For banks processing high payment volumes, dedicated cloud infrastructure may offer better long-term economics, operational control and predictability.
Public cloud has earned its place in banking. It gives institutions rapid access to infrastructure, removes large upfront hardware commitments and makes it easier to launch services without predicting capacity years ahead. Tier-1 and tier-2 banks, which usually plan their strategy a few years ahead, should think about scalability and related costs. Will services run in the public cloud be economically viable?
That matters in payments because volumes keep rising. UK Finance says the UK processed 49.7 billion payments in 2025, with cards accounting for 64% of the total. In the euro area, the ECB recorded 44.0 billion card payments in the first half of 2025, up 9.6% year on year. For established banks, authorisation, transaction processing and card payments are high-volume services that run continuously and scale.
Public cloud is not automatically the low-TCO option
The strongest argument for the public cloud is elasticity. A bank launching a digital proposition, entering a market or testing an unfamiliar volume profile can consume what it needs and scale quickly without paying for capacity before customers arrive.
The weakness appears when consumption becomes high. BCG warned in 2026 that cloud migrations without effective re-architecting and cloud financial operations can reproduce legacy complexity at potentially higher cost.
So what happens when a payments workload is predictable enough that the bank is renting the same large amount of capacity every day?
Consider a bank “A” processing 150 million transactions annually. A usage-based public-cloud processing model might cost around £1 million a year, while dedicated capacity could start at £1.25 million. At that volume, public cloud is cheaper.
Increase the workload to 450 million transactions. Even with lower unit pricing, the usage-based model could easily reach around £2.5 million. A dedicated environment might need only another £400,000 of investment, taking annual cost to roughly £1.65 million. These are illustrative figures, not market pricing, but they show the banks should model for scale.
Dedicated does not mean going backwards
A dedicated processing instance can still be cloud-based and provider-managed. The distinction is that compute, databases and recovery capacity are reserved for one institution rather than consumed from a shared processing environment.
That can suit banks that need predictable throughput, stricter isolation, greater control over release windows, or specific resilience and audit requirements. The trade-off is equally clear: the bank commits earlier to a larger baseline and gives up some pay-for-what-you-use flexibility.
European banks already show that it is not just an idea, but practice. ING reported that 68% of its physical processing cores were on its private cloud at the end of 2025 while it continued towards a hybrid model. Deutsche Bank said its 2025 cloud strategy had moved to a hybrid approach combining public cloud with a separate private-cloud environment, citing business demands and regulatory concerns including data residency. Santander reported that more than 97% of its infrastructure had moved to the cloud, while also expanding on-premises private-cloud capacity.
Why a dedicated private instance can work better for larger institutions
For tier-1 and tier-2 banks with high, sustained payment volumes, a dedicated processing instance can offer advantages that go beyond infrastructure choice:
- More predictable economics at scale. Dedicated capacity can make long-term processing costs easier to forecast, particularly when transaction volumes are consistently high, and infrastructure utilisation is relatively stable.
- Greater isolation and control. Everything can be segregated from other clients, giving institutions greater operational separation and clearer control over access.
- More flexibility over change. Dedicated environments can support more tailored release calendars, configuration changes, processing schedules and resource allocation. This matters for banks that cannot always align critical payment changes with a standard shared-service release cycle.
- Stronger operational visibility. Extended access rights, transparent monitoring, audit logs and flexible reporting give banks greater visibility into how the environment is operating, while the provider continues to manage the underlying processing service.
Where the difference is
The difference between shared processing and a dedicated instance depends largely on scale, volume predictability, operational requirements, and the level of control an institution needs.
Shared processing-as-a-service can remain attractive where volumes are lower or less predictable, speed of deployment is important, and institutions want to minimise initial infrastructure commitment. Dedicated processing becomes more relevant as volumes scale, workloads stabilise, or institutions require stronger isolation, greater release flexibility, and more control over their processing environment.
For tier-1 and larger tier-2 banks with sustained transaction volumes, however, dedicated capacity deserves a serious TCO comparison. The question is no longer simply, “Can public cloud scale with us?” Because it can. The better question is: “Once we have scaled, are we still paying for flexibility we no longer need?”
