What Cloud Core Banking Really Costs: A Worked Example for a Mid-Size Bank
A labelled five-year model for a 400,000-account bank shows why the subscription is rarely the biggest number, and where migration, parallel-run and staffing budgets overrun.
Why the subscription quote is the smallest surprise in core banking software cost
The vendor's annual subscription is the easiest number to get for a new-core business case. On its own, it's also the least useful. In our model below, five-year core banking software cost comes to roughly three and a half times the subscription line. Most of the money goes to implementation partners, data migration, running two cores at once, and the bank's own people. Boards tend to underestimate those lines, and that's where projects overrun.
The five-year model below is hypothetical. It isn't drawn from any real bank or vendor contract. It uses typical 2026 market ranges, which vary widely by contract, region and vendor. Use it to structure your own estimate and to pressure-test proposals, then confirm every figure with the vendors and integrators you're actually talking to.
The hypothetical bank
- Size: about 400,000 accounts across roughly 240,000 customers. Retail current and savings accounts, term deposits, SME accounts, and a consumer and small-business loan book.
- Starting point: a heavily customized, batch-based on-premise core more than 20 years old. It costs about $3.2 million a year to run once licences, hardware, data center and support staff are counted.
- Target: a cloud core bought as SaaS, hosted and operated by the vendor in-region.
- Approach: a phased migration over about 24 months. Deposits go first, then lending, with a controlled parallel run before each cutover.
- Account growth: 3 to 4 percent a year.
Cost falls into five buckets: subscription, implementation, data migration, parallel run and internal staffing. We've added two lines buyers often leave out, third-party integration and contingency.
The hypothetical five-year budget, line by line
The table shows the planning figure we'd use for this bank and the spread we'd expect across credible proposals. All figures are indicative USD. Local pricing, labor rates and vendor discounting differ sharply between markets.
| Cost bucket | Planning figure (5 years) | Typical range | When it hits |
|---|---|---|---|
| SaaS subscription (licence, hosting, vendor operations) | $12.3M | $9M to $18M | Ramped from year 1, full rate from go-live |
| Implementation (vendor services plus systems integrator) | $10.5M | $7M to $16M | Years 1 to 2 |
| Data migration (tooling, mapping, cleansing, mock runs) | $2.4M | $1.5M to $4M | Years 1 to 2 |
| Parallel run and legacy overlap | $3.0M | $2M to $5M | Late year 1 to year 3 |
| Internal staffing (dedicated team plus backfill) | $8.0M | $4M to $12M | Years 1 to 3 |
| Integration (payments, cards, digital, AML, general ledger) | $2.5M | $1.5M to $5M | Years 1 to 2 |
| Contingency (20% of one-off costs) | $5.3M | 15% to 30% of one-off costs | Held centrally |
| Total | about $44M | $30M to $65M |
That's roughly $110 per account over five years. About $31 of it is subscription.
How the subscription line was built
We assumed a steady-state rate of about $6.50 per account per year, or around $2.6 million a year at 400,000 accounts. Year 1 is billed at 50 percent while the platform is in build and test. After that, account growth and indexation push the bill up, partly offset by lower per-account rates as the bank moves into higher volume bands. Year by year: $1.3M, $2.6M, $2.7M, $2.8M, $2.9M.
For a bank this size, an indicative 2026 range is $4 to $12 per account per year. Where you land depends on which modules are included, how many non-production environments come with the contract, and how hard you negotiate.
Why implementation dwarfs migration tooling
The $10.5 million implementation line assumes 35 to 50 vendor and integrator consultants for about 20 months, covering product configuration, integration build, testing support and cutover. Blended day rates are the biggest variable. Integrator teams staffed mainly from India, Eastern Europe or the Philippines can cost around half as much as onshore teams in London, Frankfurt, Dubai or Sydney. The trade-off is time-zone overlap and, sometimes, local domain knowledge.
Internal staffing is real money even if nobody invoices it
The $8 million staffing line pays for a dedicated team of about 30 for 24 months: product owners, business analysts, testers, data owners, an architecture lead and change managers. It also pays for backfill for the operations staff you pull into the program. The planning figure assumes a fairly high-cost labor market. Fully loaded costs can run under $40,000 a year per person in parts of South and Southeast Asia and well over $150,000 in Western Europe, North America and the Gulf. Rescale this line to your own market before relying on it.
How core banking licensing models change the subscription line
The same bank can receive five-year subscription quotes that differ by a wide margin purely because vendors define the billable unit differently. Before comparing, convert every proposal to one basis: total five-year subscription at your projected volumes, including indexation and minimum commitments.
Common core banking licensing models
- Per account, banded: the most common SaaS model. You pay per active account, with lower rates in higher bands. Check whether dormant, closed-but-retained and internal or GL accounts are billable. At 400,000 accounts, that definition alone can move a meaningful share of the bill.
- Per customer: simpler where customers hold several products. It can penalize banks with many single-product customers.
- Transaction or event based: cheap at low volumes, hard to forecast. It also punishes growth in real-time payments, which keep rising in most markets, from UPI in India and Pix in Brazil to instant payment schemes in Europe.
- Platform fee plus modules: a fixed annual fee with priced add-ons for lending, cards, Islamic finance products, collections and so on. Predictable, but watch for core functions sold as extras.
- Term licence, bank-hosted: some cloud-native vendors license software the bank runs in its own AWS, Azure or Google Cloud tenancy. Thought Machine has commonly been deployed this way. The licence line looks smaller, but you then carry hyperscaler bills (indicatively several hundred thousand to over a million dollars a year at this scale) plus your own platform engineering team.
SaaS versus bank-hosted cloud
For most banks under a million accounts, vendor-operated SaaS is the better fit. A bank this size rarely wants to staff a 24/7 site reliability team for its core, and SaaS puts uptime accountability on the vendor. Bank-hosted cloud makes sense when a regulator or board wants direct control of infrastructure, or when you already run a mature cloud platform team.
Mambu, 10x Banking, Temenos, Infosys Finacle, Oracle and TCS BaNCS all offer some form of hosted or managed option. Hosting regions and service terms differ, so ask each vendor which regions it can serve from today, not on its roadmap.
Data residency can raise the price
Hosting rules feed straight into cost. Examples include India's localization requirements for payment system data, in-country hosting expectations from regulators in Saudi Arabia and the UAE, EU outsourcing guidance and DORA, Singapore's MAS outsourcing and technology risk guidelines, and Australia's CPS 230. Each can push you toward a particular cloud region, a dedicated tenant, or extra audit and exit provisions. Dedicated single-tenant deployments usually carry a clear premium over multi-tenant SaaS, sometimes 20 to 50 percent. Have local regulatory counsel confirm what your jurisdiction actually requires before you pay for isolation you may not need.
Where core banking migration budgets usually overrun
Overruns seldom come from one big mistake. They come from schedule slip, because almost every line in the model is time-based.
Worked example: the cost of a six-month delay
Say the hypothetical bank's deposits cutover slips two quarters because mock migrations keep failing reconciliation. Each quarter of delay costs roughly:
- Integrator and vendor services: $1.2M to $1.5M at the team size above.
- Internal team and backfill: about $1M.
- Legacy core running costs: about $0.8M, which the business case assumed would stop.
- The new subscription: about $0.65M if billing started on a fixed date rather than at go-live.
Six months adds roughly $6M to $8M, more than the entire contingency line. That's why a clause tying subscription start to production go-live is worth more than a few points off the per-account rate.
The five usual culprits
- Data quality discovered late. Old cores accumulate free-text fields, duplicate customer records and product variants nobody remembers launching. Budget for at least four full mock migrations, with reconciliation signed off by finance. Start profiling data before you sign with a vendor.
- Product rationalization skipped. Suppose a bank carries 300 deposit product variants and migrates them one-for-one. It pays to configure and test every one. Cutting to 60 first can save months.
- Parallel run longer than planned. Our model assumes about nine to twelve months of combined overlap across phases. Auditors and regulators may want extended reconciliation, especially for interest accruals and fees, and every extra month carries both cores' running costs.
- Integration counted too narrowly. Payment scheme connectivity, card processors, AML and sanctions screening, regulatory reporting, the general ledger and the digital front end all need rework. A bank this size commonly has several dozen interfaces into the core. Count them before you sign.
- Customization creep. Every request to make the new core behave like the old one adds change requests now and upgrade cost later, which undercuts the case for SaaS.
Contract terms and checks before you approve the budget
Once you've rebuilt the model with your own volumes and labor rates, use these checks to judge whether a proposal is credible.
Commercial terms to negotiate
- Subscription start: tie full billing to production go-live of the first migrated portfolio, with a capped build-phase fee.
- Indexation cap: a fixed cap, commonly 3 to 5 percent a year, or CPI with a ceiling. Banks with uncapped CPI clauses saw sharp increases during recent inflation spikes.
- Billable unit: define in writing how dormant, closed and internal accounts are treated.
- Non-production environments: confirm how many test, staging and performance environments are included. Extras are often priced in the tens to low hundreds of thousands of dollars a year each.
- Exit and data return: a defined exit-assistance fee and export format. Many regulators expect a documented exit plan for material outsourcing anyway.
- Implementation commercials: fixed price or capped time-and-materials for defined phases, with payments tied to milestones such as successful mock migrations rather than elapsed months.
Red flags in a proposal
- Implementation priced below the first two years of subscription for a bank of this size.
- No named data migration lead or method, or migration labelled "client responsibility."
- A cutover plan with no parallel run or reconciliation period.
- Rate cards that expire before your realistic go-live date.
- References that are all greenfield neobanks when you're replacing a 20-year-old core.
A decision rule for the board paper
If your five-year core banking software cost estimate puts subscription above 40 percent of the total, you've probably undercounted implementation, staffing or overlap. Rebuild those lines before the paper goes to the board. Then send every shortlisted vendor and integrator the same account volumes and growth assumptions, ask them to price against the seven buckets in the table, and have finance and your external advisers review the assumptions before you commit.
Frequently asked questions
How much does cloud core banking cost for a bank with around 400,000 accounts?
In our hypothetical model, the five-year total is about $44 million, and credible proposals could plausibly fall between $30 million and $65 million. The SaaS subscription made up under a third of that. Implementation, internal staffing, data migration and running old and new systems side by side account for the rest. These are typical 2026 market ranges that vary by contract and region, so confirm them with vendors.
Is per-account pricing better than transaction-based pricing for a core banking platform?
For most mid-size banks, banded per-account pricing is the safer choice because it's easier to forecast. Transaction-based pricing can look cheaper at first but climbs as real-time payment volumes grow. Whichever model you're offered, convert each quote into a five-year total at your projected volumes so you're comparing like with like.
How long should a parallel run last in a core banking migration?
Phased migrations commonly plan one to three months of parallel running or reconciliation per cutover. In our model that adds up to nine to twelve months of legacy overlap across the program. The right length depends on product complexity, data quality, and what your auditors and regulator expect. Budget it as its own line, because every extra month means paying to run both cores.
Can a mid-size bank cut costs by hosting a cloud-native core in its own cloud account?
Sometimes, but the savings often go to hyperscaler bills and the platform engineers needed to run the system around the clock. Bank-hosted cloud makes sense if you already have a mature cloud operations team or face specific regulatory hosting requirements. Otherwise, vendor-operated SaaS usually costs less in total and carries less operational risk. Check your jurisdiction's outsourcing and data residency rules with qualified advisers before deciding.
Written by the Adviser Hub Cloud editorial team. Prices, features and rules change often - confirm details with the vendor or a licensed professional before you decide.