What Is a CDN? Content Delivery Networks, Explained

Most people picture one server when they load a website. The request flies off to a machine in some data center, and the page comes back. For almost every site you touch daily, that picture is wrong. The page came from a server that copied it earlier and parked it a few milliseconds from you, and that server is probably not the one closest to you on a map.

That copy-and-park system is a CDN, a content delivery network, and it quietly carries a huge slice of internet traffic. This guide covers what a CDN is, the problem it solves, how it routes and caches your requests, who runs the biggest ones, and whether your own site actually needs one.

What is a CDN?

A CDN, or content delivery network, is a group of servers spread across the world that store cached copies of a website’s content and deliver each request from a server close to the user. Instead of every visitor reaching one origin machine, they pull HTML, images, scripts, and video from a nearby edge. Less distance for the data to travel means a faster page.

CDN stands for exactly that: content delivery network. They are now plumbing for most of the web. The market sat around $38.75 billion in 2026 and is forecast to reach $164 billion by 2035, which tells you how central this layer has become.

The problem a CDN solves: distance, load, and bandwidth

A CDN exists to fix three problems that hit any site with a global audience: distance, load, and bandwidth. Serve everything from one machine and far-away users wait, traffic spikes can topple the server, and large media files choke a single connection. A CDN attacks all three by moving copies closer to people.

Distance is the first one, and it is pure physics. A user in Amsterdam loading a site hosted in Virginia, and another in Tokyo loading the same page, both send requests across continents before the server even starts to reply. That delay is latency, and latency is what makes a site feel sluggish.

One machine, every continent. Distance becomes latency, and latency is what makes a site feel slow.

Load is the second. If every request hits one origin, a viral post or a product launch can fire millions of requests at once and the server buckles. Bandwidth is the third: video, images, and downloads all compete for the same pipe out of a single data center. Modern sites need a way to push content outward, lighten the origin, and ride out spikes. That is the job a CDN was built for.

How does a CDN work?

A CDN works by catching your request through DNS, routing it to the nearest edge server, and answering from that server’s cache whenever it can. The origin only gets pulled in when the edge does not already hold a fresh copy. Most of the magic happens before a single byte of the page loads.

It starts with DNS, the system that turns a domain name into an IP address. With a CDN in front, that lookup does not return one origin IP. It returns the CDN, which then decides which edge location should handle you. Your request lands at that edge, the edge checks its cache, and the path forks into a hit or a miss.

1. 01 Browser You request cdn.example.com
2. 02 CDN DNS Returns an edge IP, not the origin
3. 03 Nearest edge Your request lands at a PoP

cache?

Cache hit

The object is already on the edge. It comes straight back. The origin never even hears about you.

Cache miss

The edge fetches from the origin, stores a copy, then answers. The next visitor nearby gets a hit.

On a hit, the edge returns the cached object right away and the origin never hears about you. On a miss, the edge fetches from the origin, keeps a copy, then sends it back, so the next nearby visitor gets a hit. That single behavior is what offloads so much work from origin servers.

Why “nearest” rarely means closest

On the internet, the nearest server is the one with the fastest network path, not the shortest distance on a map. CDNs choose it with a few routing techniques, and the result is often counterintuitive: an edge a few hundred kilometres farther away can still answer before a closer one stuck behind congested links.

Three strategies do the steering. Anycast lets many servers share one IP address, and the internet’s own routing carries you to the closest reachable instance. GeoDNS answers based on where your request comes from, sending whole regions to specific server pools. Latency-based routing measures live response times and picks the quickest endpoint, even across regions.

Closer · 300 km95 ms

Farther · 800 km12 ms ✓

This is the fact that trips people up. A server 800 km away on a clean, low-hop route beats one 300 km away routed through six congested links. When a CDN says it sent you to the nearest edge, it means the fastest path, and that distinction is most of what makes the modern web feel quick.

What is CDN caching? Cache hits vs cache misses

CDN caching means storing copies of content on edge servers so future requests can skip the origin. When the edge already holds a fresh copy, that is a cache hit and the response is near-instant. When it does not, that is a cache miss, and the edge has to fetch from the origin before it can answer. The gap between the two is enormous.

Cache hit served from the edge

3 ms

Cache miss full trip to the origin

453 ms

≈151× faster when the edge already holds your content

How long content stays cached is set by rules. A short time-to-live suits API responses, hours fit news articles, and days work for images and scripts. When something changes before its time is up, developers force a refresh with a manual purge or by versioning the filename, like shipping /app.v2.js so the URL itself changes. Without those controls, people would keep getting stale pages.

What is a CDN server? Edge servers and PoPs

A CDN server, usually called an edge server, is a caching server placed at the edge of the network, close to users, that stores and serves content for the origin. Many edge servers sit together in a site called a Point of Presence, or PoP, and large networks run hundreds of PoPs in cities worldwide. The edge is what you actually talk to.

request flows up only on a miss

Edge server 01

The cache that answers you. It stores copies of pages, images, scripts, and video segments close to where you are.

Serves the responseTerminates TLSClosest to the user

↑

PoP (Point of Presence) 02

A data center, usually next to an internet exchange, holding many edge servers. Big networks run hundreds of them worldwide.

Many edge serversNear an exchangeHundreds globally

↑

Origin 03

The one authoritative source: your web server, API, or storage bucket. The CDN only contacts it on a cache miss.

Source of truthHidden behind the edgeTouched only on a miss

The key split is edge versus origin. The origin is the one authoritative source, your web server, API, or storage bucket, and it holds the current version of everything. The edge holds fast copies. Originally CDNs cached only static files like images and scripts, but modern ones also speed up dynamic content and APIs, and on platforms like Cloudflare every edge can run code, blurring the line between a CDN and full edge computing.

CDN examples: Netflix, YouTube, and game updates

The easiest place to see CDNs at full stretch is streaming and gaming, where a central server alone would never cope. Netflix, YouTube, and every major game launcher lean on edge delivery to reach millions of people at the same moment without melting their core infrastructure.

Netflix built its own CDN, Open Connect, and places appliances inside ISP networks so popular shows stream from hardware a short hop from your living room. Google serves YouTube through a planet-scale edge network and now rents it out as Media CDN, reaching viewers in over 200 countries. Gaming is the loudest example: a single Fortnite update once pushed Akamai’s CDN traffic to 106 Tbps, and Riot Games’ League patcher spreads downloads across several CDNs at once to find the fastest path for each player.

You hit CDNs constantly outside those giants too. Social feeds load images from CDN domains, news sites lean on them to survive 10x traffic surges during breaking events, and online stores use them to keep product pages quick during a sale. If you want to run one yourself, a handful of providers cover most needs.

Why do CDNs matter?

CDNs matter because they improve four things at once: speed, reliability, security, and cost. Content sits closer to users, the service survives the loss of any single region, attacks get absorbed before they reach the origin, and the bandwidth bill drops. Few infrastructure changes pay off on that many fronts.

Speed ~70%

median latency drop reported after moving to a well-tuned CDN, the difference users feel first

Reliability 100s

of PoPs to fail over between, so one bad region does not take the whole site down

Security & scale 106 Tbps

peak traffic a single CDN has absorbed at the edge, shielding the origin from the flood

Cost 90%+

of requests a busy service can serve from cache, so far fewer bytes ever reach the origin

Speed is the one users feel first, with teams reporting median latency dropping around 70% after moving to a well-tuned CDN. Reliability comes from spread: because content lives in many places, traffic reroutes around a failed region instead of going dark. Security comes from the edge standing in front of your origin, soaking up DDoS floods and filtering bad requests. And cost falls because a busy service can serve the bulk of its traffic from cache, so far fewer bytes ever reach the origin.

So, do you actually need a CDN?

If your site has a global audience, heavy images or video, or any chance of a traffic spike, a CDN is close to essential and often free to start. Faster load times also help search rankings, so the upside reaches beyond raw speed. For most serious sites the question is not whether to use one, but which.

The bigger shift is in how you think about scale. A CDN does more than speed up content. It changes where delivery happens. Your origin becomes the source of truth while the network absorbs the flood, which is why, when you load a page, you are rarely talking to one server. You are touching a distributed system built to move data efficiently across the planet.

For the animated version of all of this, the routing, the cache, and the real-world scale in about eleven minutes, watch the full explainer on YouTube.