Last verified April 2026
APRIL 2026 — CONSUMER BILL AUDIT EDITION — DIGITAL SIGNET — FOR FINANCE TEAMS
Cloud cost optimisation: a FinOps playbook for AWS, Azure, and GCP in 2026.
By Oliver, Digital Signet — Last verified April 2026
12 min readAverage cloud waste: 27% of cloud spend
Source: Flexera 2024 State of the Cloud Report. Industry average across survey respondents. Varies from 10-40% depending on FinOps maturity.
The Flexera 2024 State of the Cloud Report's 27% waste estimate means that a $1M/month cloud bill contains approximately $270,000 per month in spending that generates no business value. At the portfolio-level, the cloud industry as a whole wastes an estimated $150-200 billion per year. The FinOps Foundation's State of FinOps 2024 survey found that cloud waste reduction was the top FinOps priority for the third consecutive year, up from the second position in 2023.
The 27% figure should be treated as a rough industry benchmark rather than a precise measurement. Organisations with mature FinOps practices typically report 10-15% waste; organisations with no FinOps practice can see 35-40%. A new FinOps engagement that takes a $200k/month bill from 35% waste to 15% waste saves $40k/month, or $480k/year, which is typically well above the cost of any FinOps tool or consultant engagement.
The seven cloud waste categories
Dev/test instances running 24/7, sandbox environments, terminated-project resources still active
Instance sizes chosen for peak load, running at 5-15% utilisation 90% of the time
RIs and Savings Plans that lapsed; account silently reverted to on-demand without anyone noticing
Unattached EBS volumes, orphaned snapshots, S3 Standard for archival data that should be Glacier
Cross-AZ/region data transfer, duplicate logging, over-replicated data
Enterprise Support tier overkill, marketplace software on unused instances
Illustrative breakdown based on FinOps Foundation and industry practitioner reports. Not from a single study.
AWS-specific playbook
Cost Explorer + AWS Compute Optimizer
AWS Cost Explorer provides cost and usage data with 12+ months of history. Compute Optimizer uses ML to identify over-provisioned EC2 instances, Lambda functions, and EBS volumes and recommends right-sized alternatives. Enable Compute Optimizer in the console and run the first report within the first week of a FinOps engagement. Typical right-sizing opportunity: 20-30% of EC2 spend.
Savings Plans vs Reserved Instances
For most modern workloads, Compute Savings Plans are preferred over EC2 RIs because they apply across instance families, sizes, and regions. Commit to 1-year plans initially to preserve flexibility; move to 3-year plans for stable baseline workloads once you have 6+ months of usage data. Start coverage at 70-80% of your steady-state compute baseline, not 100%, to avoid over-committing on workloads that may shrink.
Spot Instances for batch and dev/test
Spot Instances offer 60-90% discount over on-demand for workloads that can be interrupted. Batch jobs, CI/CD runners, machine learning training, and dev/test environments are strong candidates. Use Spot Fleet or Spot with capacity-optimised allocation strategy. AWS's Spot Instance Advisor shows historical availability and interruption frequency by instance type and region.
S3 Intelligent-Tiering and lifecycle policies
S3 Standard storage at $0.023/GB/month becomes S3 Glacier at $0.004/GB/month for data not accessed in 90+ days. Set S3 Intelligent-Tiering on all buckets that contain data with irregular access patterns. Set explicit lifecycle policies on backup and log buckets to transition to Glacier-IR after 30 days and delete after 365 days. This single change can reduce storage costs by 40-70% on mature accounts with significant log accumulation.
FinOps tool map (independent assessment)
| Tool | Best for | Honest assessment |
|---|---|---|
| Vantage | Engineering teams, multi-cloud | Strong UX, unified cloud + SaaS billing, resource-level tagging. Less strong on enterprise finance workflows. |
| CloudZero | Finance-led FinOps, unit economics | Best unit-economics and showback/chargeback features. Engineering UX is less polished than Vantage. |
| ProsperOps | Autonomous RI/SP management | Fully automated commitment management. Strong if you are tired of managing RIs manually. Narrow scope outside of compute commitments. |
| Apptio Cloudability | Large enterprise, IT finance | Most mature enterprise feature set. Complex implementation. Often overkill for under $500k/mo spend. |
| Harness Cloud Cost | Teams already on Harness platform | Good if you use Harness for CI/CD; less compelling as a standalone FinOps tool. |
| Zesty | Automated right-sizing | Continuous EC2 and EBS right-sizing engine. Complements rather than replaces a FinOps platform. |
Digital Signet has no affiliate relationship with any FinOps platform. This table represents our independent assessment as of April 2026.
FinOps maturity: crawl, walk, run
The FinOps Foundation's maturity model describes three stages. Crawl: basic cost visibility, resource tagging, a shared dashboard, initial anomaly alerting, and a first pass at Reserved Instance or Savings Plan coverage. This is achievable in 30-60 days with any team that has billing data access. Walk: an established FinOps practice with regular review cycles (weekly team cost sync, monthly cross-functional review), showback or chargeback to engineering teams, and active commitment management. Run: automated optimisation, forecasting embedded in sprint planning, unit economics by feature or customer, and continuous improvement loops in the engineering process.
Most organisations entering FinOps are in Crawl. The goal of the first 90 days is to reach Walk. The most common failure mode is attempting to skip Crawl and implement a full Run-stage FinOps practice without the underlying visibility and accountability structure. Without visibility, there is no accountability; without accountability, optimisation is one-time rather than continuous.
Cloud creep and consumer bill creep share the same root cause: lack of systematic review. The Xfinity Broadcast TV Fee and the expired AWS Reserved Instance are two expressions of the same principle. See the bill anaesthesia essay for the behavioural economics of why both go unnoticed.
AI inference is the newest cloud cost vector. As enterprises integrate LLM APIs into their products, context-window costs and token usage can grow faster than any other cloud line item. See contextcost.com for the AI-context-window cost story.
Drowning in SaaS and telecom creep?
Digital Signet runs two-week bill-and-subscription audits. We map every line item, score every contract, and deliver the memo your CFO will sign.
See the engagement formatFrequently asked questions
How much of an AWS bill is typically wasted?+
The Flexera 2024 State of the Cloud Report estimates organisations waste an average of 27% of their cloud spend. Organisations with mature FinOps practices report 10-15% waste; those with no FinOps practice can see 35-40%. The FinOps Foundation found cloud waste reduction was the top FinOps priority for the third consecutive year.
What is the difference between Reserved Instances and Savings Plans?+
Reserved Instances are commitments to a specific instance type in a specific region for 1 or 3 years (40-75% discount). Savings Plans are flexible dollar-amount commitments that apply across instance families (Compute Savings Plans, up to 66% discount). For most modern workloads using containerised or autoscaled compute, Savings Plans are preferred because they accommodate instance type changes during the commitment period.
At what cloud spend level does FinOps tooling become worth it?+
Under $50k/mo: native tools (AWS Cost Explorer, Azure Cost Management) plus disciplined spreadsheet review are usually sufficient. $50k-$200k/mo: a lightweight tool like Vantage or CloudZero typically pays for itself within 1-2 months. Over $200k/mo: a full FinOps platform and often a dedicated practitioner become cost-justified.
What is the FinOps maturity model?+
The FinOps Foundation describes three stages: Crawl (basic cost visibility, tagging, initial commitment coverage, achievable in 30-60 days), Walk (established review cycles, showback/chargeback to teams, active commitment management), and Run (automated optimisation, forecasting, unit economics embedded in engineering). Most organisations entering FinOps are in Crawl; the first 90-day goal is to reach Walk.
Is cloud creep the same as bill creep in consumer services?+
Structurally, yes. Reserved Instances expire and accounts revert to on-demand pricing (like a cable promo expiring). Dev instances run 24/7 (like a gym membership on autopay). Orphaned snapshots accumulate (like forgotten software subscriptions). The mechanism is identical; the dollar scale is typically 10,000x larger.