Most GCP overspend is not exotic. It clusters into five categories that appear in nearly every estate we review: idle and orphaned resources you are paying for but not using, Compute Engine instances provisioned far larger than their workload, steady baseline usage running at full on demand price because no committed use discount covers it, storage that accumulates in the wrong class with no lifecycle rule, and network egress generated by architecture that did not have to generate it. The order matters. Idle cleanup and rightsizing pay back immediately with no commitment and no risk, while commitments and architecture come once you know the shape of the steady estate. Sequencing the work this way means you commit to the size you actually need rather than the bloated one you started with.
This is the cluster overview that sits above common GCP billing surprises and your first GCP cost optimization sprint. Read those for the specific surprises to expect and a week by week sprint plan.
Idle and orphaned resources: the free win
The cheapest savings are resources you are billed for and nobody uses. Unattached persistent disks left behind when a VM is deleted keep billing at full rate. Reserved static external IP addresses that are not attached to a running resource carry a charge precisely because they are idle. Forgotten load balancers, idle Cloud SQL instances kept running over weekends, and old snapshots that no policy ever prunes all bill around the clock. None of this needs a commitment or a redesign. The Recommender surfaces idle resources, but it recommends, it does not decide, so the saving is real only when someone reviews the list and deletes with confidence.
Oversized Compute Engine: pay for what runs
The second category is compute provisioned for a peak that never arrives. Teams pick a machine type by guess, double it for safety, and never revisit. GCP rightsizing recommendations read actual CPU and memory utilization and propose a smaller machine type or a custom machine type sized to the workload, and custom types let you tune vCPU and memory independently rather than jumping between fixed shapes. Because sustained use discounts apply automatically to running instances, rightsizing compounds: a smaller instance is both cheaper per hour and still earns the automatic discount. Rightsize before you commit, never after, or you will buy a commitment against bloat.
Uncovered steady state: the commitment gap
Once the estate is lean, the largest remaining lever is commitment coverage. Baseline usage that runs every hour of every day at full on demand price is the single biggest avoidable cost on a mature estate. Committed use discounts trade a one or three year commitment for a discount against on demand, with spend based commitments covering a dollar amount of usage flexibly and resource based commitments covering specific machine resources at a deeper rate. The discipline is risk adjusted coverage: commit to the defensible floor of your usage, not the peak, so utilization stays high and you never strand commitment on a workload you retired.
Storage sprawl and egress: the slow leaks
The last two categories leak slowly but never stop. Cloud Storage accumulates in the Standard class long after data goes cold, when a lifecycle rule could move it to Nearline, Coldline, or Archive automatically as it ages. Old buckets, duplicated datasets, and ungoverned exports add up. Egress is the other leak: cross region reads, traffic served from origin rather than a cache, and the premium network tier used where the standard tier would do. Neither is glamorous, but a lifecycle policy and a network tier review remove recurring spend that would otherwise compound for years.
A worked example
An indicative European SaaS company asked for a single number to cut and got a ranked list instead. The first pass deleted unattached disks, released idle static IPs, and pruned an unmanaged snapshot pile, which removed a meaningful slice immediately at zero risk. The second pass rightsized a fleet of oversized Compute Engine VMs using utilization data. Only then did they size committed use discounts against the now leaner steady state, and add Cloud Storage lifecycle rules. Working the categories in order meant the commitment they signed matched real demand rather than the bloat they began with. Figures are verified against billing data and anonymised.
Frequently asked questions
What is the most common GCP waste?
Should I buy commitments before or after rightsizing?
How do GCP sustained use discounts affect rightsizing?
Where this fits the cost review
These categories recur because nothing in GCP forces a cleanup, so a structured first sprint that works them in order is where the fastest savings live. We run that review as an independent buyer side advisory across AWS, Azure, GCP, and OCI, with zero provider commissions and a guarantee: we reduce your cloud spend or we reimburse our service fee.
For monthly buyer side analysis delivered to your inbox, subscribe to The Cloud Spend Navigator.
Put a defensible number on your cloud spend.
No provider in the room, no published price list. Tell us your footprint and we will scope the savings against your billing data — we reduce your cloud spend or we reimburse our service fee.
The Cloud Spend Navigator: what changed in cloud pricing, commitments, and FinOps — no vendor spin.