloud strategy translates into practice for organizations navigating this shift.
What Works: Foundations Before Fireworks
KPMG’s Global Tech Report 2026 for Financial Services found that over half of financial services organizations say data, cloud, and cybersecurity deliver the majority of their digital value. Yet 89% of technology leaders – who identify as innovators or fast followers – cite technical debt, talent gaps, and security concerns as the primary blockers to scalable outcomes.
The pattern is consistent: organizations that invest in foundations first extract more from AI programs later. Those that skip ahead discover those same gaps mid-project.
A practical foundation-first sequence for BFSI:
- Govern your data first. Clean, classified, well-governed data platforms are the prerequisite for every AI, fraud, and personalization use case.
- Build resilient cloud landing zones. Establish clear identity and access management, network segmentation, and compliance guardrails before migrating workloads.
- Instrument observability early. Unified monitoring and incident management need to be operational before AI agents or automated processes run in production.
- Sequence workloads deliberately. Start with analytics and customer-facing digital experiences. Move core banking systems only after architecture and governance are stable – informed by application modernization patterns that treat risk as a first-class concern.
What Fails: Cloud Spend, Migration Overruns, and Fragile Operations
Consider this scenario: A mid-sized US bank completes a two-year cloud migration. Cloud costs run 35% over projection. The FinOps team produces dashboards. No one has authority to act on them. Two years later, three separate monitoring tools generate alerts that no single team is accountable for resolving.
This is not unusual. It is the predictable outcome of three compounding failures:
- Lift-and-shift without capacity planning. Moving legacy workloads unchanged inflates cloud costs without improving performance or resilience.
- FinOps as a reporting function. Per McKinsey’s 2025 banking research, banks spend ~$600B annually on technology with ROE barely clearing the cost of capital. FinOps dashboards without executive ownership do not bend that curve.
- Multi-cloud and tool sprawl. No unified observability means no accountability – and slower incident response when it matters most.
Two decision rules worth applying: if workload patterns, data egress, and licensing costs make cloud structurally more expensive than modern on-premises infrastructure, repatriation deserves a serious look. If elasticity, resilience, and AI experimentation speed outweigh unit cost concerns, cloud is still the right environment — but platform monitoring and management needs to be built in from day one.
What Leaders Get Wrong: Cloud Is a Revenue and Capacity Strategy
As McKinsey’s banking review highlights, only 4% of US new credit card applicants and checking account openings now come from the loyalty loop – down from 10% and 25% respectively in 2018. More than half of US consumers already use generative AI tools, and nearly all say they would eventually switch to another provider if their bank doesn’t keep up with this shift.
That means fragile, legacy-heavy infrastructure is no longer just an IT cost problem. It is a revenue risk. The inability to support real-time personalization and seamless digital journeys drives churn – quietly, consistently, and at scale. Understand why digital experience is now central to BFSI strategy and how infrastructure underpins every customer interaction.
The capacity dimension compounds this urgency. Per McKinsey’s August 2025 analysis of US hyperscale data center capacity, primary market vacancy hit a record low of 1.9% in 2024. More than $2.8 trillion of the projected $7 trillion in global data center capex will flow into the US through 2030, with demand growing at a 16% CAGR. Banks delaying hybrid cloud expansion decisions face tighter availability, higher unit costs, and reduced negotiating leverage with hyperscalers.
Two governance changes that close this gap:
- A CFO–CIO partnership with shared KPIs across cost, resilience, and digital growth – not separate budget conversations.
- Capacity planning and data-center dependence elevated to board-level discussion, not delegated to procurement. Pair this with business continuity planning that accounts for infrastructure scarcity, not just outage scenarios.
Choosing the Right Partner Model Without Inheriting Their Mistakes
Large cloud migrations overrun when vendors are incentivized on time and materials and lack BFSI-specific security and regulatory depth. The Bank of England’s Oracle migration – widely covered in financial and technology press as having tripled in cost – illustrates what happens when scope, accountability, and domain expertise are misaligned from day one.
What good looks like for US BFSI executives:
- Outcome-based contract phases linked to resilience, cost reduction, and incident response metrics – not effort hours.
- Integrated capabilities across data, cloud, cyber, and digital experience – because migrations that touch only one tower tend to expose gaps in the others.
- Operating model support alongside the technology build – SRE, AIOps, and cybersecurity services embedded in the engagement, not added after go-live.
The right partner architects and operates environments over time. A one-off migration is a handoff. A long-term operating model is a competitive advantage.
FAQ: Straight Answers for BFSI Executives
Why does our cloud spend keep rising quarter after quarter, even though we “do FinOps”?
Most FinOps programs focus on visibility – tagging, anomaly alerts, rightsizing recommendations. Without executive ownership to enforce trade-offs, spend decisions default to speed and uptime over cost. FinOps only bends the curve when it carries organizational authority. Explore practical cloud cost optimization approaches that go beyond dashboards.
What workloads should a bank move to cloud first, and which core systems should stay on-premises longer?
Start with analytics, reporting, and customer-facing digital workloads – where elasticity and speed deliver immediate value. Move core banking systems only after you have stable governance, clean data pipelines, and a proven operating model. Core migration done prematurely is the most expensive rework in BFSI cloud programs.
Do US banks still need on-premises disaster recovery if most workloads are in the cloud?
For most banks, yes – particularly for systemically critical workloads. Cloud providers offer resilience, but shared infrastructure, region-level outages, and contract terms mean on-premises or co-location DR remains a prudent hedge. See how a leading US bank enhanced incident management with AI as part of a broader resilience strategy.
How should CIOs and CFOs share accountability for cloud cost and value?
Define joint KPIs: cost per transaction, infrastructure availability, time-to-deploy for new digital capabilities. Separate budget conversations produce separate incentives. Shared metrics produce shared decisions.
Start Where the Risk Is Highest
Cloud infrastructure decisions made in 2025 and 2026 will determine whether your bank captures AI-driven productivity gains – or absorbs the costs without the returns. Waiting for a perfect architecture or a cleaner budget cycle is itself a strategic choice, and not a neutral one.
If you want to assess where your current cloud and infrastructure strategy stands – and what a pragmatic path forward looks like for your environment – talk to our team or reach out directly at inquiries@scalence.com.
