Companies Using Fastly

Fastly is the one technology in this wave detected purely from HTTP response headers, and that makes it worth explaining what that buys you and what it costs.

When we detect Fastly, we are reading the server header and Fastly's own request identifier on a live response. The domain is not merely a customer, it is being served through Fastly's edge at the moment we fetched it. There is no markup to strip, no script to block, no consent banner in the way. Header detection is about as close to ground truth as this work gets.

The cost is that headers describe a response, not an organisation, and a single company can serve some hostnames through Fastly and others through a rival. Our 91,625 is a count of domains observed being delivered through Fastly, which is a narrower and more literal claim than a customer list.

The category ranking is wrong in a way worth fixing

Fastly ranks twelfth of forty-nine in our CDN category, on 0.48%. Taken at face value that reads as a marginal player. It is a classification problem.

Our CDN category contains two genuinely different products. There are commercial edge networks, where a company pays to have its own traffic delivered: Cloudflare, Fastly, Google Cloud CDN, Akamai. And there are free public library CDNs, where any site can reference a script without an account, a contract or a relationship: cdnjs on 11.11%, jsDelivr on 11.09%, Google Hosted Libraries on 8.24%, unpkg on 3.36%. All four outrank Fastly, and none of them is a competitor to it.

Ranking a paid edge network against a free script host is like ranking a haulage company against a public road. Both move things, only one of them is a purchase.

Among the products a company actually buys, Fastly sits behind Cloudflare and Google Cloud CDN. That is a harder position than the raw rank suggests in some ways and a much better one in others, and it is the comparison a reader should be making. We have said so on the page rather than letting the table imply something we know to be misleading.

Two customer bases wearing one name

The size distribution looks contradictory until you separate it. 79.97% of Fastly domains belong to companies with ten staff or fewer, the highest small-company share of anything we built this month. Meanwhile the top of the list is Bloomberg, TED, Block, Vox Media, Business Insider and the United Nations Development Programme.

Both are accurate, and they describe two different businesses sharing a logo. There is a small number of very large media and technology properties, the ones Fastly is known for, moving enormous volumes of cache-heavy content. And there is a long tail of individual developers and small studios on self-serve plans. Counting domains weights those two equally, which is why the median customer looks nothing like the reference customer.

Anyone using this data commercially should decide which population they want before filtering, because a list built on the raw count will be four-fifths micro-businesses.

The most technical customer base in the wave

Software Development leads Fastly's industry mix at 10.28%. That number deserves attention because of how unusual it is. On the other five profiles we built this month, no industry clears 5%: Google Workspace tops out at 4.54% advertising, Font Awesome at 4.17% construction, Elementor at 4.52% construction. Fastly's leading industry is more than twice as concentrated as any of them.

Add IT services on 7.69%, technology and internet on 5.21% and design services on 5.14%, and roughly a quarter of the identified base is technical. That is what a product configured through a domain-specific language and an API looks like in the data, rather than one configured through a dashboard.

The age profile says the same thing from another angle. 44.6% of these companies were founded in the 2010s and 35.38% in the 2020s. Four in five are under fifteen years old. Set that beside the Mimecast profile in this same wave, where 16.85% of customers predate 1960, and you have two products that could not be selling into more different generations of business.

On the switching numbers, and why there are so few of them

Before publishing any migration figures here I checked the baseline, which is a discipline I would now apply to every stackable category. 23.55% of active Fastly domains already run Cloudflare. CDNs coexist routinely, in front of each other and beside each other for different asset paths, so a domain showing both proves nothing on its own.

Measured against that baseline, most apparent flows evaporate. Domains leaving Fastly show cdnjs, jsDelivr, unpkg and jQuery CDN at rates below their ambient presence, which means those are not destinations at all. Only two flows clearly exceed the baseline: Cloudflare, which took 1,442 domains and returned 529, and Google Cloud CDN, which took 239.

Run the competitor's ambient co-occurrence rate before you run the churn query. Without that baseline you will publish a migration table that is really a popularity table.

So this page carries a deliberately short switching section. Two rows that survive the test, and nothing else. A longer table would have looked more authoritative and told the reader less.

Reading the chart

Fastly appears in our history from November 2011 on one domain, the year the company was founded, and climbs to a long-run peak of 175,167 in March 2025. Our current live crawl records 91,625. That is not a collapse. It is two measurement systems with different reach, and header-based detection in particular depends on which hosts and paths a given crawl actually requests. The decline through mid-2025 is our archive feed narrowing its coverage, which we can confirm because total domains fall in step with active ones.

We publish both segments with the seam visible at TechnologyChecker.io rather than splicing them into one smooth line, because the join is the part a careful reader needs to see.

My summary: Fastly has the most trustworthy detection basis in this wave and the most misleading category rank. It serves a genuinely technical, genuinely young customer base, split between a handful of very large publishers and a long tail of small developer shops, and it is losing a modest but consistent number of domains to Cloudflare.

One more thing the headers tell us

Because we read the response rather than the page, this profile captures infrastructure that no markup-based crawler would ever see: API endpoints, image and asset hosts, download mirrors. That is part of why our company match rate here is 53.8% rather than the eighty-odd percent we get on a DNS-detected product like Mimecast. A large share of Fastly-served hostnames are not websites at all, so there is no company record for them to match.

I would treat that gap as a feature rather than a defect. It means the count reflects delivery infrastructure as it is actually deployed, including the parts that never render a page, and it is a reminder that a detection count and a customer count are different quantities. We report both numbers separately for exactly that reason.