← all stories

After the Web · Era 3 · The early Web

1996–2024shut down by VK in 2024

ICQ

Mirabilis

The first mass-market internet messenger: it showed which of your friends were online right now and gave every user a permanent number (UIN). Built in Israel in 1996, sold to AOL for $287 million just two years later, resold to Russia in 2010 for $187.5 million, and there it lived for another 14 years, until in 2024 its new owner, VK, switched the service off with a single line on the website.

Founders Ariel (Arik) Vardi · Yair Goldfinger · Sefi Vigiser · Amnon Amir · Yossi Vardi (investor, chairman)
Domains icq.com · mirabilis.com
messagingnetwork-effectacquired-twicegeo-arbitrageproduct-as-asset

In brief

ICQ was the first messenger to turn the very idea of seeing which of your friends are online right now into a mass-market one. Four programmers in their twenties from Tel Aviv built it in 1996 without spending a single dollar on advertising, and just two years later they sold it to AOL for $287 million. After that the story went in a direction unusual for dot-coms: the product did not die along with its era but was resold a second time, to Russia, in 2010, and lived there for another 14 years, until in 2024 its owner, by then renamed from Mail.ru Group to VK, switched it off after warning users with a notice on the website a month in advance.

How it started (the founders)

Ariel (Arik) Vardi, Yair Goldfinger, Sefi Vigiser, and Amnon Amir met as colleagues at Zapa Digital Arts in Tel Aviv (3D graphics for the internet) and left the company in the middle of 1996 after disagreements. The idea for the future ICQ was born during a game of ping-pong: at first they discussed a pager-plus-internet combination, and then Goldfinger wrote a prototype of a notification that a friend had come online. The ping-pong episode is described independently by Calcalist and by a 2003 ISRAEL21c article built on an interview with Goldfinger. Amnon Amir left the project as early as its second month to return to his studies, which explains his absence from the company's later quotes and press releases.

The starting money came not from a venture fund but from Arik's father, Yossi Vardi, an investor who became the company's chairman. The sources give the amount differently: Calcalist and Goldfinger himself speak of several hundred thousand dollars at the start, while the business press of 1998 speaks of $3.2 million invested over almost two years before the sale; these are probably different stages, but no source reconciles them.

The first version was built in less than two months. The same ISRAEL21c article says the company was set up in a rented three-room apartment in San Jose, California, where internet access was cheaper; there is no other confirmation of this, and the press of 1998 called Mirabilis an Israeli startup, so the San Jose detail remains an estimate. The team worked in shifts around the clock for half a year, with no movies and no restaurants; as Goldfinger recalled, they simply worked all the time. The client came out in November 1996, free, with no advertising budget, starting from 40 of the founders' friends. The program is useful exactly to the extent that your friends are in it, which meant every new user had to genuinely invite people they knew rather than simply share a link. About six months later Yossi Vardi pointed out to the team that the product was turning into a phenomenon and predicted its growth rate with almost one hundred percent accuracy; the founders themselves had not yet grasped the scale of what was happening.

Year-by-year timeline

Lesser-known but significant facts

  1. By the time of the sale to AOL, Mirabilis did not have a single dollar of revenue, and that was a deliberate strategy. 70 employees, almost 13 million downloads, zero income: Yossi Vardi said plainly that the company had never tried to make money.
  2. The permanent UIN number gave rise to a spontaneous social hierarchy. A five-digit number meant you had been there from the very beginning, an eight-digit one that you had arrived last; the mechanic arose simply from sequential numbering, and nobody designed it estimate.
  3. The claim that Apple added support for the ICQ protocol is a common version, but an inaccurate one. In 2002 the original iChat let users sign in to AIM with an ICQ UIN not because Apple recognized the importance of someone else's network, but because ICQ and AIM already belonged to one company, AOL, and shared the OSCAR backend.
  4. In 2020, four years before the shutdown, ICQ had 439,000 genuinely active users against 11 million registered accounts. For comparison, WhatsApp in Russia at the time had 68 million monthly active users estimate.
  5. A six-digit ICQ number was a currency in the early Russian-speaking carding underground: people used such numbers to pay for services like drawing an avatar, and years later security researchers used these numbers to link the same nickname across different forums estimate.

Legend vs. the record

The first growth lever

The lever was not advertising but the mechanics of the product itself. ICQ left no choice: the program was useless until your friends were in it, so the first users had to genuinely invite people they knew rather than share a link just to tick a box. Steve Case, the head of AOL, described the result as early as 1998 as one of the fastest-growing and most loyal communities in cyberspace, before any advertising campaign on AOL's part. The growth curve held until 2003 with no advertising at all, because growth was deliberately put ahead of monetization from day one. A second, subtler lever was the UIN number itself: it did more than identify the user, since it turned into a status symbol (a way of saying you had been there since 1996), which added an emotional reason to stay on the network and show off your number to newcomers.

Parallels today (projects from the catalog)

The paired story

The pair for this season is IRC. The contrast between them is almost textbook. IRC (1988, Finland) is an open protocol that nobody owns: it can be neither bought nor shut down by a single decision of a board of directors, and the only thing that can be done is to break up an individual network inside it, which is what happened to Freenode in 2021, but IRC itself is alive in 2026 (see the IRC dossier). ICQ (1996, Israel) is its direct opposite: a private product that was sold twice as an asset, first to AOL for $287 million, then to Russia for $187.5 million, and that ended its life not on its own but by an administrative decision of its owner. The shared plot: both were born between the late 1980s and the mid-1990s as a simple answer to a local problem (a student chat on a bulletin board versus a friend-online notification for a group of friends), and both unexpectedly became global infrastructure. The lesson of the pair for the listener: who owns a technology determines not whether it takes off but how, and when, it ends.

What a builder can take from this in 2026

  1. A product that is useless without a network of friends makes its users invite real people on its own; this is stronger than any referral program, because the invitation is unavoidable rather than encouraged with a bonus.
  2. A deliberate refusal to monetize at the start is not a weakness but a strategy: ICQ had 13 million downloads and zero revenue at the time of its sale for $287 million, because what the buyer was paying for was the audience and the network, not current income.
  3. A simple identification mechanic (a permanent number) can become an emotional asset in its own right; today this is reproduced through early-bird pricing, low account numbers, and day-one-user badges.
  4. A second market can extend a product's life by years: in the West ICQ had fallen behind AIM and the Microsoft and Facebook messengers by 2010, yet under a Russian owner it kept running for another 14 years. It is worth looking not only for the market where a product will take off but also for the market where it will not die.
  5. A technology that belongs to a single company can be switched off by an administrative decision at any moment, even while it still has users: ICQ was shut down by its owner's decision, with a month's warning.

Discrepancies and what we could not verify

Sources (primary first)

Internet History Keeper

Every historical dossier here is free and open, and it will stay that way. If you want the series to keep going (new dossiers, checks against primary sources, the English edition), support it with a subscription for $9.90 a month. It is support, not access, and you can cancel at any time.

Become a keeper — $9.90 a month →

The same thing, about today

The same breakdowns, but of projects launching right now: what the product is, where the first users came from, how they charge. The card is free, the full dossier is $5 (the dossier itself is written in Russian).

Browse the dossier catalog →