Cloud Waste Self-Check
Cloud overspending is usually the sum of a few ordinary habits. Answer five questions about how your infrastructure is run and get a straight read on where you stand.
Cloud waste is spend that buys nothing: environments running while nobody works, instances sized for traffic that never arrived, storage nobody deletes, a bill nobody owns, and frontier-model API calls doing work a smaller model would do. Oleg Sotnikov, a fractional CTO who keeps a production platform serving users in 190+ countries on about $650/month of AWS, treats those five signals as the first pass over any company's cloud spend. This self-check asks the same five questions and runs entirely in your browser.
Five questions
Answer honestly. "Not sure" counts as a signal: on a cloud bill, nobody knowing is itself the finding.
Answered 0 / 5
01Do dev and staging environments run nights and weekends?
02Has anyone reviewed instance sizing against actual utilization in the last six months?
03Do old snapshots, logs, and backups have lifecycle policies?
04Does one named person see and own the cloud bill?
05Do all your LLM/AI API calls go to a frontier model regardless of task?
Answer all five to see the verdict.
Nothing is sent anywhere. The check runs in your browser, and your answers leave with the tab.
Frequently Asked Questions
How much cloud waste is normal?
There is no honest number before someone reads your bills. The five patterns in this check are common at every size, from a small startup account to a large enterprise commitment, and what they cost depends on which ones are running and for how long. A percentage quoted before anyone has seen your setup is a guess.
What are the fastest cloud savings?
Switching off what nobody uses. A schedule for non-production environments, unattached volumes and forgotten snapshots deleted, lifecycle policies on logs. Hours of work, no architecture changes, visible on the next invoice. Rightsizing and reserved capacity come after that, once real utilization is known.
What about LLM API bills?
They fail the same five-signal test: if every request goes to a frontier model regardless of task, that is question five answering itself. Try the AI cost calculator on this site with your own volumes to see what routing, caching, and batching would return; the audit then verifies it against the real bill. AppMaster runs 11B+ tokens a month, so this is daily work rather than theory.
What happens in the audit?
Five business days at a fixed $5,000. I take read-only access to your billing, infrastructure, and repos, then hand back a prioritized savings map with every finding quantified in dollars per year, plus a plan your team can execute. Nothing changes in production while I look. If the audit doesn't identify at least $50,000/year in combined payroll, cloud, and LLM savings, you pay nothing.
Find out what the bill is actually buying
Five business days, a savings map in dollars per year, and a plan your team can execute.
At least $50,000/year in identified savings — payroll, cloud, and LLM bills — or the audit is free.
Related reading
Where cloud and token bills go, and how to keep them small.
