Pitfalls, exit costs and official sources
Why is my cloud bill so high? Eight surprise-bill patterns and how to fix each
The eight most common causes of unexpectedly high cloud and hosting bills for small businesses, how to find them in your cost report, and what to change first.
When a cloud bill jumps, the cause is almost always one of eight patterns: forgotten resources, over-sized servers, storage that keeps growing, data transfer, managed-service meters, networking extras, compromised credentials, or currency and tax effects. Open the provider's cost report, group the month by service, compare it with the previous month, and work down the largest increases. Most fixes take minutes once the cause is visible; the lasting fix is a budget alert and a monthly review.
This guide is part of the pitfalls overview. Amounts are illustrative.
Diagnose in five steps
- Group by service for this month and last month.
- Sort by increase, not by size.
- Group the top increase by resource or tag to find the specific item.
- Check usage type: hours, GB-months, GB transferred, requests.
- Check the start date: when did the line begin to grow? Match it to a change, a deployment or a staff member's project.
The eight patterns
1. Forgotten resources
Test servers, demo environments, a cluster from a proof of concept, a GPU instance used once. They bill every hour.
Fix: stop or delete; tag everything with owner and expiry; schedule non-production (cost control guide).
Illustrative: one forgotten VM at $0.30 an hour costs 730 × $0.30 = $219 a month.
2. Over-sized servers and databases
Sized for an imagined peak or copied from a template.
Fix: measure and step down a size (vCPU and RAM guide).
3. Storage creep
Snapshots with no retention limit, old machine images, unattached disks, logs kept forever, object versions accumulating.
Fix: retention policies, lifecycle rules, monthly clean-up (storage guide).
Illustrative: 3 TB of forgotten snapshots at $0.07 per GB-month is 3,000 × $0.07 = $210 a month.
4. Data transfer spikes
A viral post, a new video, a misconfigured backup copying to another region every hour, a bot scraping your site.
Fix: CDN, caching, rate limiting, check replication schedules (egress guide).
5. Managed-service meters
Managed databases, serverless functions, queues, logging and monitoring services each have their own meters (requests, GB ingested, read and write units). A chatty application or verbose logging can multiply them.
Fix: reduce log verbosity, sample traces, set retention, cache frequent queries.
Illustrative: logging 5 GB a day at an illustrative $0.60 per GB ingested is 5 × 30 × $0.60 = $90 a month; cutting debug logging in production to 1 GB a day saves $72.
6. Networking extras
Public IPv4 addresses charged hourly, NAT gateways processing large downloads, idle load balancers, VPN connections left up.
Fix: quarterly networking review; release unused addresses; avoid routing bulk traffic through NAT.
7. Compromised credentials
Leaked access keys or a phished admin password used to launch large numbers of expensive instances, often in regions you never use. The customer is generally responsible for usage on their account.
Fix: immediately revoke the credentials, stop the resources, contact the provider's support, and review access. Prevent with multi-factor authentication, no keys in code repositories, alerts on new regions and a tight budget alert. If personal information may have been accessed, assess your privacy obligations (shared responsibility guide). Report cybercrime to the Canadian Anti-Fraud Centre or local police as appropriate.
8. Currency, tax and billing changes
The usage is unchanged but the CAD amount rose: a weaker dollar, a card foreign transaction fee, tax applied because your GST/HST number is missing from the account, an expired commitment now billing at on-demand rates, or a free tier or credit that ended.
Fix: compare the USD invoice with the card statement (FX guide); check tax settings (GST/HST guide); track commitment end dates.
A one-page triage table
| Symptom in the cost report | Likely pattern |
|---|---|
| Compute hours up, no new project | 1 or 7 |
| Compute steady but high | 2 |
| Storage GB-months rising every month | 3 |
| Data transfer line jumped | 4 |
| Logging, monitoring or database requests up | 5 |
| Small networking lines adding up | 6 |
| Usage in a region you do not use | 7 |
| Same usage, higher CAD total | 8 |
After the fix
- Set or tighten a budget alert and a forecast alert.
- Add mandatory tags.
- Put a 30-minute monthly review in the calendar.
- Ask the provider about a goodwill credit if the cause was a genuine mistake, and explain what you changed. Credits are discretionary.
- Use the tax, FX and renewal calculator to reset next month's budget.
FAQ
Can I set a hard spending cap?
Some platforms and services allow caps or automatic actions, especially for test accounts and credits; others only alert. Check what your provider offers, and use separate accounts with strict limits for experiments.
Will the provider refund charges from a hacked account?
Sometimes, at its discretion, particularly if you act quickly and secure the account. It is not guaranteed, so prevention matters.
Why did my bill change when I did nothing?
Look at pattern 8 first: currency, tax settings, expired commitments or ended free tiers. Then check for automatic scaling and storage growth, which can rise without anyone acting.
Sources
- Canadian Centre for Cyber Security — account security, multi-factor authentication, baseline controls
- Canadian Anti-Fraud Centre — reporting fraud and cybercrime
- Office of the Privacy Commissioner of Canada — breaches of security safeguards
- Bank of Canada — exchange rates
