← all stories
1994–1997IPO on Nasdaq · 1997

Amazon

Cadabra → Amazon.com

An online bookstore that started in a garage in Bellevue under the name Cadabra and nearly kept that name, but for a lawyer who misheard it as cadaver. In three years, from July 1994 to May 1997, it went from a list of 20 ideas and $1 million from his parents and ~20 acquaintances to an IPO on Nasdaq, building Associates along the way, the first truly mass affiliate program in the history of the internet (1996). It is alive and has grown into one of the largest companies in the world; its first GAAP-profitable quarter came only at the end of 2001.

Founders Jeff Bezos (the only officially recognized founder) · Shel Kaphan (the first employee, hired before the company was registered; he describes himself as a co-builder, formally not a co-founder)
Domains amazon.com · relentless.com (registered earlier as one of the candidate names, September 1994; still redirects to amazon.com)
ecommerceaffiliatereferralmarketplaceretaildistribution-partnershipslogistics

In brief

Amazon began not with the idea of selling everything on earth but with cold arithmetic: Jeff Bezos picked books as the product because they are light, do not spoil, and already had a wholesaler's ready-made database behind them, an ideal first wedge for online retail. In three years the company went from a garage startup with desks made of doors taken off their hinges to an IPO on Nasdaq with revenue of $147.8 million, and along the way, almost by accident, it built the first truly mass affiliate program in the history of the internet. Four and a half years after the IPO, at the end of 2001, Amazon posted its first GAAP-profitable quarter, proving to the skeptics that a bet on scale mattered more than a bet on quick profit.

How it started (the founders)

At 30, Jeff Bezos was a vice president at the investment firm D.E. Shaw on Wall Street: successful, well paid, with no reason to leave. In 1994 he came across a statistic about internet usage growing 2,300% a year and realized this was a rare window of opportunity. He made the decision through what he himself called the "regret minimization framework": he projected himself 80 years forward and asked which he would regret more, having tried and failed or never having tried at all. The answer was obvious: "I knew that when I was 80 I was not going to regret having tried this." He told his boss about his plans to open an online bookstore; his boss took him on a two-hour walk through Central Park and in the end agreed the idea was a good one, but advised him to think for 48 hours before giving up a secure career.

Bezos wrote out a list of 20 product categories that could in theory be sold over the internet and narrowed it to five: compact discs, computer hardware, software, video, and books estimate. Books won: a huge global market, a low price point, an enormous number of titles and, importantly from an operational standpoint, a ready-made database of every book in print at wholesalers such as Ingram estimate. With his wife MacKenzie he drove from Fort Worth to Seattle, writing the business plan along the way while she drove estimate. Seattle was chosen for a combination of reasons: the small population of Washington state exempted almost 99% of US buyers from sales tax; Ingram's book distribution warehouse in Oregon was nearby; and the city had enough technical talent thanks to Microsoft and the University of Washington estimate.

On July 5, 1994, the company was formally incorporated in the state of Washington under the name Cadabra, Inc., from "abracadabra." The first employee (in effect a co-builder, though formally not a co-founder) was Shel Kaphan, a programmer from Santa Cruz who came to Seattle in October 1994, before the company was registered; a month later a second technical hire followed, Paul Davis from the University of Washington. The office was a converted garage and house in Bellevue; Bezos's wife MacKenzie wrote checks from time to time and kept the early books. That same November 1994, a lawyer handling the registration heard "Cadabra" as "cadaver," and Bezos decided the name would not survive, renaming the company Amazon.com after the largest river on earth, a symbol of the scale of the catalog. The amazon.com domain was registered on November 1, 1994, two months later than relentless.com, which Bezos had registered in September as one of the candidate names and which still belongs to Amazon.

The funding at the start was not venture capital but personal connections: about 60 meetings with relatives and acquaintances to raise $1 million at roughly $50,000 a person for 1% of the company. The number who agreed is given by various secondary sources as 20 (CNBC) or 22 (Yahoo Finance, South China Morning Post, and Tom Alberg's obituary in Bloomberg); among them were Bezos's parents, who put in $245,573 (he warned them honestly about a 70% chance of losing everything), and, according to SCMP, Bezos's younger brother and sister, Mark and Christina estimate. The primary document gives a third number: per form S-1 (Item 15, "Recent Sales of Unregistered Securities"), the largest tranche that matches by amount is 3,021,000 shares to 23 investors for $1,007,000, executed not all at once but as a series of purchases from December 6, 1995 to May 16, 1996 (two smaller tranches came earlier still: 3 investors in February–July 1995 and 1 investor in August 1995). The popular version, that the round closed in early 1995 with about 20 people, is more accurately described as money from friends and acquaintances being documented in shares in stages over a year and a half, from February 1995 to May 1996; 20, 22, and 23 do not agree with one another, but the amounts and their sequence are the same. The only outside venture investor and the first member of the board of directors was Tom Alberg; separately from him, in June 1996 Kleiner Perkins (John Doerr) put in $8,000,014 for preferred stock convertible into 3,416,376 common shares, Amazon's first true venture check.

Year-by-year timeline

Lesser-known but significant facts

  1. Amazon's "first book" was a test purchase by an acquaintance of a company employee, more than three months before the public launch (with a caveat). On April 3, 1995, John Wainwright, a friend of Amazon's first employee Shel Kaphan, presumably bought the Hofstadter book through the site while it was still not open to the public; Wainwright confirmed the date from his own order history, and a building on the Amazon campus bears his name today. An important caveat: Smithsonian Magazine, the primary source for this story, writes plainly that "no one has entirely confirmed that Wainwright is the true customer": the story goes back to a post on Quora, and the buyer's identity has not been fully confirmed. The "official" version of the company's history usually names July 1995, the moment of the public launch, as the first sale estimate.
  2. The name Amazon came about because of a lawyer's warning. The company was going to be called Cadabra (from "abracadabra"), but the company's lawyer Todd Tarbert pointed out that the word sounds like "cadaver," and Bezos decided a name with that echo would not survive. Other rejected names included Relentless, Awake, Browse, MakeItSo (a nod to Star Trek), and Aard (to land at the top of alphabetical listings).
  3. The famous door desks were not a marketing legend but literal thrift on furniture. In the summer of 1995, at a house across from a Home Depot, Bezos discovered that a door with legs cost less than a desk, and he bought the door. The story was confirmed independently by two early employees (#5 and #6) on Amazon's own corporate blog. Today Amazon gives teams the "Door Desk Award," a miniature signed replica, for cost-saving ideas.
  4. Barnes & Noble sued Amazon literally three days before the IPO, over the use of the phrase "the world's largest bookstore," insisting that Amazon was not really a store but a book broker (even though only a few hundred titles were physically held in Amazon's Seattle warehouse against 170,000+ at B&N, with orders going straight from the wholesalers). The suit was settled without a single dollar of compensation five months later, when both sides decided it was more profitable to compete in the market than in a courtroom.
  5. Amazon was neither the first online bookstore nor the first affiliate program on the internet, only the first to make both things mass-market. Charles Stack's Book Stacks Unlimited opened in Cleveland as a BBS store in 1992 (Stack had the idea a year earlier), three years before Amazon began selling books (1995). Affiliate programs paying a commission existed before Associates as well: CDNow launched its BuyWeb program in November 1994, almost two years before Amazon; William J. Tobin created an early referral sales model for PC Flowers & Gifts that industry reviews of affiliate marketing usually date to 1989, but that does not hold up. Tobin's only verifiable patent on the subject (US6141666A) was filed in January 1997, and Wikipedia gives 1994 rather than 1989 as the founding year of PC Flowers & Gifts itself; the sources on the history of the industry disagree among themselves here estimate. Associates (July 1996) was in any case the first such program available publicly and at scale to any site owner, and it was the one that set the standard for the industry.

The first growth lever

Amazon's growth in 1995–1997 was not one trick but a sequence of three steps, each of which converted the previous one into a more scalable version. Step one was free distribution: almost immediately after the launch in July 1995 the site landed in Yahoo's curated "What's Cool" list and Netscape's "What's New" list, and in the very first month orders came from all 50 US states and 45 countries, while by September weekly sales had reached $20,000. That was a one-time piece of luck that could not be switched on at will; Amazon did not buy it and could not reliably repeat it.

Step two turned that one-time luck into a permanent, self-service channel: in July 1996 the Associates Program launched, and any site owner could put up a link to a specific book and earn a commission on the sale, while Amazon took on the whole order, the payment, and the delivery. The first known participant count, the earliest and the least widely circulated, comes from the original IPO filing: "over 4,800 registered members as of December 31, 1996," that is, less than six months after launch. Growth after that is documented in three of Amazon's own press releases: by September 1997 more than 15,000 partner sites, by February 1998 30,000 (with commissions of up to 15%), and four months later, in June 1998, 60,000. Not one of those primary sources, including the description of the program in the S-1 itself, gives the starting commission as a number; 3–8% appears in only one secondary source estimate. The mechanics were about as simple as that era allowed: Amazon converted the very structure of hyperlinks on the web into a sales channel, paying a commission not for a click and not for an impression but only for a completed sale, a model that had existed before (CDNow certainly from 1994; possibly Tobin's PC Flowers & Gifts earlier, though the exact date is disputed, see "Lesser-known but significant facts" #5) but not at public, self-service scale.

Step three turned the free step one into a paid one, at least as one secondary source (FundingUniverse) reads it: around the middle of 1997 Amazon signed formal partnerships with Yahoo! and America Online, the same portals that two years earlier had pointed to the company for free, and got broad promotional placement instead of a one-off mention estimate. The list of partners is confirmed by the primary 1997 letter to shareholders as well ("established long-term relationships with many important strategic partners"): it later expanded to Netscape, GeoCities, Excite, AltaVista, @Home, and Prodigy. But the letter, unlike FundingUniverse, nowhere calls those partnerships part of the Associates Program or a source of commission income. For the best publisher partners a premium rate was launched: the top 500 sites in the PC Meter rankings received a 22.5% commission on most of the catalog for a six-month bonus period. The same asset, someone else's audience, Amazon first got for free as luck, then scaled through the self-service of thousands of small sites, and then bought wholesale from the largest players, once it had something to pay with.

Parallels today (projects from the catalog)

What a builder can take from this in 2026

  1. You do not have to be first; you have to be first at scale. Book Stacks was selling books online three years before Amazon, and CDNow launched a commission affiliate program at least two years before Associates (November 1994 against July 1996), but it was Amazon that took both ideas to the point where anyone could use them. An idea is almost never invented from scratch; it is scaled.
  2. Free distribution is a signal, not a strategy. Landing in Yahoo's "What's Cool" gave the first burst of traffic, but it was precisely because the event could not be repeated that the company spent a year building Associates, a reproducible channel in place of a one-time piece of luck.
  3. A negotiating position is built from the bottom up. Amazon got the attention of Yahoo and Netscape for free in 1995, and in 1997, having accumulated traffic and data, it bought far more expensive promotional placement from the same portals: first a free mention, then a paid partnership.
  4. Cultural symbols are cheaper than they look and work for years afterward. A desk made from a door cost an order of magnitude less than a normal one, but it became an institutionalized symbol of frugality (the Door Desk Award); an operational decision turned into a narrative asset.
  5. A public commitment to thinking long-term protects against market pressure. The 1997 letter warned investors plainly that the company would prefer cash flow to attractive reported earnings, and that position, stated in advance, gave it the right to grow losses for four years for the sake of scale before showing its first profit at the end of 2001.

Discrepancies and what we could not verify

Sources (primary first)

Primary (period documents and the founder's own words):

Secondary (business journalism, specialized historical reference resources):

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 →