CompuServe

The first mass-market commercial online service in the United States: in 1969 an insurance company decided to sell the hours its mainframe sat idle in the evenings to private individuals, and out of that grew forums, chat, three million subscribers by the middle of the 1990s, the GIF format, and the first court case about platform liability. In 1998 the business was divided between WorldCom and AOL; the brand lasted as an internet provider until 2009.

In brief

CompuServe was the first truly mass-market commercial online service in the United States: out of an insurance company's side project to monetize mainframes that sat idle in the evenings (Columbus, Ohio, 1969) it grew into a network of forums, chats, and mail with three million subscribers around the world connected to it by the middle of the 1990s, before most of them had heard the word "internet." The service invented the GIF format, lived through one of the first court cases in history about who answers for someone else's words on a forum, and in 1998 broke into two parts in a three-way deal: the telecom giant WorldCom took the network infrastructure, and America Online took the brand along with the subscribers. The brand itself lasted until 2009 as an internet provider and until 2017 as a forum.

How it started (the founders)

In 1969 the Columbus insurance company Golden United Life Insurance (a document of 1979 names the holding company as Ilex Corp., see "Discrepancies") discovered that the mainframe it had recently bought sat idle in the evenings and on weekends, since the calculations ran only during business hours. Golden United's founder, Harry Gard Sr., called in his son-in-law, Jeff Wilkins, a master's student in electrical engineering in Arizona, to build the company a proper computing infrastructure. Wilkins brought in a classmate, John Goltz, someone with real hands-on technical skill. The first president of the new subsidiary, Compu-Serv Network, was Goltz and not Wilkins, contrary to the common version in which Wilkins alone counts as the founding father; Wilkins took over from Goltz as CEO only within the first year.

For its first decade Compu-Serv (CompuServe Incorporated from 1977) lived the dull life of B2B timesharing: it sold machine time to corporations and government bodies while building its own packet-switched network. The idea of turning those same idle evening hours toward private individuals came only in 1979, and it turned out to be not all that original: The Source, the service of Bill von Meister, had taken the same niche two and a half months earlier (Reader's Digest bought into its capital only in September 1980).

Year-by-year timeline

1969: Compu-Serv Network is founded in Columbus as a subsidiary of the insurer Golden United Life Insurance, to monetize the idle evening hours of a mainframe fact
1975-06-27: CompuServe is separated from the parent holding company into a public company of its own, trading on NASDAQ under the ticker CMPU fact
1977: renamed CompuServe Incorporated, building its own packet-switched network estimate
1979-09-24: the consumer service MicroNET goes to retail through the Radio Shack chain: an evening dial-up by modem, $5 an hour plus the telephone call fact
1979-11 / 1980: H&R Block agrees to buy CompuServe for ≈$23 million in stock and cash ($20 in cash per share, or 0.8 of a Block share); the deal closes in 1980 fact
1980-02-21: CB Simulator, the first mass-market multi-user chat, written by Sandy Trevor over a single weekend on the metaphor of CB radio fact
1980-07-01: MicroNET is renamed CompuServe Information Service; The Columbus Dispatch is among the first newspapers available through a home computer estimate
1987-06-15: the GIF format: a team led by Steve Wilhite solves an internal CompuServe problem, compressing graphics for slow modems, on an assignment from the manager Sandy Trevor fact
1991-10-29: the court in Cubby v. CompuServe finds the company a "distributor" rather than a "publisher" of other people's content on the forums fact
1994–1995: the conflict over the Unisys patent on the LZW algorithm inside GIF; in January 1995 CompuServe introduces a token license ($1 plus up to 1.5% per copy), and the free PNG format is born in answer within a few months fact
1995-04-28: the peak, 3 million subscribers around the world fact
1995-12: under pressure from the Bavarian prosecutor's office, CompuServe blocks access to more than 200 usenet groups for a month and a half, for every subscriber in the world and not only in Germany; the director of the German office, Felix Somm, is put on trial (conviction in May 1998, acquittal in November 1999) fact
1996-04-19: IPO on NASDAQ (CSRV): 18.4 million shares at $30, with H&R Block keeping 80.1%; for the fiscal year just ended, $793 million of revenue and 5 million subscribers counting licensees fact
1997-09-08: WorldCom announces an agreement to buy all of CompuServe from H&R Block for ≈$1.2 billion in stock fact
1998-01-31: the deal closes: AOL trades ANS Communications for CompuServe's consumer business and $175 million ($162 million at closing) fact
2009-06-30: CompuServe Classic is shut down as an internet provider after 30 years of continuous operation; the forums under the same name will run another eight years, until 2017

Lesser-known but significant facts

The company's first president was not Jeff Wilkins but John Goltz. Wilkins, who today is almost always named as the sole founder, took his place only within the first year of operation.
GIF was not given away free out of altruism; it is simply that until 1994 nobody suspected there was someone else's patent inside it. When Unisys claimed rights to LZW, CompuServe introduced a token fee in January 1995 ($1 once plus up to 1.5% per copy) and wrote plainly in its memorandum that it would earn nothing on the service itself, since the money went to Unisys. Even a levy that modest turned out to be reason enough for the free PNG to be born within a month: the first draft came out on January 4, 1995, and CompuServe announced PNG support as early as February 7.
The company's first public spin-off nearly happened 20 years before it was sold to AOL, and was cancelled by H&R Block's board of directors four months after the IPO of 1996, citing CompuServe's losses and the uncertainty of the market for online services estimate.
In 1995 CompuServe physically had no way to block usenet groups for one country alone, so the Bavarian prosecutor's demand to shut off access to the material for Germany turned into a block for every subscriber on the planet for a month and a half.
Its own attempt to copy AOL's model failed even before the sale to WorldCom. In March 1996 CompuServe launched the flat-rate service WOW! ($17.95/month) and closed it 10 months later (January 31, 1997), having signed up 102,000 subscribers against 2.2 million on the main service and having put in $50+ million.

Legend vs. the record

Legend: CompuServe was the first online service for private individuals. That is what almost every headline about the company says, including its own retrospectives. The record: The Source, the service of its founder Bill von Meister rather than of a publisher (Reader's Digest took a 51% stake only in September 1980, more than a year after the start), went live in July 1979, two and a half months before CompuServe's consumer MicroNET (September 24, 1979). Verdict: CompuServe did not invent the category, it won in it, and won literally: in 1989 it bought The Source together with its ≈53,000 remaining subscribers. The fame of being first went to the winner after the fact rather than to the pioneer.
Legend: GIF was invented for the web. The format is so bound up with internet memes that it feels like a web invention. The record: the GIF87a specification is dated and copyrighted by CompuServe on June 15, 1987, two years before Tim Berners-Lee's proposal for the WWW (March 1989) and four years before the public web. Verdict: GIF was created for CompuServe's internal needs, graphics over slow modems on its own network, and the web simply picked up a format that was already there.
Legend: liability of online platforms for someone else's content is an invention of Section 230 (1996). The popular history of internet law usually starts with that section. The record: back in 1991, five years before Section 230, the court in Cubby v. CompuServe had already solved a similar problem with ordinary case law, finding CompuServe a "distributor" rather than a "publisher" of forum content and freeing it from liability for what the company could not moderate in advance and was not obliged to. The problem was created not by that principle but by its side effect: in 1995, in Stratton Oakmont v. Prodigy, the same kind of court effectively punished another platform for the opposite, for an attempt to moderate content, which got it treated as a "publisher" and held liable. Verdict: the doctrine of platform liability began in 1991 with CompuServe; Section 230 in 1996 did not create it from nothing, it repaired the incentive, spoiled by that same logic, to moderate nothing at all.
Legend: CompuServe lost to AOL. A simple story of defeat is convenient but inexact. The record: in the real deal of 1997–1998 the main buyer was not AOL but the telecom operator WorldCom, and it bought CompuServe whole precisely for the network infrastructure (CompuServe Network Services), which it kept. The consumer brand with its subscribers WorldCom immediately traded to AOL for the ANS network division in a separate deal (closed 1998-01-31). Verdict: CompuServe did not so much lose to AOL as turn out to be a valuable asset in the eyes of a telecom company for which a retail online brand was a bargaining chip rather than a prize; the brand itself went on living under AOL's wing another 11 years as a provider and 19 as a forum.

The first growth lever

The lever was arbitrage on the price of a single asset: Golden United's mainframes were loaded with the insurance company's calculations by day and sat idle at night and on weekends. Instead of adding corporate customers, in September 1979 CompuServe opened those same hours to retail buyers under a separate brand, MicroNET: $5 an hour plus the telephone call, access only outside business hours. The key was not the pricing in itself but the channel: instead of building its own retail sales network for an audience it had never thought about before, MicroNET was sold through thousands of already existing Radio Shack stores, ready-made retail with a ready-made flow of buyers of home computers. CompuServe had fewer than a thousand private subscribers when H&R Block bought it in 1980, and the growth of the first years was a slow start rather than an explosion. The real acceleration came from what grew on top of the evening hours: from 1980 the forums and CB Simulator turned dialing in from buying machine time into coming to talk, and since billing was hourly, every minute in a chat or on a forum was a directly monetizable minute rather than a free extra on top of the product. By 1987 CompuServe had 380,000 subscribers, growth of hundreds of times over eight years on a model of somebody else's retail plus hourly billing for conversation.

Parallels today (projects from the catalog)

Midjourney (midjourney) is the same move as the forums with CB Simulator: Midjourney launched Discord-first instead of a site built from zero, taking the ready-made social graph of somebody else's platform, the way CompuServe in 1979 used Radio Shack's retail instead of sales channels of its own. The difference between the eras: for CompuServe conversation was a direct source of revenue (hourly billing), while for Midjourney the public channels are free for the company and monetized through a separate subscription; the function is the same, in that every action a user takes in front of others is a demo and an engagement at once. view this project's dossier →
Prelint (prelint) is a direct parallel to the most contested place in CompuServe's business model, payment by actual use (pay-per-use, credits) rather than a flat subscription: the same trade-off that in the end lost CompuServe the price race with AOL and its flat rate in the middle of the 1990s. The difference is that for Prelint the usage-based model is a deliberate choice for a B2B tool with uneven consumption, not a legacy of an era when there was simply nowhere, technically, to arrange a flat rate. view this project's dossier →

What a builder can take from this in 2026

Look for the price of an idle asset, not only the price of a loaded one. CompuServe had the same machines by day and by night, but two segments of demand and two prices. Today that is the evening or batch windows of GPU inference, or under-used licenses: before buying more capacity, sell the idle time you have already paid for to a patient customer more cheaply.
Do not build retail from zero if it already exists at somebody else's. Radio Shack gave CompuServe thousands of points of sale instantly; today that is marketplaces, the App Store, Discord, Slack: the road into somebody else's assembled audience is almost always shorter than the road to your own.
If people pay for time or for use, conversation is the product and not a load on support. For CompuServe the forum was retention and direct hourly revenue at the same time. With usage-based prices in 2026 that means designing community features so that they give rise to more paid minutes or tokens rather than merely reducing churn.
Open up what is not your product, because it may outlive the product. GIF was an internal instrument of CompuServe's network rather than something the company sold; open and free at the moment of its release, it lives on decades after CompuServe disappeared. Infrastructure byproducts (formats, SDKs, protocols) are worth releasing into the world freely rather than locking up together with the main product.
A regulatory norm in one jurisdiction is able to switch a product off for everyone, so build in local blocking in advance and not only global blocking. The story of the Bavarian prosecutor's office in 1995 is a scenario that any regional regulation of AI and content reproduces in 2026: without ready geo-segmentation a product has no middle option between breaking a local law and turning a feature off for everybody in the world.

Discrepancies and what we could not verify

CompuServe's parent holding company in 1969–1975: Golden United Life Insurance or Ilex Corp? All the retellings name Golden United; a contemporaneous document of 1979 names the parent company directly as "Ilex Corp., an Ohio-based life insurance holding company." A two-level structure (an operating insurer under a holding company) is likely but not confirmed; the fact of the founding in 1969 under the Gard family is not disputed unverified.
The exact day (not month) the H&R Block/CompuServe deal closed. The document of November 1979 planned for early 1980; part of the retellings name May 1980, and the discrepancy of ~4 months is unexplained.
The text of the Merger Agreement itself and of the SC 13D of 1997 (WorldCom/AOL) still has not been read directly; the dates and the sums are now confirmed from the 8-K with the press release and from the notes to AOL's audited 10-K, sources more reliable than the press at the moment of the announcement.

Sources (primary first)

Secondary (context, cross-check):