Main content
Reference

FinOps glossary: cloud cost and FinOps terms explained

Plain-English definitions of 45 terms you will meet in cloud cost management and FinOps, from allocation and showback to effective cost, commitment coverage, GPU utilization and FOCUS, with links to deeper guides.

Last updated: · 9 min read

A

Allocation (cost allocation)

Assigning cloud cost to the team, product, environment or customer that caused it, using tags, labels, account structure or rules. Allocation is the base layer for showback, chargeback and unit economics. Multi-cloud cost allocation guide.

Amortized cost

Cost with upfront and recurring commitment fees spread across the usage period they cover, so a one-year upfront reservation shows as a steady monthly cost instead of a single spike.

Anomaly detection

Automatically flagging spend that departs from its normal pattern, so a runaway workload is caught in days rather than at month end. Cloud cost anomaly detection guide.

Autoscaling

Adding or removing capacity automatically as demand changes. It controls cost only when minimum sizes and scale-down rules match real demand.

B

Billed cost

The amount that appears on the invoice for a charge. In the FOCUS specification this is the BilledCost column, used to reconcile to invoices. FOCUS explained.

Blended vs unblended cost

On AWS, unblended cost is the actual rate charged for each usage line, while blended cost averages rates across the accounts in an organization. Use unblended or amortized views for accuracy.

Budget

A spending limit for a scope and period, with alerts at set thresholds so owners hear about overspend before it happens.

C

Chargeback

Billing internal teams for the cloud cost they incur, so it lands on their budget. It requires allocation that teams trust.

Cloud waste

Spend that delivers no value: idle or orphaned resources, oversized capacity, unused commitments and non-production environments running when nobody uses them. Flexera's 2026 State of the Cloud report estimates it at 29% of IaaS and PaaS spend (a self-reported figure). Cloud waste: where it hides.

Commitment coverage

The share of eligible usage that is covered by commitments such as Savings Plans, Reservations or committed use discounts. Low coverage on steady workloads means paying list prices.

Commitment utilization

The share of a commitment you actually use. Low utilization means you pay for capacity you are not using. Track it alongside coverage.

Committed use discount (CUD)

Google Cloud's one- or three-year commitment in exchange for lower prices, available as resource-based commitments for Compute Engine and spend-based commitments across several services. Google Cloud cost optimization.

Container cost allocation

Splitting the cost of a shared Kubernetes cluster across namespaces, workloads and teams, based on requested or used resources. Kubecost vs OpenCost.

Contracted cost

In FOCUS, the cost after your negotiated discounts, before commitment amortization. Useful for comparing contract terms.

Crawl, walk, run

The FinOps Foundation's maturity model: begin small, then grow each capability as the business value justifies the effort. What is FinOps?.

D

Data egress

Data leaving a cloud provider's network, or crossing regions and sometimes zones, which providers charge for. A common source of surprise bills.

E

Effective cost

Billed cost with commitment fees amortized across the resources that used them. Use it for allocation and unit economics, and billed cost to reconcile invoices.

F

FinOps

An operational framework and cultural practice for maximizing the business value of technology, with financial accountability shared between engineering, finance and business teams, as defined by the FinOps Foundation. What is FinOps?.

FinOps for AI

Applying FinOps practice to AI spend: LLM tokens, managed model services and the GPU infrastructure behind them, with unit metrics tied to business value. FinOps for AI.

FinOps Framework

The FinOps Foundation's reference model: principles, three iterative phases (Inform, Optimize, Operate), capabilities, personas and the crawl, walk, run maturity model.

FOCUS

The FinOps Open Cost and Usage Specification, an open standard that defines common columns and semantics for billing data across cloud and SaaS providers. FOCUS explained.

Forecasting

Predicting future spend from history and known changes. Forecast accuracy is usually tracked as an error percentage such as MAPE. FinOps KPIs.

G

GPU utilization

How busy provisioned GPUs actually are. Because GPUs cost dollars an hour, idle GPU time is among the most expensive waste in AI infrastructure. Idle GPU cost.

Graviton

AWS's Arm-based processors. AWS states Graviton instances cost up to 20% less than comparable x86 instances, though workloads must be compatible.

I

Idle resource

A resource that is running and billing but doing little or no useful work, such as a load balancer with no traffic or an instance with near-zero CPU for weeks.

Inform, Optimize, Operate

The three phases of the FinOps Framework: gain visibility and allocation, improve efficiency, then embed the practice in how the organization runs.

L

Label and tag

Key-value metadata attached to resources (labels on Google Cloud and Kubernetes, tags on AWS and Azure) used to attribute cost to owners. Tags are usually not retroactive.

List cost

In FOCUS, cost at public list price, before discounts. Comparing it with contracted and effective cost shows how deep your discounts are.

M

Multi-cloud

Using more than one cloud provider. It raises the value of a normalized cost view because every provider bills in a different vocabulary. What is a cloud management platform?.

N

NAT gateway

A managed service that lets private subnets reach the internet. On AWS it bills hourly plus per GB processed, so heavy traffic through it is a common cost driver. Hidden AWS savings.

O

On-demand pricing

Paying the standard rate with no commitment. Flexible, but the most expensive way to run steady workloads.

Orphaned resource

A resource that outlived what it was created for, such as an unattached volume, a snapshot of a deleted volume or an unused IP address.

P

Prompt caching

Reusing a repeated prompt prefix across LLM requests at a reduced price for the cached portion, which can cut input token cost substantially for repetitive prompts. FinOps for AI.

R

Reserved Instance (RI)

A one- or three-year commitment to specific capacity in exchange for a discount versus on-demand pricing, offered in some form by the major clouds. Savings Plans vs RIs vs Spot.

Rightsizing

Matching resource size to actual need, such as moving an over-provisioned instance to a smaller type based on observed utilization.

S

Savings Plan

A commitment to a consistent hourly spend over one or three years in exchange for lower rates. AWS lists savings of up to 66% for Compute Savings Plans and up to 72% for EC2 Instance Savings Plans; Azure lists up to 65% for its compute savings plan. Figures are maximums. Savings Plans vs RIs vs Spot.

Shared cost

Cost that benefits many teams, such as networking, monitoring, support plans and idle cluster capacity, which needs an agreed rule to split fairly.

Showback

Showing teams the cloud cost they cause without billing them. It builds awareness and trust in the numbers and is a common step before chargeback.

Spot instance

Discounted spare cloud capacity that the provider can reclaim at short notice. Google lists Spot VM discounts of up to 91%. Best for fault-tolerant, interruptible work.

T

Tag coverage

The share of resources, or of cost, that carries the required tags. The leading indicator of allocation quality.

Token (LLM)

The unit in which language models count input and output text, and usually the unit providers bill in. Cost scales with tokens per request and request volume.

U

Unit cost (unit economics)

Cloud cost divided by a business output such as customers, transactions or resolved tickets. It shows whether spend is growing faster than the value it supports. FinOps KPIs.

W

Waste rate

Identified waste as a share of total spend for a scope. Track your own trend before comparing with industry benchmarks.

Workload

An application or service and the resources it uses. Cost reporting by workload is more useful to engineers than cost by service.

Z

Zombie resource

Informal term for an orphaned or forgotten resource, often non-production, that keeps running and billing with no owner.

Frequently asked questions

What does FinOps stand for?

FinOps is a blend of Finance and DevOps. It describes an operational framework and cultural practice, stewarded by the FinOps Foundation, for managing technology spend with shared accountability between engineering, finance and business teams.

What is the difference between showback and chargeback?

Showback reports the cost each team causes without billing them. Chargeback bills the cost to the team's budget. Most organizations start with showback to build trust in the data and move to chargeback once allocation is reliable.

What is the difference between effective cost and billed cost?

Billed cost is what the invoice shows for a charge. Effective cost spreads commitment fees across the usage they cover. Use billed cost to reconcile to invoices and effective cost to allocate and measure unit economics.

Sources

  1. FinOps Framework — FinOps Foundation
  2. FOCUS specification — FinOps Foundation
  3. State of FinOps 2026 — FinOps Foundation, 2026
  4. Compute Savings Plans pricing — Amazon Web Services
  5. AWS Graviton — Amazon Web Services
  6. Optimize your cloud investment with Cost Management — Microsoft Learn
  7. Spot VMs — Google Cloud documentation
  8. Flexera 2026 State of the Cloud findings — Flexera, 2026

Provider pricing and discount figures were checked against the linked documentation on 7 October 2026. Providers change pricing and programs regularly; confirm current terms before making purchasing decisions.

Put these terms to work on your own bill

Connect a cloud account and see allocation, waste, commitments and forecasts for your environment. Free trial, no credit card.