Cloud Infrastructure Cost Optimisation in Cloud Native Fintech: Balancing Innovation and Budget

Fintech sits at a genuinely interesting inflection point right now. Born in the cloud and built for scale, these organisations have leaned into cloud native architecture to get the agility that defines modern financial services. But that innovation brings a real challenge with it: managing infrastructure spend without giving up an inch on security, compliance, or the speed that gives fintechs their edge.

This article looks at cloud cost optimisation in cloud native fintech, and how these companies balance the pull toward innovation with the discipline of budget control. It draws on public case studies from global players including Nubank, PicPay, KOHO and others, alongside homegrown New Zealand fintech stories from Dosh, PaySauce, Hnry, Wedge and Akahu, to build a picture of the strategies, tools and organisational habits that make sustainable cloud cost management possible.

For New Zealand fintechs operating in a small, tightly regulated market with its own data sovereignty expectations, these lessons matter more than most. The organisations that get the balance right between innovation and cost efficiency will not just survive, they will out compete rivals who treat cost control as an afterthought.

1. Introduction: the fintech cloud cost challenge

When I talk with fintech leaders across Aotearoa and further afield, one worry keeps coming up: how do we keep innovating while keeping cloud costs under control? It is a question that goes right to the heart of what it means to run a modern fintech.

The numbers back this up. Global banking, financial services and insurance IT spend is projected to top $761 billion in 2025, and cloud cost management has become a genuine board level concern rather than an engineering side project. Flexera's State of the Cloud research has, for two years running, found that managing cloud spend now ranks as the single biggest challenge organisations report, ahead of security. For fintechs, the pressure is doubled: they operate in a heavily regulated environment while still needing the speed and agility that defines the sector.

The stakes are real. A migration that is not planned properly can send costs upward instead of down, leading to frustration and lost revenue rather than the promised efficiency gains. For New Zealand fintechs, where margins tend to be thinner and scale more limited than for US or European peers, the impact of cloud overspend can threaten the viability of the whole business.

Here is the thing though: cost optimisation and innovation are not actually opposing forces. Done well, cloud cost management becomes an enabler of innovation rather than a handbrake on it. A well run cloud setup creates mental margin. Less firefighting, more building, so scarce engineering hours go toward features rather than infrastructure triage. Engineers who are not manually hunting down idle resources can spend that time shipping the things customers actually value.

This article asks a central question: how can cloud native fintech organisations, in New Zealand and globally, optimise cloud infrastructure costs while protecting security, compliance and the capacity to keep innovating? Through real world case studies and emerging best practice, it sets out a framework for cost optimisation suited to the particular constraints of financial services.

2. Why cloud costs spiral in fintech

2.1 The innovation versus cost tension

Fintech companies are defined by their ability to move fast, experiment constantly, and scale quickly. That imperative often leads to infrastructure choices that favour speed over efficiency. Kubernetes adoption is a good example: the Cloud Native Computing Foundation has found that container orchestration has driven up cloud spend for a large share of organisations, with over provisioning frequently cited as the direct cause of overspend.

The tension is structural. Engineering teams are rewarded for shipping features and keeping systems reliable, while finance teams are judged on cost control. Without deliberate alignment between the two, teams end up lacking unified cost visibility, clear attribution of spend, anomaly detection, and automation for FinOps, which is exactly the gap that lets costs run away unnoticed.

2.2 Common cost drivers in fintech cloud environments

Case studies and industry research point to a handful of recurring patterns:

Resource sprawl and over provisioning. Teams often err on the side of caution, over allocating capacity to avoid performance problems during traffic spikes. This shows up strongly in fintech, where transaction volumes can spike hard around events like Black Friday, Mother's Day, or national payday cycles. It is common to find a development environment running at well under a quarter of its allocated capacity, quietly burning budget month after month.

Lack of cost visibility. Without granular, near real time dashboards, finance and engineering teams cannot see which products or teams are actually driving spend. It is not unusual for organisations starting a cost optimisation journey to find that only around half of their cloud spend can be confidently allocated to a business unit or product. In New Zealand, where many fintechs run lean teams, this visibility gap can persist for years before anyone tackles it properly.

Inefficient data retention. Compliance mandated logs and backups often sit indefinitely on expensive, high performance storage tiers. Financial services regulation demands extensive audit trails, but organisations frequently fail to tier that storage down over time. The Reserve Bank of New Zealand's record keeping expectations add another layer of complexity for local fintechs to manage here.

Decentralised decision making. Individual teams making independent cloud procurement calls leads to duplicate resources, overlapping tools, and weaker negotiating leverage with vendors. This pattern shows up often in growing fintechs where decentralisation is seen, rightly or wrongly, as a way to preserve speed.

Static commitment strategies. Many fintechs rely on fixed Reserved Instance or Savings Plan commitments that cannot flex with highly dynamic, cyclical workloads. As one global fintech's FinOps team put it, their cyclicality and dynamic workloads really needed something different from a static Compute Savings Plan approach.

2.3 The security and compliance dimension

What sets fintech apart from other cloud native sectors is that security and compliance are not negotiable. Cost cutting cannot come at their expense. Disabling controls such as data loss prevention, SIEM or web application firewalls to save money exposes sensitive financial data to real threats, and rapid, poorly planned deletion of resources can wipe out audit trails, backups or encryption keys that regulators expect to see.

Fintechs cannot simply switch off workloads or delay upgrades if doing so puts regulatory compliance or customer trust at risk. Cost strategies have to be designed with security and compliance built in from the start, not bolted on afterward.

For New Zealand fintechs, the Privacy Act 2020 adds a further layer. Offshore hosting can expose sensitive financial and legal data to foreign jurisdictions, including exposure under the US CLOUD Act, which complicates Privacy Act compliance and shapes architecture decisions that carry direct cost implications.

3. The FinOps framework: a strategic foundation

3.1 Understanding FinOps

FinOps, short for financial operations, has become the dominant framework for cloud cost management. It is best understood as an operational framework and cultural practice aimed at maximising the business value organisations get from cloud and technology, bringing engineering, finance and business teams into the same conversation about trade offs between speed, cost and performance, rather than treating cost control as finance's problem alone. As Google Cloud's explainer on FinOps puts it, FinOps is not about becoming cheap, it is about removing blockers so engineering teams can deliver better features and migrations faster while making informed, data driven decisions about where to invest.

The framework runs across three phases:

Inform: giving teams visibility into cloud usage and cost
Optimise: improving efficiency through rightsizing, discount programmes and architectural change
Operate: embedding cost awareness into culture and process on an ongoing basis

3.2 A broader scope for 2025 and beyond

FinOps practice has widened out well beyond public cloud infrastructure. The current framework increasingly covers SaaS applications, AI workloads, licensing and even on premises data centres, giving organisations a more nuanced way to manage cost across every category of technology spend rather than just the AWS or Azure bill. For fintechs running across multiple clouds and a growing stack of SaaS tools, this broader scope matters because it stops cost governance falling through the cracks between systems.

3.3 Why FinOps matters for fintech

Financial services organisations face a genuinely tough balancing act: they need operational scalability alongside strict regulatory obligations and real profitability pressure. FinOps addresses this by embedding financial intelligence directly into the infrastructure lifecycle rather than treating cost review as a quarterly afterthought.

For New Zealand fintechs, adopting FinOps practice matters even more given the limited scale of the local market. The emergence of homegrown FinOps providers like Yarken and Spotto, alongside global platforms such as Vantage and CloudHealth being taken up by New Zealand firms, suggests a market that is taking cloud cost management seriously.

4. Technology strategies for cost optimisation

4.1 Compute optimisation

Rightsizing. One of the fastest wins available is eliminating over provisioned instances. AWS Compute Optimizer, for example, recommends rightsized instances based on actual workload need, letting teams identify underutilised capacity and downsize accordingly, often onto ARM based Graviton instances for further savings.

Graviton processors. AWS Graviton has become a genuine lever for fintech cost control. Nubank, serving around 94 million customers across Brazil, Mexico and Colombia, migrated key workloads to Graviton processors and achieved a 14% reduction in cloud costs alongside a 21% cut in carbon emissions. Similar migrations reported by other financial services firms, including Techcom Securities and Paytm, point to compute cost reductions in the range of 30% to 50% once workloads are fully migrated.

Spot Instances. For non production workloads and fault tolerant production services, Spot Instances can deliver savings of up to 90% compared with on demand pricing, making them a natural fit for the bursty, test heavy workloads common in fintech engineering.

Serverless computing. AWS Lambda and AWS Fargate remove the need for always on servers, charging only for actual execution time. This suits fintech workloads well, given how much their demand varies through the day and across seasons. New Zealand fintech Wedge has architected its platform around NestJS with TypeScript Lambda functions, a clear local example of serverless principles being put into practice.

4.2 Container and Kubernetes optimisation

Kubernetes has become the default orchestration platform for cloud native fintechs, but it introduces its own cost management headaches. The dynamic nature of container workloads makes proper resource allocation genuinely difficult to get right by hand.

Automated resource management and node optimisation. Several fintechs report meaningful savings from automating Kubernetes optimisation rather than relying on manual tuning, using tools such as Cast AI and DoiT to rebalance workloads, autoscale automatically, and cut overprovisioning even in environments that already made heavy use of Spot Instances.

EKS Auto Mode. Canadian fintech KOHO reduced infrastructure costs by 30% and achieved five times faster service scaling by adopting Amazon EKS Auto Mode, which fully automates cluster management for compute, storage and networking. Observability company StormForge reported a similar outcome, cutting infrastructure costs by 30% to 40% and cluster upgrade time by 65% using the same approach.

4.3 Storage optimisation

Fintechs deal with enormous datasets for real time transactions, risk analysis and compliance reporting. Tiered storage helps bring the cost of that down considerably:

Amazon S3 Intelligent Tiering automatically shifts rarely accessed data to cheaper tiers, typically saving 20% to 40%
Amazon Glacier and Deep Archive store long term compliance data at up to 90% lower cost than standard S3
Amazon Aurora Serverless auto scales databases so firms only pay for what they actually use, rather than provisioning for peak load year round

Payments platform Pismo, the cloud native core banking provider Visa acquired in 2024 and now used by Dosh here in New Zealand, is a good example of what this kind of usage based database architecture looks like in practice for a fintech built on Aurora Serverless style infrastructure.

4.4 Commitment management

Reserved Instances and Savings Plans. Pre committing to AWS resources can save fintechs up to 72% over on demand pricing with Reserved Instances, or up to 66% with Savings Plans.

But static commitment strategies often fail for dynamic workloads. This is particularly relevant for New Zealand fintechs, whose workloads may be more variable due to seasonal patterns or the concentration of their customer base in a single market. Moving from a fixed commitment model to a dynamic, continuously optimised one is where a lot of the remaining savings tend to sit once the obvious rightsizing work is done.

4.5 Data transfer and API optimisation

Fintechs often spend heavily on API usage and inter region data transfer. Useful levers here include:

Amazon CloudFront to cache frequently requested content closer to users
VPC Peering and AWS PrivateLink to cut inter region data transfer fees
API Gateway throttling and caching, which can reduce API usage costs by as much as 40%

For New Zealand fintechs, distance to the nearest AWS region in Sydney adds a data transfer cost that larger, more centrally located markets simply do not face, which makes optimising cross region traffic an unusually high value exercise here.

4.6 The New Zealand cloud market: sovereign alternatives

New Zealand has also seen the rise of sovereign cloud alternatives that claim meaningful cost advantages over the hyperscalers. Providers such as ASI and TEAM Cloud New Zealand offer local hosting with flat, predictable pricing, and report cost savings of 40% to 70% compared with running equivalent workloads on the major international clouds, along with case studies of large public sector repatriation projects delivering substantial savings by moving workloads back from offshore hyperscaler regions.

For fintechs with data sovereignty requirements, these alternatives are worth serious evaluation, though they come with trade offs in service breadth and ecosystem compatibility that need to be weighed carefully against the cost benefit.

5. Organisational practice: culture and governance

5.1 Cost ownership and accountability

Perhaps the single biggest success factor in cloud cost optimisation is establishing clear cost ownership. PicPay, the Brazilian digital finance super app serving more than 60 million customers, has reported meaningful savings by unifying its data platform and giving business units real visibility into the cost impact of their own decisions, including around $10 million in savings tied to smarter, data driven cash back rewards and a 50% reduction in analytics platform costs after consolidating its data infrastructure.

That kind of transformation tends to follow a common pattern: give teams real time visibility into the financial impact of their own choices, and ownership and accountability follow naturally.

For New Zealand fintechs, where teams are often smaller and more cross functional than at global scale players, establishing cost ownership can actually be simpler to set up in practice, but the underlying principle is the same. Every engineer should understand the cost implications of the architectural choices they make.

5.2 Tagging and cost allocation

Implementing tagging and cost allocation from day one is one of the most consistent recommendations across every fintech case study reviewed here. Good tagging gives finance teams real visibility of infrastructure spend, which makes cost easier to allocate, forecast and defend at budget time.

The practical steps that tend to matter most: tag resources from day one, set budgets and alerts, build dashboards teams will actually look at, automate rightsizing and shutdown of idle resources, and run periodic architecture reviews rather than waiting for a crisis to force one.

5.3 Shift left governance

Modern FinOps practice increasingly emphasises shift left governance, embedding cost controls into the DevOps pipeline before resources are ever provisioned. In practice this looks like policy as code for cost guardrails, automated tagging enforcement, budget aware CI/CD pipelines, and cost anomaly detection built into the deployment process rather than discovered a month later on the bill.

New Zealand fintech Wedge's use of Terraform for infrastructure as code is a good local example of how this shift left approach gets embedded into everyday engineering workflow, since infrastructure defined in code makes cost governance part of the same process engineers already use to ship changes.

5.4 Multi cloud governance

As fintechs adopt multi cloud strategies to avoid lock in and gain negotiating leverage, cost management gets more complex, not less. Unified cost visibility tools such as IBM Cloudability exist specifically to give teams a single view of spend across multiple providers, which becomes essential once a business is running production workloads on more than one cloud.

Wellington based digital accounting service Hnry appears to run a genuinely multi cloud stack across AWS and Google Cloud Platform, a strategy that offers flexibility but demands robust FinOps capability to manage well. Hnry's public commentary on engaging infrastructure partners for an AWS Control Tower maturity assessment across security, business continuity and DevOps shows the seriousness with which the company approaches this kind of governance.

5.5 Engineer first approaches

Most fintechs run lean, and a well run cloud setup creates mental margin: less firefighting, more building, so scarce engineering cycles go into features rather than infrastructure triage. Automation is not just a cost saving measure here, it is also a talent retention strategy, because engineers generally prefer building products to manually hunting down idle resources every sprint.

6. Balancing security, compliance and cost

6.1 The integration imperative

Treating security, compliance and cost optimisation as three separate disciplines no longer works. Fintechs need to architect for scalability and compliance from day one, and treat security as a foundation rather than an add on bolted onto the architecture after the fact.

6.2 Where FinOps meets SecOps

An emerging practice worth watching is the integration of FinOps with SecOps. Security monitoring tools can be used to correlate cost anomalies with potential security events, since the two often move together: an unexpected cost spike can be an early signal of a compromised resource, and security incidents almost always carry cost implications of their own.

6.3 Avoiding the false trade off

A useful insight from across the industry is that security and cost optimisation are not fundamentally in tension with each other. Cost optimisation must never undermine a fintech's security and regulatory posture, but effective spend management has to go hand in hand with security monitoring, continuous compliance and granular access control, not sit apart from it.

The false trade off tends to appear when organisations treat cost cutting as an emergency measure rather than an ongoing discipline. When central teams cut too aggressively and without consultation, business units can end up bypassing controls and spinning up unsanctioned cloud resources of their own, which creates both a cost problem and a security problem at the same time.

7. Global case studies: fintechs leading the way

7.1 Nubank: scale and efficiency

Nubank is one of the largest digital financial platforms in the world, serving around 94 million customers across Brazil, Colombia and Mexico. Its cost optimisation programme has been multi pronged: migrating critical workloads to AWS Graviton processors for a 14% cost reduction and a 21% cut in carbon emissions, shifting from a small number of massive AWS accounts toward a multi account architecture that isolates environments and enables far more granular cost management, and adopting more dynamic, automated approaches to commitment management rather than relying on static, once a year negotiated discounts.

7.2 PicPay: from growth at all costs to efficiency

PicPay has moved through a genuine shift in organisational philosophy, from an all in focus on top line growth toward a steadier focus on efficiency. Consolidating its data infrastructure onto a single platform delivered around $10 million in savings from smarter, data driven cash back rewards, cut analytics and visualisation platform costs by 50%, and lifted the number of employees able to self serve their own business analytics by 25%, freeing up engineering time that had previously gone into stitching disparate tools together.

7.3 KOHO and StormForge: EKS Auto Mode in practice

Canadian fintech KOHO provides no fee spending and savings accounts, cash back and credit building tools to Canadians who have been underserved by traditional banks. By migrating to Amazon EKS Auto Mode, KOHO cut infrastructure costs by 30%, achieved five times faster service scaling, and reduced the time to scale new services from around ninety seconds down to five, all while meeting the stringent banking security and compliance requirements its regulator expects. Observability company StormForge reported comparable results from the same underlying technology, a 30% to 40% reduction in infrastructure costs and a 65% cut in the time spent on cluster upgrades and maintenance.

7.4 Graviton migrations across payments and fintech

The pattern seen at Nubank repeats itself widely across the sector. Techcom Securities, a Vietnamese digital wealth management platform, cut overall compute costs by up to 30% after migrating to Graviton, while Paytm achieved a 35% reduction in its payment gateway's EC2 spend alongside a 47% improvement in throughput. These results are consistent enough across independent organisations that Graviton migration has become one of the lowest risk, highest confidence levers available to any fintech running meaningful compute workloads on AWS.

8. New Zealand case studies: local lessons for global challenges

8.1 Dosh: cloud native banking from day one

Dosh, Auckland's first locally owned digital wallet and aspiring digital bank, is progressing toward registration as a bank with the Reserve Bank of New Zealand. Rather than building custom banking infrastructure from scratch, Dosh partnered with Pismo, the cloud native core banking and payments platform Visa acquired in 2024 for around a billion US dollars.