CDN Pricing in 2026: What Providers Actually Charge

Take a 50 TB per month HTTPS workload, price it against six major providers using their public 2026 rate cards, and the spread between the cheapest and most expensive invoice lands near 4.7×. Same bytes, same regions, same cache-hit ratio. That gap is why CDN pricing comparisons built on headline per-GB rates are useless: the rate card is maybe 60% of the invoice. Request fees, regional multipliers, origin egress from cache misses, and metered add-ons do the rest. This article gives you the 2026 pricing models side by side, the four variables that actually move your CDN cost, a workload-fit matrix, and an arithmetic walkthrough so you can compute your own effective cost per delivered GB before your next renewal.

What actually drives CDN cost in 2026?

Four levers dominate. Everything else is rounding error on most invoices.

• Region mix. As of 2026, North America and Europe remain the cheapest delivery zones on every metered provider. APAC, South America, the Middle East and Africa carry premiums that typically run 1.5× to 3× the NA/EU rate. A workload that shifts 20% of traffic from EU to India or Brazil can raise a blended rate by 25–40% without a single byte of extra volume.
• Requests per GB. This is the number most teams never instrument. Four-second HLS or LL-HLS segments at 5 Mbps produce roughly 2.5 MB per request, or about 400 requests per GB. At $0.0075 to $0.010 per 10,000 HTTPS requests, that adds $0.0003–$0.0004 per GB — trivial. An API or tile-serving workload averaging 20 KB per object generates 50,000 requests per GB, which adds $0.037–$0.050 per GB. Request fees can exceed the transfer fees outright.
• Cache-hit ratio and origin egress. Origin pull is billed by your cloud, not your CDN. Dropping from 95% to 85% CHR on a 100 TB month moves origin egress from 5 TB to 15 TB. At typical 2026 cloud egress of $0.08–$0.09 per GB in NA/EU, that extra 10 TB is roughly $800–$900 per month bolted onto a bill that has nothing to do with your CDN contract.
• Commit structure. Committed-volume contracts in 2026 still discount 30–50% off metered list, but the floor is real. Traffic that seasonally drops 40% below commit means you are paying for bytes you never delivered. Commit to the trough, not the average.

Add-ons come fifth: real-time log streaming, image optimization, edge compute invocations, dedicated certificates, and premium support routinely add 15–40% on top of delivery. Log egress in particular surprises people — streaming every request event from a high-RPS property can cost more than the traffic that generated it.

2026 CDN pricing comparison: models, not just rates

Rate cards change; billing models change slowly. This table compares how each provider charges as of Q1–Q3 2026, with indicative NA/EU public rates where they are published.

Two structural notes for 2026. First, hyperscaler egress-to-CDN remains the quiet arbitrage: S3 to CloudFront transfer is free, so an AWS-native origin gets a discount no third-party CDN can match on the origin leg. Second, 95th-percentile Mbps billing has not disappeared. If your traffic is spiky, per-GB pricing is almost always cheaper; if it is flat, percentile billing can beat it.

How to estimate your own CDN bill

Six inputs, one formula, no vendor calculator required. Pull these from your current provider's analytics for a full billing month.

1. Delivered GB at edge, split by billing region.
2. Total billable requests, and derive requests per GB by dividing.
3. Cache-hit ratio, by volume rather than by request count.
4. Origin egress GB — measure it at the cloud provider, not the CDN.
5. Peak-to-mean ratio, for percentile or commit modelling.
6. Add-on units: log events, edge function invocations, image transforms.

Then: effective cost per GB equals (transfer charges + request charges + add-on charges + origin egress charges) divided by delivered edge GB. That single number is the only fair basis for a CDN pricing comparison.

Worked example: 10 TB of NA/EU HLS video

Assume 10,240 GB delivered, 4-second segments averaging 2.5 MB, so roughly 4.1 million requests. Cache-hit ratio 95%, origin egress 512 GB at $0.085/GB.

• CloudFront at $0.085/GB: $870 transfer plus about $31 in request fees, roughly $901, plus $44 origin egress if the origin is not S3 — call it $945 total, or about $0.092 per delivered GB.
• Bunny at $0.010/GB: $102 transfer, no request fees, roughly $146 with origin egress. About $0.014 per GB.
• BlazingCDN on the 25 TB plan: $100 flat for the month, plus origin egress. About $0.014 per GB and falling as volume rises into higher tiers.

Now change one variable. Make it an API workload with 20 KB average objects instead of video: 10 TB becomes roughly 512 million requests. CloudFront request fees alone reach about $3,840 — more than four times the transfer cost. That inversion is the single most common cause of CDN bill shock in 2026, and it is entirely predictable from requests per GB.

Best-for matrix: matching workload profile to pricing model

The multi-CDN split follows directly from this matrix. Route heavy static and video egress to the cheapest per-GB provider, keep personalised and compute-dependent paths on a premium edge, and steer with DNS or a client-side selector. Teams delivering multi-hundred-terabyte months on this pattern find that the bulk-egress leg is where the savings live: BlazingCDN's volume-tiered pricing starts at $4 per TB and drops to $2 per TB ($0.002/GB) at 2 PB, with stability and fault tolerance comparable to Amazon CloudFront at a fraction of the delivery cost. It runs on 100% uptime, allows flexible per-property configuration, and scales quickly through release-day and live-event spikes, which is why media and software companies including Sony use it for bulk delivery.

Cheap CDN tactics that survive scrutiny

• Prune the cache key. Strip tracking query parameters, unnecessary cookies, and vary-inducing headers. Every unnecessary key dimension fragments the cache and converts hits into billed origin pulls.
• Lengthen segments where latency allows. Moving VOD from 2-second to 6-second segments cuts request volume by two thirds with no impact on start-up time for on-demand playback.
• Sample logs. A 1% or 5% sample preserves statistical validity for most analytics and cuts log egress proportionally. Keep 100% only for error classes.
• Tier your origin shield. A mid-tier cache in front of origin raises effective CHR on long-tail objects and directly reduces cloud egress.
• Commit to the trough. Size annual commits against your lowest month, price bursts on-demand, and re-baseline quarterly.
• Audit regional distribution monthly. A marketing push into a premium zone can move your blended rate before anyone notices the invoice.

FAQ

What is a realistic CDN cost per GB in 2026?

Public list rates in NA/EU span roughly $0.002 to $0.085 per GB as of 2026, a spread of more than 40×. Enterprise negotiated rates on premium providers typically land between $0.020 and $0.050 per GB at scale. Your effective rate — total invoice divided by delivered edge GB — is usually 10–30% higher than whatever rate card you signed.

Why is my CDN bill higher than my per-GB rate implies?

Three usual causes. Request fees on small-object traffic, regional premiums on APAC or LATAM delivery, and origin egress from a cache-hit ratio lower than you think. Compute requests per GB first; if it exceeds about 10,000, request charges are probably your largest line item after transfer.

Is a cheap CDN a false economy for video delivery?

Not on cost grounds. Bulk video egress is a commodity operation dominated by peering quality and cache efficiency, and low-cost providers with zero request fees frequently deliver identical byte-for-byte performance. The genuine trade-offs are in edge compute maturity, log tooling, and configuration depth — not in the delivery of large cacheable objects.

How do I compare CDN pricing across providers fairly?

Normalise everything to effective cost per delivered GB using your own traffic profile: delivered GB by region, requests per GB, cache-hit ratio, origin egress, and add-on units. Then run the same numbers through each candidate's 2026 rate card. Never compare headline rates directly, because the billing dimensions differ.

Does committing to volume always reduce CDN cost?

Only if your baseline is predictable. Commits typically discount 30–50%, but unused commit is unrecoverable spend. Size against your lowest expected month and handle seasonal peaks with on-demand or a secondary provider.

Should I run multi-CDN purely for cost?

It is one of the more defensible reasons. Splitting bulk static and video egress to a low-cost provider while retaining a premium edge for dynamic and compute paths often cuts total delivery spend by half or more. The operational cost is duplicate configuration and a steering layer, which is worth it above roughly 50 TB per month.

Run this benchmark this week

Pull last month's CDN analytics and compute four numbers: requests per GB, cache-hit ratio by volume, percentage of traffic in premium regions, and total invoice divided by delivered edge GB. Most teams find at least one of those four is materially different from what they assumed when the contract was signed. Then re-price the same traffic profile against two alternative billing models from the table above — one request-free volume-tiered provider, one plan-bundled provider — and see whether the delta justifies a migration or a renegotiation. If your requests per GB comes back above 20,000 and you are still on per-request billing, that is the conversation to have first. What did your effective cost per GB turn out to be?