Cloud spending has a direction: up. Not because anyone decides to spend more, but because every force in the system points the same way — adding is instant and removing requires a decision, storage only accumulates, vendors nudge tiers upward, and the bill arrives as an auto-paid line item nobody itemizes. The result is a ratchet: each month’s spend becomes next month’s baseline, and eighteen months later the number is 40% higher with nothing obviously different. Pushing back doesn’t take a finance initiative — it takes knowing where cloud waste specifically hides, and one recurring hour. Here’s both.
Where the waste hides, by layer
Licenses and subscriptions — the biggest, easiest layer. Seats assigned to nobody, tiers above what roles need, bundled tools bought twice: the Microsoft 365 licensing review and the SaaS keep-consolidate-kill pass cover this ground, and together they’re routinely worth 15–25% of the whole cloud line. If you do nothing else in this article, do those two.
Storage — the layer that only grows. Cloud storage is metered forever and pruned never: the file shares that migrated with a decade of history, backup retention set generously and never tiered, mailboxes and Teams recordings accumulating on defaults, the departed-employee data preserved (correctly!) but on premium storage indefinitely. The fixes are structural, not heroic: lifecycle policies that shift aging data to cheap archive tiers automatically · retention set by decision rather than default — including the deletion half of the decision · and one honest look at what’s being stored twice.
Infrastructure — where the meter really runs. If you’ve moved servers to Azure or similar, the classic waste patterns apply: oversized VMs (specified generously at migration, never right-sized against actual utilization — the single biggest infrastructure saving, usually) · things running at full price during empty hours — dev and test systems, workloads with business-hours rhythms, all schedulable · paying on-demand rates for steady loads that reserved-capacity pricing discounts substantially — the cloud equivalent of committing annually for the stable core of your licenses · and zombie resources: the unattached disks, idle IPs, and orphaned snapshots left behind by every project, invisible individually, real in aggregate. The platforms’ own cost tools (Azure Advisor and kin) will literally list these — the recommendations sit unread in most tenants, which tells you where the hour goes.
Egress and plumbing — the fine print layer. Moving data out of clouds costs money, which surprises businesses during migrations, backup designs, and vendor exits. It’s rarely the biggest line; it’s the one nobody budgeted, and one more reason the exit question belongs in vendor selection.
The practice: one hour, quarterly, with teeth
Cost optimization fails as a project (one heroic cleanup, then the ratchet resumes) and works as a rhythm: quarterly, one hour, three moves — read the bill itemized against last quarter and interrogate every line that grew (“what is this, who owns it, why is it bigger?” — the same owner-and-reason discipline as the admin-account census) · run the platform’s recommendation report and action the top three items · and log the decisions, because “we chose to keep X at $Y for reason Z” is what makes next quarter’s hour fast. Fold the output into the annual budget and the quarterly review — in a managed relationship, bill-watching should simply be part of the service, for the self-evident reason that a provider who hands you savings has earned the renewal conversation.
Two guardrails keep optimization honest. Don’t cut resilience to cut cost: backup retention, immutable copies, and failover capacity are cheap against what they insure — trim the zombie disks, not the safety systems. And price the alternative fairly: cloud line items are visible in a way server rooms never were — power, cooling, hardware refresh, and Saturday failures billed nobody monthly. The goal isn’t the smallest bill; it’s a bill where every line has an owner and a reason.
The bottom line
Cloud costs drift because adding is easy and removing takes a decision — so institutionalize the decision: licenses and SaaS reviewed on their cycles, storage lifecycled and retention chosen, infrastructure right-sized against real utilization, and one quarterly hour reading the bill like it’s someone’s job. Businesses that adopt the rhythm typically claw back 20–30% in the first pass and hold the line thereafter — not by spending less on what matters, but by finally not paying for what doesn’t.