Every serious head of streaming engineering has cleaned up after the kickoff thundering herd, has the encoding ladder memorised, and has stopped pretending the per-GB conversation is going anywhere useful. CDN vendors sell aggregate. Live needs instantaneous. Nobody serious about live is still renting the moments that matter.
Media commit rates flatlined for five years. Underneath the same acronym, the market split into two products that share fibre, hardware, and three letters on the invoice. One sells delivery. One sells perimeter. The buyers, the failure modes, and the pricing logic were never the same shape.
Not a pivot. A late acknowledgment of two products the buyers had always treated as different sales.
The serious ones built it
Akamai shipped FreeFlow in 1999, with streaming media delivery in the portfolio from day one. The catalogue had recognised for years what the buyers already knew. Streamers and enterprise apps were different sales.
Netflix made it an industry. When Watch Instantly launched in 2007, streaming was a rounding error. By 2013, Netflix was a third of US peak downstream traffic. The largest commercial CDN customer in the world, on a single growth curve. Three CDN vendors carried it. Akamai. Limelight. Level 3. Per-GB delivery, head to head.
Then Netflix left.
Open Connect launched in 2012. Appliances shipped to ISPs, traffic served from inside the access network. By 2016, Netflix served the vast majority of streaming via Open Connect. The largest buyer in the market built the exit and walked through it.
Every streamer at meaningful scale now owns their delivery stack. YouTube runs on Google’s own edge, embedded inside telco networks via Google Global Cache. Amazon Prime Video runs on AWS. Disney spent years on their stack, then bought BAMTech in 2017 for the answer. ESPN+ and Disney+ run on it. CAPEX bought ownership of the peak.
The CDN vendors built for the buyers that left never recovered. Limelight absorbed Edgecast and rebranded as Edgio. Edgio went under in 2024. Level 3 was absorbed into Lumen. Lumen still trades, but sold its CDN business to Akamai in 2023 and exited the category. Two of the three companies that defined commercial media CDN are extinct. The third reorganised around perimeter, with media as one product among many in the catalogue.
The commercial media buyer at meaningful scale is leaving. The CDN vendors built to serve them are consolidating.
Media side: bytes are the whole sale
Media CDN was sold on delivery from day one. Edge nodes near the audience. Interconnect with ISPs. Egress to the last mile. Caches filled ahead of demand. Platform features made the delivery credible. Bytes were the product.
How media CDN is actually priced:
- Per-GB egress. The floor, the ceiling, the negotiation. Rates sit under NDA. Published numbers are the asking price nobody at scale pays.
- Commit discount. Nobody at scale commits multi-year. Monthly/annual minimums, tier discounts on the committed meter. The buyer trades flexibility for a lower floor.
- Regional differentials. APAC and LATAM priced well above US and EU. Transit and peering asymmetry, not delivery difficulty. The streamer optimises traffic shape around the map.
The platform pitch was a courtesy. The bytes were the sale.
And the meter stopped going up. FHD is enough for most viewers. Traditional studios are producing less. Attention fragments across short-form, parallel screens, and AI-generated feeds.
What’s left is live. A Matildas goal. State of Origin III. AFL finals. Collective moments that command shared screens at scale, where peak concurrency hits in seconds and there is no second chance.
Live is where commercial CDN runs out. CDN vendors sell aggregate. Live demands instantaneous. Kickoff is a thundering herd. Edge servers hit their ceiling. Network links saturate. Fresh segments slow. Manifests over-cache, latest segments lag behind live. Every part of the stack has to hold under load. The peaks that matter don’t average out.
The tournaments carried on Optus Sport. At the moment a nation tunes in, the CDN vendors don’t have the capacity you need. You build it yourself, or you ship a degraded event. There is no third option. The CDN we built at Optus Sport ran in country alongside three commercial CDNs as a content delivery portfolio.
Enterprise side: bytes were never the sale
Open Cloudflare, Fastly, or Akamai today. The lead position isn’t CDN anymore. AWS never had to make the move. CloudFront was always a service in the catalog, never the lead. November 2025 bundled it like the others: flat-rate tiers with WAF, DDoS, edge compute and Route 53, attack traffic off the meter.
Akamai’s enterprise products sold as website performance through 2011 to 2013. Dynamic Site Accelerator, application acceleration, Ion. Kona Site Defender shipped as web application firewall in 2012. Bot Manager shipped in 2016. The 2014 Prolexic acquisition sealed it. Performance became table stakes. Security became the product. The incumbent had already built it, in the same window the security-first entrants arrived. The market and the incumbent agreed.
Enterprise CDN sells security. WAF, DDoS, bot management, edge compute, observability, audit. Bytes make the platform credible. The platform is the product. Nobody hires a CDN to deliver bytes to an enterprise app. They hire it to stop the attack and shield the origin from the public internet.
How enterprise CDN is actually priced:
- Commitment discounts. Savings Plans, Commits, Enterprise Discount Programs. The customer pays on a schedule regardless of usage. Smooth cashflow trades at a higher multiple.
- Bundled meters. WAF requests, edge invocations, log ingestion, bot inspections. Each a line item above the bytes floor. New services arrive already metered.
- Single subscription. Cloudflare’s flat-rate model from inception. Pooled across tenants, usage variance cancels.
Attack absorption is the differentiator. AWS Shield Standard at $0 on every account. Fastly’s Zero Attack Fees. The CDN vendor takes attack-traffic risk onto its own balance sheet.
The bytes line was a courtesy. The platform was the sale.
And the platform keeps growing. More SaaS, more APIs, more public-facing apps. Every perimeter has to hold against credential stuffing, scraping, DDoS, and bots that write their own evasion. The number of buyers grew. The number of attacks grew faster.
AI added a second growth vector, same shape. Defence. Inference runs in centralised GPU clusters. Every token costs real money. Every abuse case costs more. Prompt injection defence. Abuse controls on agents that can spend real money. Audit trails for regulated deployments. Each one a SKU above the bytes floor.
Cyber compounds. AI compounds. Bytes don’t.
Migration up-stack
The capital is moving. Enterprise compounds. Media doesn’t. Public CDN vendors face shareholder pressure they can’t satisfy by staying on both sides. The migration is structural and it’s underway.
Every survivor moved up-stack by buying security. Akamai shipped Kona Site Defender in 2011 and bought Prolexic in 2014. Akamai bought Linode in 2022 for alternative cloud. Fastly paid $775M for Signal Sciences in 2020. Thales bought Imperva for $3.6B. Limelight and Level 3 didn’t move. Both are gone from the market. Migrate up-stack or fold. None compounded back down.
The forward-looking moves are shipping now. Fastly Object Storage previewed in 2024, collapsing storage and CDN into one product. Zero Attack Fees took attack-traffic risk onto Fastly’s balance sheet. Cloudflare R2 ran the storage play in 2022. AWS stings hard on egress transit out of S3. Zero-egress storage pulls workloads onto the platform that hosts it. The CDN market isn’t where this gets decided. It’s downstream of where it gets decided.
AWS is the exception in keeping media front and centre as a product offering. Elemental wires the full media supply chain onto the AWS invoice. MediaLive, MediaPackage, MediaTailor, MediaConvert. The invoice is the moat.
Race is on until public markets decide it’s over.
If you’re buying media CDN in 2027, you’re buying it from a hyperscaler, you’re operating your own, or you’re buying from a specialist that treats live as craft. The commodity middle disappears in between.
Reading the bill
If you’re buying media CDN, the floor is the whole fight and the vendor pool is shrinking out from under you. If you’re buying enterprise CDN, the floor is rounding error and the vendor pool is consolidating toward you.
Bytes don’t pay the rent in either market. They never did in enterprise. In media, the market will never build enough capacity for the moments that actually matter.
If those moments are your business, you already know what the bill is telling you.
Build the moments that matter. Build your capacity.