9. Four genuine gaps remain in the FinOps market, and three of them affect financial services disproportionately

The situation

The FinOps tool market now has 40+ vendors (FinOps Foundation landscape). Cloud-native tools are free and improving rapidly. Enterprise platforms cover governance and reporting. Kubernetes-specific tools handle container cost allocation. The visibility problem that defined early FinOps is largely solved for public cloud. What remains unsolved is harder: bridging visibility to action, normalising hybrid environments, governing AI costs, and embedding regulatory constraints into cost decisions.

The complication

Financial services organisations face these gaps more acutely than technology companies. A SaaS startup running 100% on AWS with a single product has a FinOps problem that existing tools solve well. A UK bank running hybrid infrastructure across three public clouds, on-premises data centres, HPE GreenLake private cloud, 100+ SaaS subscriptions, and regulated workloads across multiple jurisdictions has a FinOps problem that no tool fully addresses.

Gap 1: IDE-level real-time cost feedback

Current state: Infracost estimates costs at PR time. It has a VS Code extension that runs on save, not inline. There is no tool that shows cost implications as an engineer types Terraform code the way a linter shows syntax errors.

The gap is between cost-at-PR-time and cost-at-typing-time. During that gap, engineers commit to architectural decisions that embed cost for months. A developer choosing between a NAT gateway and a VPC endpoint does not see the $480/month difference until the PR is raised. By then, the architecture is set, the tests are written, and the engineer has moved on to the next feature. The switching cost at that point is measured in days of rework, not the 30 seconds it would take to choose differently at typing time.

The technical barrier is latency. Pricing API calls take 200-500ms. Inline feedback needs to be sub-100ms to feel responsive. Caching solves some of this (instance pricing changes infrequently), but dynamic costs like data transfer depend on configuration context that changes with every line. The first vendor to solve this will own the shift-left FinOps category.

Financial services impact: medium. This gap affects all organisations equally. A bank has no more or less need for IDE-level feedback than a technology company. The opportunity is real but not sector-specific.

Gap 2: Regulatory-aware FinOps for multi-jurisdiction banks

Current state: No tool combines cost allocation tags with data residency metadata and regulatory compliance requirements. A workload that is cheapest in us-east-1 may legally need to run in eu-west-2 or ap-southeast-1 at 15-20% higher cost. No FinOps platform includes a regulatory filter on its optimisation recommendations.

This matters because every cost recommendation that ignores regulatory constraints is at best useless and at worst dangerous. If a FinOps tool recommends moving a workload to us-east-1 to save $2,000 per month, and the data in that workload is subject to GDPR residency requirements, acting on that recommendation creates a compliance violation. The cost of a PRA/FCA enforcement action or a DORA penalty exceeds any cloud savings by orders of magnitude.

The data model required is complex. Each workload needs cost metadata (current region, current spend, optimised region, optimised spend) alongside regulatory metadata (data classification, applicable regulations, permitted regions, required certifications). No cloud provider or FinOps vendor maintains this regulatory layer because it varies by organisation, jurisdiction, and data type. Building it requires deep financial services domain knowledge alongside cloud cost expertise.

The competitive field is empty. CoreStack has compliance-aware governance but does not tie compliance metadata to cost optimisation recommendations. Apptio has no regulatory layer. CloudHealth's policy engine can enforce region restrictions but cannot calculate the cost premium of compliance. The gap is not a missing feature in an existing product. It is a missing product category.

A viable product in this space would need three components: a regulatory rules engine (which regulations apply to which data classifications in which jurisdictions), a cost comparison engine (what does this workload cost in each compliant region versus the cheapest non-compliant region), and a recommendation engine that only surfaces cost optimisations within the compliant region set. The regulatory rules engine is the hard part because regulations change, interpretations vary, and each bank's compliance team has its own view of what is permissible.

Standard Chartered operates across 60 markets. HSBC operates in 62 countries. Barclays spans UK, US, and Asia. For these banks, every optimisation recommendation needs a regulatory overlay. The vendor that solves this will own the financial services FinOps niche, which is large enough to sustain a company.

Financial services impact: critical. This gap is sector-specific and urgent.

Gap 3: AI/LLM cost governance across multiple providers

Current state: 98% of FinOps practices now manage AI spend (State of FinOps 2026), up from 31% in 2024. CloudZero has early AI cost tracking. Helicone and Portkey provide LLM observability with token tracking and latency monitoring. But no tool provides unified LLM costs across Bedrock (AWS), Azure OpenAI, and Vertex AI (GCP) with per-team attribution and budget controls.

The problem is structural. LLM pricing is per-token, not per-resource. A team that provisions an m5.xlarge shows up in AWS billing as a clear line item attributable to a specific account and tag. A team that makes 100,000 API calls to Azure OpenAI shows up as aggregate token consumption that may or may not be attributable to a specific team depending on how API keys are organised. The attribution model that works for infrastructure does not work for API-based AI services.

The cost unpredictability compounds the governance challenge. Infrastructure costs are relatively stable: a running instance costs the same per hour regardless of what it does. LLM costs vary by model, by prompt length, by response length, and by whether the output is cached. A single query to GPT-4 can cost 100x more than the same query to GPT-3.5-Turbo. Without per-team token budgets and model-level tracking, AI spend grows without governance until the quarterly bill surprises the CFO.

Lloyds has 300+ data scientists on Vertex AI. HSBC is migrating analytics to BigQuery with Vertex AI integration. Barclays is deploying Copilot to 50,000+ colleagues. JPMorgan's LLM Suite serves 200,000+ employees. Every one of these deployments is generating AI costs that are harder to attribute, harder to forecast, and harder to optimise than traditional cloud spend.

Financial services impact: high and growing rapidly. This gap will be critical within 12-18 months.

Gap 4: Private cloud + public cloud normalised unit economics

Current state: HSBC's Natalie Daley described this exact problem at FinOps X 2023. No tool normalises costs across public cloud, private cloud (HPE GreenLake, VMware), SaaS, and on-premises into a common cost-per-transaction metric. TBM (via Apptio) provides partial coverage but requires significant manual data integration for anything beyond public cloud.

The problem is that each infrastructure type uses different cost models. Public cloud costs are variable and consumption-based. Private cloud costs are fixed or semi-fixed (you buy the hardware, you pay for the licence, you amortise over 3-5 years). On-premises costs include real estate, power, cooling, and staff that are allocated using activity-based costing models. SaaS costs are per-user or per-feature.

Normalising these into a single cost-per-transaction metric requires converting each cost model into a comparable unit. What does it cost per transaction to run a workload on GCP versus HPE GreenLake versus an on-premises data centre? The answer determines workload placement strategy. Without the answer, placement decisions are made on availability and familiarity rather than cost efficiency.

Barclays has 100,000+ workloads on HPE GreenLake private cloud alongside AWS, Azure, and GCP. Lloyds is building cloud-native core banking on Thought Machine while maintaining legacy on-premises systems. Every UK bank runs hybrid infrastructure. Without normalised unit economics, executives cannot answer the fundamental question: is our technology investment efficient?

Financial services impact: critical. Every UK bank faces this problem. The vendor that solves it will have a significant first-mover advantage in financial services.

Why these gaps persist

The four gaps share a common root cause: the FinOps tool market was built by and for cloud-native technology companies. A SaaS startup running 100% on AWS with 50 engineers and a single product has a FinOps problem that the current tool market solves well. The startup has one cloud provider, one billing system, no regulatory constraints on region placement, no hybrid infrastructure, and predictable workload patterns. Every tool in the market was designed for this profile.

Financial services organisations look nothing like this. They run three or four cloud providers plus private cloud plus on-premises. They operate under regulatory constraints that limit where workloads can run. They outsource operations to system integrators with different incentive structures. They manage hundreds of SaaS applications alongside cloud infrastructure. They deploy AI at scale without purpose-built cost governance tools. The gap between what the market provides and what financial services needs is structural, not incremental.

The startup opportunity is real but narrow. Gap 2 (regulatory-aware FinOps) and Gap 4 (hybrid unit economics) require deep financial services domain knowledge that most cloud-native startup founders lack. Gap 3 (AI cost governance) has the broadest addressable market because it affects technology companies as much as financial services. Gap 1 (IDE-level cost feedback) is most likely to be absorbed by Infracost as a feature extension rather than creating a new company.

For organisations that cannot wait for the market to fill these gaps, the interim solution is custom tooling. Build a regulatory metadata layer in your existing FinOps platform. Build AI cost attribution using CloudWatch metrics and custom dashboards. Build hybrid unit economics using Apptio TBM with manual data integration for private cloud. These custom solutions are expensive to build and maintain, but they are the only option until purpose-built tools exist.

Market whitespace validation