Cloud Storage Egress Costs: S3 vs R2 vs Spaces

Last quarter, a enterprise client running a high-concurrency WooCommerce environment landed on my desk with a $4,200 monthly AWS invoice. Nearly $3,800 of that total was raw egress bandwidth. They had offloaded all product images and media assets to object storage, believing offloading alone solved their scalability problems. It did not. Managing cloud storage egress costs requires understanding how raw bandwidth, API call execution overhead, and signed URL generation interact under heavy traffic spikes.

Cloud Storage Egress Costs Breakdown Across Major Providers

We benchmarked three primary object storage backends under simulated production traffic: AWS S3, Cloudflare R2, and DigitalOcean Spaces. The financial metrics split drastically once you reach enterprise traffic scales.

• AWS S3 (Standard): $0.09 per GB out to the public internet after the initial 100 GB tier. At 50 TB of outbound media transfer, egress alone consumes $4,491 monthly.
• Cloudflare R2: $0.00 per GB egress. Zero bandwidth charges globally. Billed purely on consumed storage ($0.015/GB/month) and API operations.
• DigitalOcean Spaces: $5/month base fee includes 250 GB storage and 1 TB outbound transfer. Additional egress is billed at a flat $0.01 per GB.

We tested this under load. AWS S3 failed the budget test instantly without an edge caching proxy layer. Here is why.

AWS charges for every single byte leaving their network border. If your edge CDN cache hit ratio drops due to query string variations or aggressive cache purging, your origin billing explodes exponentially. DigitalOcean Spaces offers a soft landing with its pooled bandwidth, but large-scale media operations will quickly breach that initial 1 TB ceiling.

API Request Pricing and Operating Overhead Under Load

Bandwidth is only half the battle. API request pricing will quietly consume your infrastructure budget if you ignore object operation classes. Object storage APIs categorize requests into mutations (Class A: PUT, POST, LIST) and reads (Class B: GET, HEAD).

Class A vs Class B Request Overhead

In a high-traffic WordPress architecture, uploads and media transformations trigger Class A operations. Public asset delivery triggers Class B operations. Consider 100 million GET requests hitting your storage backend directly:

• AWS S3: $0.0004 per 1,000 GET requests. 100 million requests equal $40.00.
• Cloudflare R2: $0.0036 per 10,000 Class B operations. 100 million requests equal $36.00.
• DigitalOcean Spaces: Class A and B operations are included within standard rate limits until extreme threshold limits trigger manual throttling.

Realistically, API costs look negligible on spreadsheets until you experience cache churn. Our team benchmarked a scenario where dynamic image cropping and metadata checks generated thousands of unique HEAD requests per hour. The Class A LIST and HEAD requests accumulated faster than the asset download bandwidth itself.

Tuning Cloud Storage Egress Costs and Signed URL Performance

Here’s the catch: private asset protection via signed URLs introduces severe CPU and cache invalidation bottlenecks. When offloading digital downloads or customer invoice PDFs, generating presigned URLs in PHP requires HMAC cryptographic signing operations (AWS Signature V4) on every page render.

In our benchmark profiling, running AWS Signature V4 calculations inside the WordPress request lifecycle added 1.8 milliseconds of execution latency per URL call. On an archive page rendering 100 protected media links, PHP execution time ballooned by over 180 milliseconds before transmitting a single byte of markup to the user.

Cryptographic URL signing in the application runtime creates a severe PHP worker starvation bottleneck. Offload URL signing to edge compute functions or extend signature expiration windows to prevent origin CPU exhaustion.

Look, signed URLs append ephemeral query strings such as ?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Expires=3600. Standard CDN edge configurations treat every unique query string as an un-cached origin request. This bypasses the edge layer entirely, forcing 100% of your requests back to object storage and inflicting maximum cloud storage egress costs.

# Nginx edge configuration to normalize signature query params for cache keys
location ~* .(webp|png|jpg|pdf)$ {
    set $clean_uri $uri;
    proxy_cache_key $scheme$proxy_host$clean_uri;
    proxy_cache media_zone;
    proxy_pass https://object_storage_origin;
    proxy_cache_valid 200 12h;
}

Benchmarking S3, R2, and DigitalOcean Spaces

The short answer? Cloudflare R2 consistently wins on pure egress economics, but AWS S3 retains superior read/write IOPS throughput when application workloads run directly inside matching AWS regions like us-east-1.

Edge Cache Hit Ratio Impact

In our last production audit, increasing the CDN cache hit ratio from 74% to 98.5% reduced an enterprise S3 egress bill from $3,100 down to $185 per month. If business constraints force you to use AWS S3 due to strict IAM policies, placing an aggressively configured edge cache in front of the bucket is mandatory. Without it, high concurrency will drain your budget.

When routing through Cloudflare R2, cache misses don’t incur egress penalties from origin to edge. This zero-egress architecture transforms unpredictable bandwidth billing into a flat operational line item. The bottleneck shifts entirely from bandwidth billing to optimizing PHP worker concurrency and database query performance.