This chapter explains what the individual cloud services cost: which unit a provider meters, what sets the price of that unit, and which line items a service produces on the invoice and in billing data. It walks along the services, from the virtual machine through storage and networking to the platform services, and asks the same three questions of each: what is being paid for, what drives the price, and how the line item can be recognised in the data.

What this chapter covers

The chapters before this one explained how a cloud bill comes about, how billing data is structured, and under which pricing models the providers sell. This chapter starts one level below, at the single service. Every provider bills every service in its own unit, per second, per gigabyte and month, per million requests, per gigabyte processed, so the same architectural decision produces differently cut line items at AWS, Azure and Google Cloud. Knowing the unit lets you read a bill; knowing the price driver lets you change it.

Four sections structure the chapter, each by the question it answers:

  • Compute answers what you pay for per hour: how a machine, a container or a function is named and metered, what sets the hourly price, and what is billed beside the running time.
  • Storage and data answers what it costs to hold, read and move data: block, object and file storage with their classes and lifecycle, managed databases and analytics, backup and recovery, each with separate line items for capacity, operations and transfer.
  • Network answers which network boundary costs money and why the path sets the price: from egress to the internet through gateways, traffic distribution and public addresses to delivery through a CDN.
  • Platform answers how managed services bill that are neither machine nor storage: logs and metrics, identity and security services, AI models, rented and brought-along licences, whose line items often hang on data volume or requests rather than on running time.

Every page follows the same shape: how the service works and what the providers document, what it means for the cost, where it goes wrong in practice, and where the providers publish the numbers. List prices appear only where a provider publishes them with region, currency and date.

Key terms

  • Data transfer: traffic across a network boundary, which the chapter treats as a cost category of its own beside running time and storage.
  • Rightsizing: matching a resource to its actual utilisation; this chapter supplies the price factors that get changed in the process.
  • Effective cost: the cost of a line item after discounts and amortised commitments, the amount the pages here mean by the price of a unit.
  • Savings opportunity: a concrete change to a service with a calculated saving; each page names which changes those are for its service.
  • Reserved Instances: a purchase option that lowers the price of a unit without changing the unit, which is why it appears here as a price factor beside size and region.
  • Tag: the key-value pair through which a service's line items can be assigned to an owner, and which some line items, data transfer among them, do not carry.

Neighbouring chapters

This chapter sits between the pricing models and ownership. "Pricing models and commitments" before it explains how the providers set the prices that act per service here: on-demand, Spot, reservations and Savings Plans. "Allocation and ownership" after it assigns the line items that arise here to teams and cost centres. The costs described here are reduced in "Optimisation and savings", which draws on the price factors of this chapter. Readers looking for the frame start the handbook with "FinOps and cloud cost management"; the particulars of AI services and of European currency and tax questions have chapters of their own.