Loading
Skip to content
Miracle Minds AI
  • Try “HIPAA”, “model risk”, “FinOps” or “agentic”.

Managed Cloud, Infra & FinOps

Cloud cost and reliability are both decided at the moment of change: what gets deployed, at what size and under whose budget. A managed service that only reports the bill after month end is too late. Engineers have to own the cost of what they build, with the operator holding the guardrails.

Part of Cloud, Infrastructure & Applications, in the Run mode. 9 sources, checked .

estimated share of cloud spend wasted, rising for the first time in five years (Flexera 2026)
29%
estimated share of cloud spend wasted, rising for the first time in five years (Flexera 2026)
Flexera, 2026(opens in a new tab)
of FinOps practitioners now manage AI spend, up from 31% two years earlier (State of FinOps 2026)
98%
of FinOps practitioners now manage AI spend, up from 31% two years earlier (State of FinOps 2026)
FinOps Foundation, 2026(opens in a new tab)
of operators include systemic third-party risk in their resiliency assessments (Uptime 2026)
39%
of operators include systemic third-party risk in their resiliency assessments (Uptime 2026)
Uptime Institute Intelligence, 2026(opens in a new tab)

The problem

Cloud waste has stopped falling. Flexera's 2026 State of the Cloud survey of more than 750 cloud decision-makers and users found that estimated wasted cloud spend rose to 29%, the first increase in five years, as cloud-based AI workloads surged1. Managing cloud spend remains a top challenge for 85% of respondents, even though 63% now have established FinOps teams and 71% run a Cloud Center of Excellence1.

AI has widened the job. The FinOps Foundation's State of FinOps 2026, with 1,192 respondents representing more than $83 billion in annual cloud spend, found that 98% now manage AI spend, up from 31% two years earlier2. The remit has spread to SaaS, licensing, private cloud and data centers. Practitioners also report diminishing returns from cloud optimization: the big rocks of waste are gone, and governance, forecasting and influencing decisions before commitments are made now matter more2.

Running cloud well is a resilience question too. In Uptime Institute's 2026 resiliency survey, 24% of respondents said public cloud is not resilient enough for any of their mission-critical workloads, and software or configuration errors were the most common cause of major third-party outages4. Only 39% include systemic third-party risk in their resiliency assessments4. For banks, outsourcing does not move the accountability: US agencies state that using a third party does not diminish a bank's responsibility5.

Good looks like the FinOps Framework in daily operation: engineers own the cost of what they build from architecture design through operations, cost data is accessible, timely and accurate, and FinOps is enabled centrally3. The operator runs the estate to agreed service levels with controlled, recorded change, keeps the evidence your examiners will ask for, and reports cost per product or unit of business value instead of per account.

What you get

  1. 01

    Operations to agreed service levels

    Monitoring, incident response, patching, backup and capacity management for your cloud and infrastructure, with service levels and reporting you can audit.

  2. 02

    Cost allocation and showback

    Tagging standards, allocation rules and cost data in the FinOps Open Cost and Usage Specification (FOCUS) format, so every dollar maps to a product, team or cost center.

  3. 03

    Continuous optimization

    Rightsizing, scheduling, commitment and discount management, and regular waste reviews, with savings tracked against an agreed baseline and confirmed with finance.

  4. 04

    AI and GPU cost governance

    Budgets, quotas and unit cost tracking for model APIs, AI platforms and GPU capacity, so AI spend is visible before it surprises finance.

  5. 05

    Controlled change and audit evidence

    Infrastructure as code with reviewed and recorded changes, plus evidence packs mapped to your control framework and your regulators' third-party expectations.

Where it goes wrong

The failure modes we design against from the first week, and the control for each.

  • Reporting the bill instead of changing behavior

    A report after month end does not change an engineering decision. The FinOps Framework pushes cost ownership to engineers from architecture design onward3. Put cost in front of teams where they work, with a budget per product and alerts on anomalies.

  • Treating AI spend as a separate problem

    AI workloads are where estimated waste is now growing1. Bring GPU, model API and AI platform costs into the same allocation, budgeting and unit cost process as the rest of cloud from the first pilot.

  • Outsourcing operations and the accountability with it

    US banking agencies hold you responsible for outsourced activities as if you performed them in-house5, and their September 2026 proposal keeps that principle6. Contract for audit rights, incident notice, exit support and evidence, then check performance against them.

  • Uncontrolled change in a managed estate

    Software and configuration errors are the leading cause of major third-party outages4. Apply configuration change control with review, records and rollback to every change, whether your staff or the operator makes it7.

Standards and rules

  • FinOps Framework (FinOps Foundation)

    Primary text on finops.org(opens in a new tab)

    Sets the principles and capabilities for cost ownership, allocation and optimization; use it to define what the managed service covers and how maturity is judged.

  • Interagency Guidance on Third-Party Relationships: Risk Management (2023)

    Primary text on govinfo.gov(opens in a new tab)

    US banking agencies expect risk-based oversight of a managed provider across planning, due diligence, contracting, monitoring and termination.

  • EU Digital Operational Resilience Act, Regulation (EU) 2022/2554

    Primary text on eiopa.europa.eu(opens in a new tab)

    EU financial entities must manage ICT third-party risk, include key contractual provisions and keep a register of information on ICT providers; your operator must support all of it.

  • FFIEC IT Examination Handbook, Architecture, Infrastructure, and Operations booklet

    Primary text on fdic.gov(opens in a new tab)

    Examiners use it to assess how you govern architecture, infrastructure and operational controls, so the managed service must produce evidence against it.

How to measure it

  • Share of cloud and AI spend allocated to a named owner
  • Unit cost per transaction, customer or model call, trended monthly
  • Forecast accuracy against actual spend
  • Realized savings against baseline, confirmed by finance
  • SLO attainment and change fail rate for the managed estate

Questions buyers ask

What is FinOps and how is it different from cost cutting?

FinOps is an operating practice in which engineering, finance and business teams share responsibility for technology spend and make decisions on value, not just price. Cost cutting is a one-off exercise. FinOps is continuous: allocate every cost to an owner, show teams their spend while they can still change it, optimize what is wasteful, and measure unit cost against the business outcome the spend supports.

What should a managed cloud contract include for a regulated firm?

At a minimum: service levels tied to business impact, incident notification timelines, audit and access rights for you and your regulators, data location and security obligations, subcontractor disclosure, change control and evidence requirements, and exit support with a tested transition plan. Banking regulators expect you to oversee the provider as if the work were done in-house, so the contract has to give you the information to do that.

How do we control the cost of AI workloads?

Treat AI like any other cloud spend, only earlier. Tag and allocate GPU, model API and AI platform costs from the first pilot. Set budgets and quotas per team and use case. Track unit cost, such as cost per request or per resolved case, alongside the business value. Review model choice and capacity commitments regularly, because pricing and utilization change quickly and idle GPU capacity is expensive.

Sources

Every figure on this page comes from the sources below. We checked each link on .

  1. 1.Flexera Finds Cloud Value is Rising While AI Waste Grows(opens in a new tab)Flexera, March 18, 2026 · flexera.com
  2. 2.State of FinOps 2026 Report(opens in a new tab)FinOps Foundation, February 2026 · data.finops.org
  3. 3.FinOps Principles(opens in a new tab)FinOps Foundation · finops.org
  4. 4.Annual outage analysis 2026: The causes and impacts of IT and data center outages(opens in a new tab)Uptime Institute Intelligence, May 2026 · datacenter.uptimeinstitute.com
  5. 5.Interagency Guidance on Third-Party Relationships: Risk Management (88 FR 37920)(opens in a new tab)Federal Reserve, FDIC and Office of the Comptroller of the Currency, June 9, 2023 · govinfo.gov
  6. 6.Proposed Third-Party Risk Management Guidance(opens in a new tab)OCC, Federal Reserve, FDIC and NCUA (Federal Register), September 15, 2026 · federalregister.gov
  7. 7.SP 800-53 Rev. 5, Security and Privacy Controls for Information Systems and Organizations(opens in a new tab)National Institute of Standards and Technology, December 10, 2020 · csrc.nist.gov
  8. 8.Digital Operational Resilience Act (DORA)(opens in a new tab)European Insurance and Occupational Pensions Authority (EIOPA) · eiopa.europa.eu
  9. 9.Updated FFIEC IT Examination Handbook: Architecture, Infrastructure, and Operations Booklet (FIL-47-2021)(opens in a new tab)Federal Deposit Insurance Corporation, June 30, 2021 · fdic.gov