The figure that decides the bill is not the 1 TB
“1 TB of images” sounds like a single number, but the annual bill depends on how those bytes leave the server, how many times they leave, and whether anyone pays for the request count separately. On a $6 plan with 1 TB of monthly transfer, 1 TB per month is exactly at the limit; 1 TB per year is a rounding error. Budgeting starts with fixing the unit before comparing options.
The worked example below assumes a photography or product site pushing 1 TB of image traffic per month, about 33 GB per day, served from a $6 VPS. Plan assumptions come from the budget VPS comparison table, which lists included transfer and overage rates so the arithmetic can be checked against real plans rather than hypothetical ones.
Baseline: what the $6 VPS bills with no CDN
A $6 plan typically includes 1-3 TB per month. If it includes 1 TB, a 1 TB month bills at exactly $6 but leaves zero headroom, and any month with a traffic spike or a search crawler pass pushes into overage. At a typical $0.01/GB, 200 GB of extra traffic costs $2, a 33% increase on the plan for one busy week.
The more expensive problem is upstream bandwidth contention. Serving 1 TB/month averages about 0.4 MB/s, which is trivial. But a single gallery page that loads 8 MB of images can be requested by 400 visitors in ten minutes, producing a 50-90 MB/s burst that saturates a budget VPS NIC and drags TTFB for HTML along with it. The plan price does not change; the experience does.
Option A: keep it on the VPS and prepay overage
Prepaying extra transfer is the cheapest option when traffic is steady and geographically concentrated. Buying a 3 TB allowance often costs $2-$4/mo more than the 1 TB tier, which is $24-$48 per year and removes the overage risk entirely. That beats most CDN pricing for a single-region audience, where a CDN adds a hop but no latency benefit.
The catch is burst protection. Prepaid transfer does not increase the NIC or the disk read throughput. If the 1 TB is dominated by a few large galleries, caching those on the box with expires 30d and converting to WebP cuts bytes by 25-40% before any infrastructure change, which is often the highest return action available.
Option B: full CDN in front, origin shielded
A CDN with a 90% hit rate moves 100 GB/month through the origin and 900 GB through the edge. At typical budget-CDN pricing of $0.01-$0.04 per GB in North America and Europe, 1 TB at the edge costs $10-$40 per month, or $120-$480 per year, which is 2-5 times the VPS itself. That is a real cost, and on a $6 box it dominates the budget line.
The CDN becomes worth it when hit rates are high and the origin plan is small. A $6 plan with 1 TB included plus a CDN at 92% hit rate handles roughly 12 TB of edge traffic per month before the origin allowance is touched. If your traffic grows past that, the CDN is cheaper than upgrading to a $30 origin plan with 10 TB included, but not cheaper than a $9 plan with 3 TB for a 1 TB workload.
Option C: object storage plus CDN for images only
Splitting images to object storage removes them from the VPS transfer line entirely. Storage at $0.015/GB/month for 50 GB of images is $0.75/mo, and serving from an object store with a CDN front is typically $0.005-$0.02 per GB out. For 1 TB per month that is $5-$20, plus $9 per year in storage, which is competitive with option B and removes image load from the VPS disk as well.
The hidden line is operations: request counts, egress from the object store to the CDN, and cache invalidation. Providers that bill per 10,000 requests add roughly $0.40 per million requests, which is minor for images but noticeable if you serve thumbnails in many sizes. Measure requests per page before choosing a per-request-priced tier.
Annual cost table for 1 TB per month
The table compares the three options over a year, using the $6 VPS as the anchor and mid-range CDN and object-storage rates. Figures exclude the VPS compute itself except where the origin plan must be upgraded.
| Line | A: VPS + prepaid | B: VPS + CDN | C: VPS + object + CDN |
|---|---|---|---|
| Compute plan | $6/mo | $6/mo | $6/mo |
| Included transfer | 3 TB prepaid, +$3/mo | 1 TB origin | 1 TB origin |
| Edge egress (1 TB) | $0 | $20/mo | $15/mo |
| Object storage (50 GB) | $0 | $0 | $1/mo |
| Request charges | $0 | $0 | $2/mo |
| Monthly total | $9 | $26 | $24 |
| Annual total | $108 | $312 | $288 |
| Break-even traffic | under 3 TB/mo | above 4 TB/mo | above 3.5 TB/mo |
For a flat 1 TB/month, option A wins by a wide margin. The CDN options only become rational at sustained traffic far above the origin allowance, or when the burst problem matters more than the annual line, such as a launch or a seasonal peak where the 50-90 MB/s spike is the actual constraint.
Modeling the year, not the average month
Annual budgeting needs the monthly distribution, not the mean. Use twelve rows of actual egress from your server logs, then multiply each by the per-GB rate including overage. A site averaging 1 TB with two months at 2.6 TB will bill differently under option A than a site at exactly 1 TB every month, because overage applies in those two months.
Add a 15-20% buffer for crawlers, retries and log shipping. Bots that ignore robots.txt can add 5-10% of image traffic alone, and hotlink scraping can add far more if images are unprotected. Referrer or user-agent blocking at the CDN edge removes most of it for free.
Reducing bytes before buying bandwidth
Compression is cheaper than any CDN tier. Converting JPEG to WebP at quality 80 typically cuts image bytes 25-40%; serving AVIF where supported adds another 10-15%. Setting Cache-Control: public, max-age=2592000, immutable on hashed asset names drops repeat requests to near zero. Resizing originals so no image is served above 2x its display width removes the largest single source of waste.
On the server side, enabling Brotli for HTML and CSS and keeping images out of the gzip path avoids wasting CPU on already-compressed data. On a 2 GB VPS with one vCPU, image transformation through a dynamic resizer costs more CPU than the bandwidth it saves; pre-generating three sizes at upload time is the better trade.
Decision checklist before the year starts
Pull the last 12 months of egress, note the peak month and the number of months above the allowance, and pick the option whose break-even line sits above your peak rather than your average. Recheck at quarter boundaries, because a single viral gallery can move a site from option A to option C in one month.
If the peak month is more than double the average, buy headroom on the plan rather than a CDN, since burst tolerance comes from the origin. The budget VPS plan comparisons sort by included transfer and overage rate, which is the only pair of numbers that matters for this line item.
