Fastly CDN Pricing in 2026: Plans, Costs and Hidden Fees

The line item that surprises most teams on a Fastly invoice is not bandwidth. It is the second and third meters running alongside it. Fastly pricing in 2026 remains a pure usage model with no permanent free tier and no flat plan: you pay per GB delivered, per 10,000 requests served, and separately for every optional service you switch on. On an API-heavy workload with modest egress, requests and add-ons can outweigh transfer entirely. This article breaks down how the meters interact, walks a full worked bill, and gives the volume thresholds where the math stops favoring Fastly.

How Fastly Pricing Works in 2026

Three structural facts drive every Fastly bill, and none of them have changed materially heading into 2026.

First, regional metering. Bandwidth and requests are priced by the region where the edge node served the byte, not where your account lives. A workload split across North America, Singapore, and Mumbai is billed against three different rate cards.

Second, volume tiers are per-region, not global. This is the detail that breaks most spreadsheet estimates. Discount thresholds sit at roughly 10 TB, 50 TB, 150 TB, and 500 TB per month — but you climb them independently in each region. A team pushing 60 TB globally, spread across five regions at 12 TB each, sits in the entry tier everywhere. The same 60 TB concentrated in North America clears the 50 TB break.

Third, requests are a first-class meter. Every HTTP and HTTPS request is billed per 10,000, regardless of response size. A 200-byte JSON response and a 4 MB video segment cost the same request fee. For API gateways, ad-tech bidders, and telemetry ingest, this meter dominates.

Published rate ranges as of 2026

Fastly's list rates for the entry tier remain in the ranges below. Committed contracts negotiate downward, sometimes significantly, so treat these as a ceiling for anything above a few hundred TB per month.

The spread matters more than the headline. Entry-tier India egress costs more than twice entry-tier North America, and the gap persists at volume. If your growth is coming from South Asia, your blended cost per GB rises even as total volume climbs.

The Add-Ons That Move the Bill

Delivery is the advertised product. Everything below is metered separately and is where 2026 bills diverge from 2026 forecasts.

• Next-Gen WAF. Negotiated enterprise pricing. Contracts commonly land near $3,000/month at the low end and scale with protected request volume. This is frequently the single largest non-delivery line.
• Image Optimizer. Around $0.0025 per unique transformation. "Unique" is the operative word: every distinct combination of source asset, dimensions, format, and quality parameter is its own billable event. Responsive image sets with six breakpoints across AVIF, WebP, and JPEG generate eighteen transformations per source image.
• Dedicated TLS. Roughly $100/month per certificate slot. Multi-brand and white-label setups accumulate these fast.
• Origin Shield. No separate SKU, but shielded traffic traverses an extra internal hop on cache miss, and that hop is metered. Shielding lowers origin egress and raises CDN-side transfer.
• Compute at the edge. Billed on request count and compute duration. Wasm workloads with heavy per-request logic can quietly exceed their own delivery cost.
• Log streaming. Fastly's own endpoints are typically included, but the destination bill is yours. Full-fidelity logs at a few hundred million requests per day are a real observability tax.

The hidden fee nobody models: cache-miss amplification

Here is the angle most Fastly pricing write-ups skip. Your bill is not a function of your traffic. It is a function of your traffic divided by your hit ratio, plus origin fetch volume. Drop your hit ratio from 96% to 88% — a routine outcome of a cache-key change, a new Vary header, or a query-string parameter added by a marketing team — and origin fetches roughly triple. Those fetches cost you at your cloud provider's egress rate, which on major public clouds still runs well above what the CDN itself charges per GB in North America.

Instrument hit ratio per POP and per content type, not as a global average. A global 94% can hide a 60% hit ratio on your API surface where request fees are highest.

Worked Example: A Realistic 2026 Monthly Bill

Take a media-and-commerce workload: 22 TB North America, 6 TB APAC, 2 TB India, 200 million requests globally, 5 million unique image transformations, two dedicated TLS slots.

Read the table again. Delivery is roughly $4,100 of a $17,000 bill. Image transformation is 74% of it. The optimization lever here is not renegotiating bandwidth — it is collapsing the breakpoint matrix and caching derivatives aggressively so the "unique transformation" count stops growing linearly with catalog size.

Fastly vs Alternatives: A Workload Decision Matrix

Fastly earns its premium on three things: configuration changes that propagate in seconds, instant purge across the network, and genuinely programmable edge logic via VCL and Wasm. If your architecture depends on those, the premium is rational. If you are moving bytes, it is not.

Where the cost math flips

Run the crossover yourself. At roughly 25 TB/month in North America on Fastly list rates, delivery alone runs into the low thousands. Volume-priced CDNs in the same league quote flat commitments for that band. BlazingCDN's volume pricing starts at $100/month for up to 25 TB, with overage at $0.004/GB, and steps down to $2,500/month for 1,000 TB and $0.002/GB beyond 2 PB. For bandwidth-dominated delivery — media libraries, patch distribution, software updates — that is a different order of magnitude, with stability and fault tolerance comparable to Amazon CloudFront, 100% uptime, NVMe SSD edge storage, and roughly one-hour onboarding.

The honest caveat: BlazingCDN does not ship a VCL-equivalent programmable edge, and if your delivery layer is running auth logic or request rewriting in Wasm, Fastly remains the right tool. Many teams end up splitting — Fastly for the dynamic, logic-heavy surface, a volume CDN for the static bulk. That split is often where the largest single-month savings appear.

FAQ

Does Fastly have a free tier or fixed plans in 2026?

No permanent free tier and no standard plans. Fastly pricing is usage-based across bandwidth, requests, and optional services, with negotiated committed contracts available at enterprise volume. Trial credits appear periodically but should not factor into a production cost model.

Why is my Fastly bill higher than my bandwidth estimate?

Almost always requests, image transformations, or WAF. Request fees are charged per 10,000 regardless of payload size, so API and telemetry workloads accumulate cost independent of GB delivered. Image Optimizer bills per unique transformation, and a large responsive breakpoint matrix multiplies that count fast.

How do Fastly's regional volume tiers actually work?

Discount thresholds at approximately 10, 50, 150, and 500 TB are evaluated per region group, not against global totals. A globally distributed workload can miss every discount tier while a geographically concentrated one of the same size clears several. Model each region separately before forecasting.

Does Origin Shield increase or decrease my Fastly cost?

Both, in different columns. Shielding reduces origin fetches and therefore your cloud egress bill, but the extra internal hop on cache miss is metered transfer on the Fastly side. It usually nets positive when origin egress rates exceed CDN rates, which is common on major public clouds.

At what volume should I evaluate Fastly alternatives?

Around 25 TB/month of bandwidth-dominated traffic is a reasonable trigger for a comparison. Below that, Fastly's operational advantages typically justify the rate. Above it, and especially past 100 TB, per-GB differences compound into five- and six-figure annual gaps.

Can I negotiate Fastly pricing down?

Yes, on committed contracts. Published rates function as a ceiling; sustained volume above a few hundred TB per month with a term commitment is where meaningful discounts appear. Bring a modeled forecast with regional splits and request counts — vague volume claims get vague discounts.

What to Measure This Week

Pull last month's Fastly invoice and split it into four buckets: bandwidth, requests, transformations, and everything else. If bandwidth is under 40% of the total, your optimization work is not in your CDN contract — it is in your cache keys and your image pipeline. If bandwidth is over 70% and you are past 25 TB, run a shadow quote against a volume-priced provider on your actual regional split, not a global average.

Then instrument one metric you probably are not graphing: unique image transformations per new catalog item. If that number is above eight, you have a breakpoint matrix problem that will scale linearly with your content and quietly outgrow your delivery cost. What is your current ratio?