← all stories
1996–2004IPO · 2004

Google

Two Stanford graduate students studied the link structure of the web as an academic problem, with no intention of building a business, and ended up with an algorithm more accurate than any search engine of the time. In six years they went from a student project that was nearly sold to a competitor for less than a million dollars to a company with an advertising model (AdWords) that lifted revenue almost fourfold (+302%) in a single year, and to an IPO worth $23+ billion.

Founders Larry Page · Sergey Brin
Domains google.com · google.stanford.edu (historical, 1997–98)
searchadtechseoauction-modelbackend-infrab2b2cplatform

In brief

Google began not as a business but as an academic problem for two Stanford graduate students: to work out which pages on the internet matter more than others if you look not at keywords but at who links to whom. A government grant, donated hardware from IBM and Intel, a conference paper — and only two years after the research began did a company grow out of it, one that in 1999 was nearly sold to a competitor for less than a million dollars. What followed was four years with almost no advertising budget, on pure word of mouth, embedding as a backend inside other people's portals (Netscape, Yahoo, AOL) and, finally, an advertising model copied from a competitor and taken to perfection, which lifted revenue from $86 million to $348 million in a single year (+302%) and carried the company to a $23+ billion IPO in 2004.

How it started (the founders)

Larry Page was looking at Stanford for graduate school in 1995; the person assigned to show him the campus was another future graduate student, Sergey Brin. Brin was born in Moscow in 1973 and emigrated with his family to the United States in 1979, at the age of six, escaping antisemitic discrimination in the USSR; he has said of it: "I feel like I have gotten a gift by being in the States rather than growing up in Russia." Page grew up in a family of CS professors at Michigan State University; his father, Carl Page, specialized in AI and died in 1996, the year the project began.

In 1996, by Page's account, the idea came from his advisor: "I started collecting the links on the Web, because my advisor and I decided that would be a good thing to do." The search engines of the day ranked pages by keyword frequency and broke easily: "If you typed 'Stanford,' you got random pages that mentioned Stanford." Brin joined almost immediately, not as a search specialist but as a data mining one: "we thought, 'Oh, we should be able to make a better search engine.'" The project was named BackRub, "a 'web crawler' which is designed to traverse the web," and from the very beginning it was not a business: the PageRank patent application was filed in Stanford's name with a priority date of January 10, 1997, and the domain google.com was registered on September 15, 1997, a year before the company appeared.

By the time the main paper, "The Anatomy of a Large-Scale Hypertextual Web Search Engine," was published in April 1998, BackRub had more than two authors: Brin and Page separately thank Scott Hassan and Alan Steremberg, calling their contribution "critical" and "irreplaceable," although neither became a formal co-founder of the company. The research was funded entirely not by angel investors but by the Stanford Integrated Digital Library Project, a grant from the U.S. National Science Foundation (NSF) with participation from DARPA and NASA; the hardware, a Sun Ultra II, Pentium II servers, and an IBM RS/6000, was donated by IBM, Intel, and Sun rather than bought with investment money.

Everything changed in August 1998, when Sun Microsystems co-founder Andy Bechtolsheim wrote a check for $100,000 after a short demo, made out to "Google Inc.," a company that did not legally exist at that point. The check sat in Page's desk for two weeks while the founders hurriedly registered the entity. Google Inc. was incorporated on September 4, 1998; the first full-time employee, on September 21, was Craig Silverstein, another Stanford CS graduate student, who dropped his PhD for the company. The first "office" was the garage of Susan Wojcicki's house in Menlo Park, at $1,700 a month; Wojcicki herself later became Google's 18th employee and then CEO of YouTube.

Year-by-year timeline

Lesser-known but significant facts

  1. Google was nearly sold to Excite for less than a million dollars, and not because of the quality of the search. In early 1999 Excite investor Vinod Khosla put the founders in touch with Excite CEO George Bell; by Khosla's own telling, Page and Brin asked for $1 million, then agreed to $750,000, and Bell turned down both prices. The reason for the refusal, in Bell's own words (a 2014 interview): "Larry said, 'If we come to work for Excite, you need to rip out all the Excite technology and replace it with Google's search.' And, ultimately, that's, in my recollection, where the deal fell apart." That is, the sticking point was the demand to replace Excite's search technology in full, not a comparison of search quality. Bell flatly rejects the version in which a test failed to show Google as fundamentally better for the ordinary user: "I think that's baloney. That's a factor that never occurred to me." That detail was in the analysis by mistake and was removed by the verification pass. The exact amount of the deal differs across three sources ($1M→$750K per Khosla; $750,000 plus about 1% of Excite per Bell himself; $1.6 million per Steven Levy's book) estimate.
  2. The $100,000 check was made out to a company that did not yet exist. In August 1998 Andy Bechtolsheim wrote a check to "Google Inc." after a short demo, but no such legal entity existed yet. The check lay in Page's desk for two weeks while the founders hurriedly completed the incorporation (which took place on September 4, 1998) so that they could cash it at all.
  3. The founders' own 1998 paper warned about the very flaw the business was later built on. Brin and Page wrote plainly in their academic publication: "we expect that advertising funded search engines will be inherently biased towards the advertisers and away from the needs of the consumers," giving the example that a search about cell phones honestly surfaces research on the danger of talking while driving, which is exactly what an advertiser would not like. Two years later they launched AdWords.
  4. The PageRank patent belonged not to Page but to Stanford, and it brought the university $336 million. Because the research was part of the Stanford Digital Library Project grant, the patent was filed in the university's name; Google bought an exclusive license for 1.8 million shares of the company. Stanford sold part of the package during the 2004 IPO itself and earned $336 million on those shares in total.
  5. Jeff Bezos personally put $250,000 into Google as early as 1998, a year before the institutional round, and he was not alone. The investment was made at about 4 cents a share, and Bezos explained it to the journalist Ken Auletta for the book "Googled: The End of the World as We Know It" simply: "I just fell in love with Larry and Sergey." At least three others put the same amount into the same early round: Stanford CS professor David Cheriton, investor Ram Shriram, and Andy Bechtolsheim (the same man who wrote the $100,000 check). By the 2004 IPO, Bezos's stake had turned into 3.3 million shares worth about $280 million.

The first growth lever

For the first two years after incorporation Google had no marketing budget; growth ran on two mechanics, and neither of them was advertising. The first was product quality plus word of mouth: in the press release about the $25 million round (June 1999) the company reports that traffic is growing 50% a month "word-of-mouth only." The second, and strategically the more important one, was embedding as a search backend in other people's properties with a ready-made audience instead of building an audience of its own. In June 1999 Netscape (by then part of AOL) switched its search to Google technology (60+ million pages in the index); a year later Google took the same role at Yahoo!, then the largest portal on the internet; in May 2002 the same thing happened again with AOL Time Warner, this time together with AdWords. Google did not spend a cent on acquiring end users; they were brought in by the partner that already had the traffic.

The second lever, the monetization one, switched on in 2002. AdWords, launched in October 2000, was an ordinary CPM model: payment for impressions, not a breakthrough. The breakthrough was AdWords Select (February 20, 2002): payment only for a click, with the position of an ad determined by the bid multiplied by CTR, that is, by relevance and not by money alone. The model was copied from GoTo.com (Overture from 2001), which since 1998 had been the first to sell places in search results by cost-per-click bids; Google improved it with a relevance factor and got a patent suit for it on April 5, 2002, 44 days after the launch. The effect was abrupt: according to Google's own form S-1 (a primary source, with three independent tables inside the document giving the same figure), revenue grew from $86.4 million (2001) to $347.8 million (2002), up 302% year over year, meaning revenue rose roughly fourfold rather than fivefold as had been assumed before the check against a secondary source. The suit was not settled until August 2004, for 2.7 million Google shares, but by then the auction model had already done its work.

Parallels today (projects from the catalog)

What a builder can take from this in 2026

  1. A research problem with no business goal can become a company, but it does not have to be designed as one from the first day. BackRub existed for two years as an NSF/DARPA/NASA grant without a single thought about monetization; the business model appeared four years after the research started.
  2. Embedding as the backend of someone else's platform with ready-made traffic is cheaper than building your own audience. Three partnerships in a row (Netscape → Yahoo → AOL) gave Google millions of users without a dollar spent on acquisition; the company was selling invisible infrastructure under someone else's brand.
  3. Copying someone else's business model is fine if you improve it at its main weak point. GoTo/Overture came up with the cost-per-click auction back in 1998; Google added ranking by relevance (CTR × bid) rather than by the size of the wallet alone, and it was that improvement, not the borrowing of the idea, that drew the patent suit.
  4. A deal that looks bad today can turn out to have been decisive several years later, and the reverse is also true. Excite's refusal to buy Google for less than a million dollars in 1999 (the exact amount and structure are disputed among the sources, though the refusal itself is not) became one of the most expensive mistakes in the history of technology; at the same time, the founders' willingness to sell the company that cheaply shows how heavily "confidence in success" is exaggerated in hindsight. The real reason for the refusal was the unwillingness to change Excite's technology completely, rather than a comparison of search quality (which Bell himself later refuted): the lesson is more about deals falling apart from operational and political resistance more often than from a cold assessment of the product.
  5. An idealistic principle put down in writing at the IPO is an asset too, not only a declaration. "Don't be evil" appeared as an internal formulation at a meeting about the company's values somewhere between 2000 and 2001 (the exact date is disputed, and the dedicated investigation by Quote Investigator leans toward July 19, 2001), and it was publicly on Google's site by March 2002, which means the 2004 IPO made the phrase world-famous but was not its first public disclosure, contrary to an earlier version of this analysis.

Discrepancies and what we could not verify

Sources (primary first)

Primary (period documents and the founders' own words):

Secondary (analysis, journalism, and encyclopedic biographies):

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 →