Committed use discountCUD
A discount a provider gives you for committing to a minimum spend or usage level for a fixed term.
Read the full definition →Glossary
Cloud & AI Cost Glossary
Answer-first definitions for the cost terms that turn up on your AWS bill, your Anthropic invoice, and every FinOps review in between.
Last updated 50 terms
Cloud cost
The pricing models, cost-accounting bases, and waste patterns behind an AWS-style bill.
Reserved instanceRI
A commitment to one instance type in one region or zone, priced below on-demand.
Read the full definition →Savings planSP
A commitment to a dollar-per-hour spend level. It applies automatically across compute, whatever the instance family.
Read the full definition →On-demand vs. spot instances
On-demand pricing charges the full published rate with no commitment. Spot pricing sells unused capacity at a steep discount, but the provider can reclaim it on short notice.
Read the full definition →Data egress
The fee a cloud provider charges for data that leaves its network. On most providers, this includes transfer between availability zones.
Read the full definition →Amortized, blended, and unblended cost
Three bases for reporting the same spend. Unblended is the raw rate per resource. Blended averages a rate across accounts. Amortized spreads upfront commitment costs over their term.
Read the full definition →Billed vs. effective vs. list vs. contracted cost
The four cost columns FOCUS defines for one charge: cash-basis, amortized, pre-discount, and negotiated-rate. This is why two dashboards can quote two numbers for the same month.
Read the full definition →Rightsizing
Matching a resource's provisioned size to the usage you observe, not to the peak it was first sized for.
Read the full definition →Idle and orphaned resources
Resources that keep running and billing while they do no useful work. Idle resources are underused. Orphaned resources have lost their owner and purpose.
Read the full definition →Storage class
A pricing and durability tier for stored data. The same bytes cost a different amount, and take a different time to read back, depending on the class.
Read the full definition →Lifecycle policy
A rule that moves or deletes stored data on its own after a set condition, most often an age threshold.
Read the full definition →Incremental snapshot
A backup that stores only the blocks changed since the previous snapshot, not a full copy. Deleting one frees far less space than its listed size suggests.
Read the full definition →Allocatable capacity
What is left for pods to request on a Kubernetes node after the kubelet, container runtime, OS, and DaemonSets reserve their share.
Read the full definition →Requests vs. limits
A request is what the scheduler reserves for a container. It decides whether the container fits on a node. A limit is a ceiling that the kubelet enforces at runtime.
Read the full definition →Cluster idle
Node capacity that is schedulable but that no pod requested. It is a bin-packing and autoscaling problem, not a request-sizing one.
Read the full definition →Workload idle
Capacity that a pod requested and had reserved but does not use. It is a request-sizing problem, fixed by sizing to observed usage.
Read the full definition →Bin packing
How tightly a scheduler packs pods onto the fewest nodes, given their requests. Loose packing strands capacity as cluster idle across many half-empty nodes.
Read the full definition →Cost allocation tags
Key-value labels on cloud resources. They let you group spend by team, project, environment, or customer, not see one undifferentiated total.
Read the full definition →Tag coverage
The share of spend, measured in dollars and not in resource count, that carries a required set of tags.
Read the full definition →Unallocated spend
The share of a bill that you cannot attribute to a real owner. The cause is a missing tag, a wrong tag, or spend that cannot carry a tag at all.
Read the full definition →Showback vs. chargeback
Showback reports each team's cloud spend for visibility only. Chargeback bills that spend back to the team's budget.
Read the full definition →Unit economics
Infrastructure cost expressed per customer, per request, or per transaction, not as one company-wide total.
Read the full definition →RI/SP coverage and utilization
Coverage is the share of eligible usage that a commitment discounts. Utilization is the share of the commitment you bought that you actually use.
Read the full definition →Observability tax
The cost of the monitoring stack itself, such as hosts, custom metrics, traces, and logs. It grows on its own, apart from the infrastructure it watches.
Read the full definition →Metric cardinality
The number of unique combinations of metric name and tag value that a monitoring platform stores and bills for. It is not the number of metric names your code emits.
Read the full definition →High-water-mark billing
A billing method that charges close to a resource's peak usage in a period, not its average. A spike that lasts long enough can set the charge for the whole month.
Read the full definition →Log ingest vs. index
Two separate charges that most observability platforms bill independently. Ingest is charged when a log event arrives. Index, or retention, is a recurring charge to keep it searchable.
Read the full definition →Snowflake credit
Snowflake's unit of compute billing. A virtual warehouse uses credits every second it runs, at a rate that doubles with each warehouse size.
Read the full definition →Auto-suspend / resume minimum
Snowflake's default of suspending an idle warehouse to stop billing, then resuming it on the next query. Each resume bills at least 60 seconds, however short the query is.
Read the full definition →BigQuery slot
A unit of query-processing capacity in BigQuery. On-demand queries use a shared pool billed by bytes scanned. An Editions reservation buys dedicated slots billed by the slot-hour.
Read the full definition →Bytes scanned & partition pruning
On-demand BigQuery bills by how much data a query reads, not by how long it runs. Partition pruning and column selection are the two levers that shrink that number.
Read the full definition →Time Travel & Fail-safe
Two Snowflake features that keep changed or deleted data recoverable, and keep billing storage for it. Time Travel covers a configurable window. Fail-safe covers a fixed 7 days after that.
Read the full definition →AI & LLM cost
How usage-based model API spend is metered, priced, and controlled.
Token
The basic unit of text that an LLM API meters and bills by. It is roughly three to four characters of English text, not a whole word.
Read the full definition →Input vs. output tokens
Most LLM APIs price input tokens, the prompt you send, and output tokens, the response generated, separately. Output usually costs several times more.
Read the full definition →Context window
The maximum number of tokens a model can process in one request. It counts the prompt, the conversation history, and the generated output together.
Read the full definition →Prompt caching
A pricing and speed feature that discounts and speeds up tokens repeated across requests. You do not pay the full input rate each time.
Read the full definition →Batch API
An asynchronous request mode. It trades a longer turnaround, from minutes to hours, for a discount against real-time API calls.
Read the full definition →Cost per 1M tokens
The standard unit LLM providers use to publish prices: the dollar cost of one million input or output tokens. A single token is too small a unit to price.
Read the full definition →Per-request cost
The full cost of one API call, shown as one number. It includes input tokens, output tokens, and any tool calls.
Read the full definition →Fine-tuning vs. inference cost
Fine-tuning cost is a one-time or periodic training expense to specialize a model. Inference cost is the ongoing, usage-based cost of running it.
Read the full definition →Model tiering
Routing each request to the cheapest model that can handle it well, not sending every request to the most capable and most expensive model.
Read the full definition →FinOps & process
The habits and metrics that turn a bill into something a team can manage.
FinOps
The practice of managing cloud, and increasingly AI, spend as a shared job between finance and engineering, on an ongoing schedule.
Read the full definition →FOCUSFOCUS
The FinOps Open Cost and Usage Specification. It is an open, vendor-neutral standard for the shape of a provider's cost and usage export. The FinOps Foundation maintains it.
Read the full definition →Cost anomaly detection
Flagging spend that moves far from a service's or workload's own history. It happens automatically, so you do not wait for a person to notice it on the invoice.
Read the full definition →Budget and budget alert
A budget is a target spend ceiling for a scope of cost. A budget alert tells an owner when actual or forecast spend nears or crosses it.
Read the full definition →Cost forecasting
Projecting future spend from the historical trend and known upcoming changes. You use it to catch a budget overrun before it happens, not after.
Read the full definition →Run-rate
Current spend annualized, or projected to month-end, as if the recent pace continued unchanged. It is a quick but naive look forward.
Read the full definition →Total cost of ownershipTCO
The full cost of a system or vendor decision over its useful life. It includes costs beyond the sticker price: migration, integration, ongoing operation, and eventual replacement.
Read the full definition →Bill shock
An invoice that arrives far higher than you expected, with no earlier warning that the amount was building.
Read the full definition →Percentage-of-savings pricing
A cost-tool pricing model that charges a cut of the spend it manages, or of the savings it finds, not a flat fee.
Read the full definition →A rate you looked up will change again.
Provider pricing pages change without notice. CostMon tracks the bill that pays for it.