What the machine remembers

For two centuries, Encyclopedia Britannica built one of the most successful direct sales forces in American history by targeting middle-income families with a proposition equal parts aspiration and guilt: the greatest gift you could give your children was access to knowledge. By 1990, the company was selling 120,000 sets a year and generating $650 million in revenue. A 32-volume set cost $1,500 to $2,200. The company had not had an unprofitable year since the 1930s.

Britannica was, in the most literal sense, the original answer engine. Someone had a question. They went to Britannica. The company had spent more than two centuries making itself the answer, across every topic, in every home that aspired to know. The brand was older than the United States. It felt like it would outlast everything.

The Offer

In 1985, a Microsoft executive approached Britannica with a proposal to license their content for a multimedia encyclopedia on CD-ROM. The company refused. As Britannica's director of public relations stated that year, "The Encyclopedia Britannica has no plans to be on a home computer." The home computer market was small. Only 4 or 5 percent of households had one. A digital product sold cheaply would undercut the premium print sets that the sales force was built to move. Protecting the sales force meant protecting the format.

What complicates the story is that Britannica was not oblivious to digital media. The company actually launched the first multimedia CD-ROM encyclopedia in 1989, four years before Microsoft entered the market. But they released it under the Compton's brand name, a less prestigious encyclopedia they owned, specifically to keep the Britannica name from association with a cheaper medium. They had the technology and deliberately hid it. The mission was right. Protecting the format above all else was what failed them.

Microsoft went to Funk and Wagnalls instead, licensed their content, and began building what would become Encarta.

The Collapse

Encarta launched in 1993 at $395, dropped to $99 within months, and was eventually bundled free with new computer purchases. The marginal cost to Microsoft of giving it away was roughly $1.50. The marginal cost to Britannica of selling a print set was over $750, including commission.

Revenue fell 50 percent in three years. By 1994, the company was selling only 51,000 sets. Its belated response illustrated exactly why the format had become the trap. Britannica finally launched its own CD-ROM in 1994, priced at $995, ten times the cost of Encarta. The price was set to avoid cannibalizing print sales and keep the sales force from revolting. It solved an internal problem and accelerated the external one.

By 1995, the company put itself up for sale. Microsoft declined. So did every technology and media company approach. In 1996, financier Jacob Safra purchased Britannica for $130 million, less than half its book value. The sales force was dissolved. The brand that had defined authoritative knowledge for two centuries was nearly worthless.

The Survival

Most people who know this story assume it ends there. It does not.

The new ownership did what the old company could not: it separated the mission from the format. Britannica pivoted to institutional subscriptions, selling access to its content to schools and libraries at prices described by one executive as about the same as a glass of milk per subscriber. When the company ended print entirely in 2012, less than 1% of its revenue came from physical books.

By 2022, revenue was approaching $100 million, with more than 500,000 institutional subscribers and 100 million students worldwide. The company, now rebranded simply as Britannica, is pursuing an IPO. The reported valuation: nearly $1 billion, driven by a pivot to artificial intelligence.

The irony is complete. AI companies building large language models need one thing above all else: vetted, authoritative, human-edited content. Models trained on the raw internet produce hallucinations because the raw internet is full of errors and misinformation. Britannica's 256-year archive of expert-reviewed knowledge is exactly what the machines need to produce reliable answers. The company that was nearly destroyed when the medium for answering questions changed is now a valued supplier to the systems that broker answers today.

What the Mission Was Actually For

Britannica's failure was not a failure to understand digital technology. The company launched a CD-ROM encyclopedia in 1989. It built an online version in 1994. It understood the technology. What it failed to do was separate the mission from the format. The mission was to curate authoritative answers to questions people actually ask. The format was print. When protecting the format became the priority, the mission became impossible to serve.

What the machine remembers is not the format. It does not care whether an answer arrived in 32 leather-bound volumes, on a CD-ROM, or through an institutional subscription portal. It cares whether the content is trustworthy, specific, and authenticated. Those qualities survived every disruption because they were never about the medium. They were about the mission.

Britannica confused the two for a decade. When it finally stopped, it turned that mission into a company worth nearly a billion dollars.