Cloud storage advertises prices like “$0.023 per GB per month” and it sounds incredibly cheap. You calculate that storing 10TB costs $230 monthly and think you’ve found a bargain. Then the first bill arrives at $2,400 and you start digging through line items wondering what happened.

Cloud providers structure pricing to look cheap in the calculator but expensive in production. Understanding where costs actually come from helps you avoid nasty surprises.

Storage Is Actually Cheap

The advertised storage price is real. Keeping data sitting in an object storage bucket genuinely costs pennies per gigabyte per month. If you upload data and never touch it again, your bill will match the calculator.

The problem is nobody uses storage that way. Data exists to be accessed, moved, processed, and retrieved. All of those actions cost money, often much more than the storage itself.

Egress Fees Are the Real Cost

Moving data out of cloud storage costs anywhere from $0.08 to $0.12 per GB, depending on volume and destination. That’s 4-5x the monthly storage cost per gigabyte. Retrieve that 10TB dataset once and you’ve paid $800-1,200 in egress fees alone.

Egress fees hit several ways. Data transferred to the internet costs money. Data transferred to different cloud regions costs money (though less). Data transferred between availability zones might cost money. Only data staying within the same zone is typically free.

If you store video files and serve them to users, every stream generates egress charges. A 1GB video file costs $0.023 to store but $0.09 to deliver to a viewer. Serve it to 100 viewers and you’ve paid $9 in bandwidth while the storage cost $0.023 total.

Companies have been hit with five-figure surprise bills after a file went viral, getting downloaded millions of times. The storage cost was trivial. The egress fees were catastrophic.

API Request Pricing

Every operation on cloud storage costs money. LIST operations to see what files exist. GET operations to retrieve files. PUT operations to upload. DELETE operations to remove files. Each action costs fractions of a cent, but those fractions add up fast.

AWS S3 charges $0.0004 per 1,000 PUT/COPY/POST/LIST requests and $0.0004 per 1,000 GET/SELECT requests. That sounds negligible until you’re processing millions of small files.

Storing log files as individual objects? Each log line written is a PUT request. Processing them means GET requests for each file. You might execute millions of API calls daily without realizing it. At $0.40 per million, that’s hundreds of dollars monthly in API fees for files costing dollars to store.

Listing bucket contents seems free but counts as API calls too. If your application lists directory contents to find files, you’re paying for every list operation. Applications that scan buckets frequently or recursively can rack up significant API charges.

Storage Class Confusion

Cloud providers offer multiple storage classes with different pricing. Standard storage costs more but allows instant access. Archive storage costs much less but retrieval takes hours and costs extra.

The pricing calculator shows archive storage at $0.004 per GB monthly compared to $0.023 for standard. That’s 85% cheaper! But archive retrieval costs $0.02-0.10 per GB depending on retrieval speed, plus additional API charges.

Archive makes sense for truly cold data you’ll almost never access. But if you retrieve that data even occasionally, the retrieval fees exceed the storage savings quickly. Retrieve archived data monthly and you’re paying more than keeping it in standard storage would have cost.

Glacier Deep Archive offers even cheaper storage at $0.00099 per GB monthly, but retrieval costs $0.02 per GB plus $0.05 per request, and takes 12+ hours. It works for compliance archives you hope to never need. It’s terrible for anything you might actually want to access.

Minimum Storage Duration Charges

Archive storage classes have minimum storage duration requirements. Upload a file to S3 Glacier and delete it the next day? You still pay for 90 days of storage. The minimum duration is 90-180 days depending on the storage class.

This catches people moving data around trying to optimize costs. You realize files don’t need to be in standard storage, move them to archive, then later realize you do need them and move back to standard. Between early deletion fees, retrieval costs, and data movement charges, you’ve paid far more than just keeping files in standard storage the whole time.

Cross-Region Replication

Setting up replication across regions for disaster recovery sounds prudent. It also doubles or triples your costs. You pay for storage in both regions, egress fees to transfer data between regions, and replication API calls.

A 10TB dataset with cross-region replication costs storage in both regions ($230 x 2 = $460), plus roughly $800-1,200 to transfer the initial copy, plus ongoing replication costs for every change. Your $230 monthly storage bill just became $500+ monthly plus a large one-time transfer fee.

Replication might be necessary, but understand you’re paying for every byte multiple times over.

Lifecycle Policies Gone Wrong

Lifecycle policies automatically transition data between storage classes based on age. Files older than 30 days move to infrequent access storage, files older than 90 days move to glacier, etc.

Automatic transitions sound smart until you realize each transition is a COPY operation (API charge) plus early deletion fees if you transitioned files that were already subject to minimum duration charges from a previous transition.

Complex lifecycle policies can create situations where moving files around costs more than the storage savings from cheaper classes. The more complex your lifecycle rules, the more likely they’re costing rather than saving money.

Monitoring and Data Transfer Fees

Even monitoring your storage costs money. Cloud provider APIs to check bucket sizes, object counts, and access patterns count as API calls. Third-party cost monitoring tools often scan your buckets repeatedly, generating API charges while trying to help you save money.

Every time you copy data between buckets, between storage classes, or between accounts, you’re paying. Data doesn’t move for free inside cloud environments like it might on local storage.

What Actually Works

Know your access patterns before choosing storage classes. If you access data regularly, standard storage is often cheapest despite higher storage costs. The lack of retrieval fees and minimums makes it cheaper overall.

Batch operations reduce API costs. Instead of 1,000 separate GET requests for small files, bundle data into larger objects when possible. Fewer larger operations cost less than many tiny ones.

Cache frequently accessed data closer to users. Pay egress fees once to move data to a CDN or edge locations, then serve from there. The CDN costs money too, but less than repeated egress from primary storage.

Watch your bills closely early on. The first month reveals your real costs. Small patterns that seem fine at low volume might become expensive at scale. Better to catch a $50 surprise than a $5,000 one.

Choose regions carefully. Serving global users from a single region means international egress fees. Using regional storage closer to users reduces egress but increases management complexity and potentially storage costs.

When It Still Makes Sense

Despite all these gotchas, cloud storage often remains the best option. Building your own storage infrastructure costs more than cloud storage for most use cases. The ability to scale instantly and pay-as-you-go provides real value.

The key is going in with eyes open. Storage isn’t as cheap as the advertised price suggests. Your real costs depend on access patterns, data movement, and operational needs. Budget for 3-5x the basic storage cost and you’ll be closer to actual expenses than the pricing calculator suggests.