What Actually Breaks During a Payments Platform Cloud Migration · Midcore Operations
Midcore Operations
Cloud & Infrastructure

What Actually Breaks During a Payments Platform Cloud Migration

Federal banking regulators found that migrations rarely fail on infrastructure. They fail at the exact boundary nobody wrote down clearly.

Published
18 October 2026
Reading time
9 min read
Author
Cloud & Infrastructure practice
Topic
Cloud & Infrastructure
Key takeaways
  • The FFIEC's Joint Statement on Risk Management for Cloud Computing Services describes a specific, recurring failure mode: management assumes effective security and resilience controls exist simply because systems now run in a cloud environment, when the actual division of labor depends entirely on the service model.
  • NIST SP 800-145 defines three cloud service models, Software as a Service, Platform as a Service, and Infrastructure as a Service, each shifting a different set of responsibilities between the financial institution and the provider. A migration that does not nail down which model applies to which workload is the single most common root cause of a control gap.
  • Access credentials, default settings, and overprovisioned permissions migrate right along with the data unless someone deliberately resets them, and regulators specifically flag this as a common post-migration gap rather than a hypothetical risk.
  • Change management processes and software development life cycles built for an on-premises environment do not automatically work in a cloud environment, and the regulators' own guidance says management should expect to augment, not just port over, these processes during a migration.

What actually breaks during a payments platform cloud migration is rarely the infrastructure itself. Federal banking regulators who studied this question directly found a more specific and more preventable failure: the division of responsibility between the institution and the cloud provider gets assumed rather than documented, and the gap between what someone assumed and what the contract actually says is exactly where the real damage happens.

The assumption that causes the actual damage

The FFIEC’s Joint Statement on Risk Management for Cloud Computing Services, issued jointly by the Federal Reserve, the FDIC, the OCC, the NCUA, and the CFPB’s principals, opens with a warning that reads like it was written specifically for a migration in progress. Management should not assume that effective security and resilience controls exist simply because the technology systems are operating in a cloud computing environment. The contractual agreement between the financial institution and the cloud service provider is what actually defines who is responsible for what. The fact that a workload now runs on someone else’s infrastructure decides nothing on its own.

That single sentence describes the actual mechanism behind most migration failures in this industry. A team moving a payments platform to the cloud is making a long list of decisions under time pressure. “Who patches this” or “who manages these credentials” quietly defaults to “the cloud provider, probably,” because the infrastructure feels like it should be managed by whoever owns it. The regulators’ point is that this default assumption is often wrong. It is wrong in a specific, structural way, tied directly to which service model the migration actually uses for that particular workload.

Why the service model is the real dividing line

NIST SP 800-145 defines the three cloud service models that the FFIEC’s own guidance builds on directly. Software as a Service runs the provider’s applications on the provider’s infrastructure. The financial institution is mainly responsible for user access, identity management, and the relationship itself. Platform as a Service lets the institution deploy its own applications on the provider’s platform. That adds responsibility for the development and deployment pipeline on top of the SaaS baseline. Infrastructure as a Service goes further still. The institution deploys and operates its own operating systems and applications, and owns the controls over operations, data, and data storage, while the provider manages only the physical data center underneath it all.

Each model draws the line in a different place. A payments platform migrating a customer-facing application to a SaaS tool is taking on a narrow slice of responsibility. The same platform migrating its core transaction processing infrastructure to an IaaS environment is taking on a much wider one. That wider slice includes operating system patching and most of the operational security controls that used to live inside the institution’s own data center walls by default. A migration plan that treats “moving to the cloud” as one single decision, rather than a different responsibility boundary for every workload and service model involved, is building its security assumptions on a line nobody actually drew.

The credentials that migrate whether anyone wants them to or not

Federal guidance is specific about a second failure mode that shows up constantly in practice. Access credentials and permissions that existed on the old system do not disappear during a migration. They come along, by default, unless someone deliberately intervenes. The guidance calls out default access credentials that are never changed, and the risk of overprovisioning access credentials, as common, avoidable problems in cloud environments specifically.

A payments platform’s on-premises environment typically accumulates access sprawl over years. Service accounts nobody remembers creating. Admin permissions granted for a one-time task and never revoked. Default credentials on a tool that was supposed to get locked down before launch and never was. A migration that lifts this environment into the cloud without a deliberate access review does not fix any of that. It relocates the entire mess into an environment where the provider’s own tooling may actually make overprovisioned access easier to exploit, not harder, because cloud platforms are built to make access easy to grant at scale. The guidance’s recommended practices are specific: limiting account privileges, multifactor authentication, frequent review of account access, and changing default credentials before anything goes live, not after an incident reveals they were never changed.

A concrete example: the patching responsibility nobody actually assigned

Picture a payments platform migrating its transaction processing stack to an IaaS environment. The engineering team’s migration plan covers moving the application, setting up the new network configuration, and validating that transactions process correctly in the new environment. It does not include an explicit line item for who patches the operating system going forward. Under the previous on-premises setup, the infrastructure team handled that automatically, and nobody specifically re-assigned that responsibility for the new environment.

Three months after the migration completes, a critical vulnerability is disclosed in the operating system the platform runs on. The cloud provider patches its own underlying hypervisor and physical infrastructure, exactly as its IaaS responsibilities require. It does not patch the guest operating system the payments platform’s own team is running, because under the IaaS model, that was never the provider’s job in the first place. The vulnerability sits unpatched for weeks. Nobody decided to accept that risk. The migration plan simply never explicitly assigned operating system patching to anyone on the new side of the line. The infrastructure did not break. The assumption about who was watching it did.

Why change management has to change too

The FFIEC’s guidance makes a point that is easy to skip past: change management controls that worked fine on-premises do not automatically transfer to a cloud environment, and management should expect to augment existing change management processes and the software development life cycle rather than simply reuse them unchanged. Cloud implementations also frequently lean on microservice architectures to get faster, more agile deployment, which the guidance notes comes with its own tradeoff. Multiple microservices can meaningfully increase the platform’s attack surface even as they make development faster.

For a payments platform, this means the change management process that reliably caught a bad deployment in the old environment may simply not have visibility into a new, more distributed cloud architecture built around many smaller services instead of one larger application. A migration that moves the application without also revisiting how changes get reviewed, tested, and approved in the new architecture is carrying over a process built for a system that no longer exists in the same shape.

The encryption tradeoff a migration forces into the open

One decision the guidance highlights deserves specific attention because it is genuinely a tradeoff, not a simple best practice. Many cloud providers offer cloud-based key management services for encrypting sensitive data, which integrate conveniently with other cloud services. The guidance is direct about the catch: cloud-based key management may allow administrators from the cloud service provider to access encrypted information, which is precisely why some institutions choose to use their own encryption and key management instead, built into the application itself.

That choice has a real cost on both sides. Provider-managed keys are simpler to operate and integrate automatically with other cloud tooling. Institution-managed encryption keeps the provider out of the picture entirely but requires building and maintaining that capability inside the application, which can interfere with the automated controls the cloud platform otherwise offers for free. A migration plan has to make this decision deliberately, for the specific sensitivity of the data involved, rather than accepting whatever the default happens to be.

What a migration plan actually needs to cover

The fix for all of this is not more caution in the abstract. It is a migration plan that explicitly documents, workload by workload, which service model applies, what the resulting responsibility split actually is, and who owns each specific control on the institution’s side of that line. Access credentials need a deliberate review and reset as part of the migration itself, not a follow-up task scheduled for after launch. Change management and the software development life cycle need to be re-evaluated against the actual architecture the migration produces, not assumed to carry over unchanged. And the encryption key management decision needs to be made explicitly, weighing the convenience of a provider-managed service against the control of managing it in-house.

Building a migration around these specific decisions, rather than a generic lift-and-shift checklist, is exactly the discipline behind Midcore’s cloud migration work. Where the underlying question is less about moving an existing system and more about how the target architecture should be designed in the first place, cloud strategy and architecture is where that gets settled before the migration plan is written. And because the access, encryption, and change management gaps described here are exactly what a sponsor bank’s own review or an examiner’s visit will eventually test, multi-cloud security and compliance work is built to produce the evidence that those specific gaps were closed, not just assumed away.

Frequently Asked Questions

What is the most common thing that actually goes wrong during a cloud migration for a payments company?

Federal banking regulators describe the core failure as a misunderstanding of shared responsibility: management assumes the cloud service provider handles a control that, under the specific service model in use, the financial institution is actually responsible for. This is not a hypothetical risk. The FFIEC’s own joint statement on cloud computing names it directly as a key driver of control gaps after migration.

Do access permissions and credentials need to be rebuilt during a cloud migration, or can they just be copied over?

They need to be rebuilt deliberately, not copied over. Federal guidance specifically flags default access credentials that are never changed and overprovisioned account access as common, avoidable risks in cloud environments. A migration that copies an existing on-premises permission structure into the cloud without review tends to carry every excess permission and stale credential along with it.

What are the three cloud service models, and why do they matter for a migration?

NIST SP 800-145 defines Software as a Service, Platform as a Service, and Infrastructure as a Service. Each shifts a different line between what the cloud provider manages and what the financial institution manages. In SaaS, the institution is mainly responsible for user access and configuration. In PaaS, it also owns application development and deployment controls. In IaaS, it owns the operating system, data, and most operational controls. A migration plan that does not identify which model applies to each specific workload is building on an assumption rather than a documented division of responsibility.

Sources: FFIEC Joint Statement on Risk Management for Cloud Computing Services, April 30, 2020, and NIST Special Publication 800-145, The NIST Definition of Cloud Computing.

Cloud & Infrastructure practiceMidcore Operations · 18 October 2026
All insights
Free consultation

Get your free Assess

Tell us what is happening. We reply within one business day with a US-based practice lead, not a salesperson.

  1. 01Within 1 business dayA US-based practice lead replies and books a 30 minute call.
  2. 02On the callWe map the problem, the volume, the partners, and the constraints.
  3. 03After the callYou get a written read of one to three pages, yours to keep.
  • Handled under NDA
  • Written, not a slide deck
  • No obligation to continue

No obligation. We will tell you if you do not need us.

Before you go

Take the written read with you

Every engagement opens with an Assess: a written read of what is happening, what it costs, and what to do about it. It is quoted on its own and you can stop after it.

+1 (786) 619-0152