Cloud costs (FinOps)

Play this lesson

The bill doubled

Shard Garden's cloud bill doubled this month. Kai asks Theo: was it more players, or a mistake?

The FinOps Foundation defines FinOps as "an operational framework and cultural practice which maximizes the business value of technology", "through collaboration between engineering, finance, and business teams."

Six principles

The FinOps principles, in no particular order:

Quick check

The team that runs weed races watches what weed races cost. Which principle is that?

  1. Everyone takes ownership for their technology usage
  2. FinOps should be enabled centrally
  3. Take advantage of the variable cost model of the cloud
Show the answer

Everyone takes ownership for their technology usage

The people who use the cloud own its cost, because they are the ones who can change it.

Cost per player

A bill on its own says little. Divide it by the players it served: the cost per player.

A bill that doubles while players double is healthy growth. A bill that doubles while players stay the same is a leak.

Do the math

bill
               Sep     OctCloud bill    $600  $1,200Players     20,000  24,000

What happened to the cost per player?

  1. It rose from 3 cents to 5 cents
  2. It stayed the same
  3. It halved
Show the answer

It rose from 3 cents to 5 cents

$600 ÷ 20,000 is 3 cents. $1,200 ÷ 24,000 is 5 cents. Players grew 20%, but the bill grew 100%: something else is costing money.

Labels say who spent it

In Google Cloud, a label is "a key-value pair that you can assign to Google Cloud resources", such as service=weed-races or environment=staging.

Labels go to billing, which "lets you break down your billed charges by label."

Quick check

Theo finds half of October's rise labelled environment=staging. What is the likely leak?

  1. Staging servers left running at full size
  2. More players in production
  3. Players playing in staging
Show the answer

Staging servers left running at full size

Staging has no real players. A staging bill that grows means servers someone forgot to turn down.

Down as well as up

The cloud scales down as easily as up. Shard Garden's staging shrinks every night and over weekends, and each race server size is reviewed every month against the load tests.

Bolt says: Next: the Keep it running boss.

Fact-checked 12 October 2026