Instance cores and RAM

Instance Ram running

The memory half of a virtual machine, billed per gibibyte-hour alongside the vCPU SKU. On memory-optimised families this line is usually larger than the CPU one, which is worth knowing before assuming CPU is where the money goes.

Billed: Per gibibyte-hour of memory, priced per machine family and region. Separate from the vCPU SKU.

Compute EngineOn-demand

What triggers this charge

  • Every hour a VM runs, proportional to its provisioned memory rather than its used memory.
  • Memory-optimised machine families (M-series), where the per-gibibyte rate and the quantity are both high.
  • Custom machine types, where memory can be scaled independently — and often was, upward, during an incident and never back down.

How to reduce it

Right-size against observed working set

Provisioned memory is billed whether or not it is touched. Compare peak actual usage over a month against what is allocated.

Use custom machine types

Predefined types force a fixed CPU-to-memory ratio. A custom type buys exactly what the workload needs instead of the next size up.

Check the family fits the workload

A memory-optimised family for a CPU-bound job pays a premium for memory the workload never uses.

Rates

Rates for this SKU group are not shown yet — the catalog sync has not run. We publish rates from Google’s Cloud Billing Catalog API or not at all; a hand-typed rate that looks authoritative and is six months stale is worse than no rate. In the meantime, cloud.google.com/skus has the current list price.

Related charges

Now find out how much of this you are actually paying.

Finitizer breaks your Google Cloud bill down to the SKU, shows how each one has moved since last week, and separates rate changes from usage changes — so you know whether to talk to engineering or to procurement.