Cloud billing glossary: the terms on Azure, AWS and Google Cloud price lists
Cloud price lists are full of terms that mean similar things on each cloud, under different names. This glossary explains the ones you’ll meet when comparing costs.
Pricing models
On-demand (pay-as-you-go): Paying the list price for what you use, with no commitment. All prices on this site are on-demand.
List price: The public price before any discount. Large customers often pay less through negotiated agreements.
Reserved instances / reservations: Committing to a specific resource (such as a VM size in a region) for one or three years in exchange for a large discount. Azure calls them reservations, AWS Reserved Instances, and Google uses committed use discounts for similar deals.
Savings plans: A commitment to spend a fixed amount per hour on compute for one or three years, in exchange for a discount on whatever compute you use. More flexible than reservations. Offered by AWS and Azure; Google’s spend-based committed use discounts are similar.
Spot (preemptible) capacity: Spare capacity sold at a deep discount that the cloud can take back at short notice. Good for batch jobs, unsuitable for always-on services. Azure and AWS call it Spot; Google uses Spot VMs (formerly preemptible).
Sustained use discounts: Google Cloud’s automatic discount on some VM types when they run for a large part of the month, with no commitment.
Tiered pricing: The unit price drops as monthly usage passes set thresholds, for example cheaper messages after the first billion. Our estimates apply tiers where the vendor publishes them.
Free tier: A free monthly allowance, or free trial credit, for some services. Our estimates leave these out unless the price list itself includes a free tier.
Units
vCPU: A virtual CPU, usually one hardware thread. Most compute is priced per vCPU-hour, or per VM-hour for a fixed size.
GB vs GiB: A GB is 1,000,000,000 bytes; a GiB is 1,073,741,824 bytes, about 7% more. Google often prices in GiB, others in GB. We convert so the comparison is like-for-like.
GB-month: Storing one GB for a whole month. Storing 30 GB for one day is about one GB-month.
Hours per month: Cloud estimates usually assume 730 hours, the average month (8,760 hours in a year ÷ 12).
Throughput unit (TU): Azure Event Hubs’ unit of capacity: about 1 MB/s or 1,000 events per second in. See event streaming costs.
Streaming unit (SU): Azure Stream Analytics’ unit of processing capacity.
KPU: Kinesis Processing Unit, AWS Managed Flink’s unit: 1 vCPU and 4 GB of memory. See stream processing costs.
Shard: A unit of capacity in a Kinesis stream (provisioned mode), with fixed read and write limits.
Message metering size: The block size a service counts messages in: 4 KB on Azure IoT Hub, 5 KB on AWS IoT Core. See message size and batching.
Storage terms
Storage tier (class): Hot, cool, cold or archive storage: cheaper to keep as you go colder, but dearer to read. See object storage for sensor archives.
Minimum storage duration: On colder tiers, the shortest time you’re billed for, even if you delete the data sooner.
Requests / operations / transactions: Charges per read, write or list on object storage. Google groups them as Class A (mostly writes) and Class B (mostly reads).
Rehydration / restore: Bringing data back from an archive tier so it can be read. Can take hours, and is charged.
LRS, ZRS, GRS: Azure’s redundancy options: locally redundant (copies in one data centre), zone-redundant (across zones in a region) and geo-redundant (a copy in a second region). AWS and Google have equivalents under other names.
Network terms
Ingress: Data coming into the cloud. Usually free.
Egress (data transfer out): Data leaving the cloud to the internet, or moving between regions or zones. Usually charged per GB, and one of the most common surprises on a bill. Not included in our estimates.
Region and availability zone: A region is a geographic location (such as Mumbai); zones are separate data centres within it. Prices are set per region. See cloud costs in India.
Billing and management
Billing account, subscription, account, project: How each cloud groups costs. Azure: billing account → subscriptions → resource groups. AWS: organization → accounts. Google: billing account → projects.
Tags / labels: Key-value labels on resources (for example site=pune) that let you split a bill by factory, team or product.
Amortised cost: Spreading an upfront reservation payment across the months it covers, so each month shows its fair share.
Budget and alerts: A spending limit with email or webhook alerts when you approach it. All three clouds offer them free.
FinOps: The practice of managing cloud spending as a shared responsibility between engineering, finance and the business.
TCO (total cost of ownership): All the costs of running something, including staff time, support plans and migration, not just the cloud bill.
See also: how to read a cloud bill and how our estimates are calculated.