For fifteen years, a small team has been quietly stitching a federated edge into 2,196 corners of the global network. Most viewers will never know its name. That, more or less, is the point.
It is a Tuesday night in Milan and forty thousand people press play on the same football match. The video doesn't come from a hyperscaler hall in Frankfurt. It comes from a beige rack inside a Fastweb central office a few neighborhoods over, running software written in Tel Aviv, sold by a company headquartered in a Redwood City strip mall. The viewer never knows. The viewer also never buffers.
That rack, and roughly 2,195 others like it, is Qwilt's quiet empire. The company calls it the Open Edge Cloud. Most people in the industry have a more grudging name for it: the CDN the telcos finally got to own.
"We don't build a CDN that bypasses the carrier. We build a CDN that is the carrier - and rent it back to whoever wants to stream."— The Qwilt thesis, paraphrased
By 2010, anyone with a calculator could tell where streaming was headed. Netflix had just gone unlimited. YouTube was eating broadcast. The CDN business, then dominated by Akamai, had built a clever workaround - thousands of servers in colocation halls, halfway to the user. Halfway was good enough, until it wasn't.
Carriers carried the freight and got paid nothing for it. Publishers paid CDNs that delivered traffic they did not control. Viewers, on bad nights, watched a little spinning wheel. Everybody hated the arrangement; nobody had the leverage to change it.
Qwilt's founders thought the leverage was sitting in plain sight. The carriers already owned the last mile. They owned the racks. They owned the cooling, the IPs, the relationships. What they did not own was the software.
"The cheapest place to deliver a video is the place the video has already been delivered."— Edge caching, restated for the millionth time
Alon Maor, Dan Sahar and Mark Fisher started Qwilt the way most ambitious infrastructure companies start: with résumés full of Cisco and Juniper, a deck full of nervous-making diagrams, and a wager that the incumbents would not see them coming. Their wager was specific. They believed a federated CDN, owned in pieces by hundreds of service providers and orchestrated by a single neutral software layer, would eventually deliver more video than any centralized network ever could.
It was an unfashionable idea. The decade that followed was the decade of the hyperscaler, of AWS regions and Cloudflare PoPs and big logos in big metros. Qwilt spent it doing the opposite: shipping appliances to carrier branch offices in Argentina and Italy and Turkey, then convincing the carriers' procurement teams to deploy them.
The Israeli press, never overly generous with hometown hype, called it patient capital. Industry analysts called it slow. Cisco, eventually, called it a $70 million check.
"Open Caching was a polite way of telling Akamai that the carriers had hired a lawyer."— A telecom analyst, off the record, 2022
To make the bet work, Qwilt could not own the whole thing. A federated CDN only federates if everyone agrees on the rules. So the company helped found the Streaming Video Technology Alliance, drafted the Open Caching specification with anyone who would join, and then - and this is the part that still surprises people - shipped the first commercial implementation. Standards bodies almost never produce running code that quickly. This one did, because Qwilt needed it to.
The Open Edge Cloud is technically two products that depend on each other. Carriers get a turnkey edge cloud they can install in their own facilities - hardware from Cisco, software from Qwilt, financing from Digital Alpha. Publishers get an API that lets them publish into thousands of carrier-owned caches at once, paying for delivery instead of haggling with each operator individually.
It is, in other words, a marketplace. The carriers supply the racks; Qwilt supplies the matchmaking. Both sides take a cut of the bill that used to go entirely to a centralized CDN.
If you publish video, Qwilt sells you delivery into the network instead of toward it - lower latency, fewer hops, better quality of experience on the third quarter of the championship game. If you operate a network, Qwilt sells you a way to turn your access infrastructure into a revenue line instead of a cost center. If you write applications that need to be close to users - low-latency multiplayer, AR, real-time analytics, edge AI inference - the same API is starting to look interesting for things that are not video at all.
"The Open Edge Cloud is the only place where the carrier, the publisher and the standards body have all stopped suing each other long enough to ship."— Closer to the truth than anyone admits
You can argue with a deck. You cannot argue with a tier-one carrier letting you install your own software inside their headends. Qwilt's customer list looks less like a startup roster and more like a regulator's filing cabinet.
Qwilt's stated mission - and one of the few mission statements written by an infrastructure company that survives contact with reality - is to make the Open Edge Cloud the default place to deliver content and applications. Default is a stronger word than it looks. Default means engineers do not have to think about it. Default means it is faster to opt in than to opt out. Default is what the cloud became to data centers, what AWS became to colos, and what Qwilt would like the edge to become to the cloud.
Whether they get there depends on a small number of unsexy things continuing to work: the Open Caching standard staying open, carriers continuing to share revenue rather than fork the stack, and publishers deciding that one API into a federated edge is preferable to a procurement matrix of separate operator deals. None of these are guaranteed. All of them are, for once, plausible.
"The carriers spent thirty years building the world's most valuable infrastructure and giving it away. Qwilt's mission, more or less, is to stop doing that."— What the polite version of the pitch leaves out
Streaming is the wedge. It will not be the prize. The same federated edge that terminates a football match three milliseconds from your couch is, in principle, a perfectly good place to run a game server, a generative inference call, a real-time translation engine, or an industrial control loop. All of those need to be close to the user. None of them are well-served by centralized clouds shoving packets across the continent.
Qwilt's job, over the next five years, is to convince the people who build those things to use the Edge API the way they currently use a Lambda function. The fact that Cisco, BT, Verizon, Vodafone and Comcast are already pre-installed is a head start that is hard to engineer around.
Back to that Tuesday night in Milan. The match ends. The viewer closes the app. Somewhere, a Qwilt node logs a session and goes back to waiting. The carrier counts the revenue it did not used to make. The publisher counts a quality score it did not used to hit. The hyperscaler counts a stream it did not used to deliver. The viewer counts nothing, because nothing went wrong.
The internet, against all odds, got a little less centralized today. The company that did it is named after a quilt.