9

Virulent Word of Mouse

The Internet is a telephone system that’s gotten uppity.

—Clifford Stoll¹

Electronic mail does not resemble Tupperware in any tangible dimension. Yet in 1995, Tim Draper saw a connection, and his flash of insight came at a propitious moment, when commercial motives would predominantly begin to shape the rate, direction, and tone of collective invention for Internet applications and infrastructure. Tim Draper had taken a circuitous path to this moment. In 1985 Draper formed a venture capital (VC) firm in Menlo Park, California, eventually partnering with John Fisher and Steve Jurvetson to form Draper, Fisher, and Jurvetson, or, DFJ, which specialized in funding entrepreneurs aspiring to start new firms. The company especially sought intrepid entrepreneurs who shot for the moon. Not a business for the faint of heart in practice, DFJ took many big risks on many big bets, so that the entire return did not depend on only a few eggs in one basket. In practice, however, usually only a few big winners emerged, and these generated large returns for the company’s entire portfolio of investments.

In the summer of 1995, DFJ backed a start-up by two unknown entrepreneurs, Sabeer Bhatia and Jack Smith. The firm was called HoTMaiL, embedding HTML in the firm’s name.² Bhatia and Smith did not reveal their business idea to DFJ right away. They were paranoid about venture capitalists—not of DFJ in particular, but of anyone and everyone.

Bhatia and Smith had met while working at Apple Computer, and both later left to work at a start-up called FirePower Systems. The idea of HoTMaiL came to them while they worked at FirePower. Aspiring to start a company, they could not e-mail back and forth without increasing the risk that their supervisor would see them discussing nonwork matters. They each had AOL e-mail, but that system also raised risks, because it relied on software on the desktop—that is, the software had to reside within the PC—and the office did not permit AOL software on its systems.³ This restriction helped generate Smith’s idea to design e-mail accounts that were accessible over the web, using any browser from any location.⁴ Because both entrepreneurs believed that other Internet users faced the same problem that they themselves encountered at work, Bhatia and Smith thought that a general service of web-based e-mail would have wide appeal. They quickly developed a business plan.

Their paranoia and sense of urgency emerged after they hatched their plan. The two entrepreneurs correctly assumed that any technically skilled programmer familiar with the growing web could understand the plan, and, therefore, imitate it. The key idea was rather easy to copy once communicated to a knowledgeable listener, as would happen during a presentation to VCs. If Bhatia and Smith made the presentation but did not receive funding, then they would have given their idea to someone else who could copy it. In fact, they had more reason to be paranoid than they knew: other firms and inventors were independently pursuing similar ideas at the same time, though most of these efforts were not widely known.⁵

Bhatia and Smith decided to test the business savvy and trustworthiness of the VCs by using a fake business plan. They invented a Potemkin business.⁶ If the two of them liked how the VCs reacted to their Potemkin business, then they would reveal their actual plan. The fake concerned an idea for a business that Bhatia and Smith had previously considered and rejected.⁷

Bhatia and Smith persevered in the face of twenty rejections, or in their view, twenty VCs to whom they would not reveal their true aspirations. At that point they reached DFJ and ran through the routine in front of Tim Draper and the intellectually peripatetic Steve Jurvetson. The VCs were not impressed, but in Bhatia’s view, Jurvetson did something no other VC had done. He asked the right questions about the Potemkin business. The two entrepreneurs indicated to each other that they were satisfied, as the VCs impatiently got ready to dismiss them if there was nothing else. At that point, Bhatia and Smith revealed their big idea for web-based e-mail, which struck a chord with the VCs. The conversation became animated. And somewhere in the ensuing cacophony, electronic mail met Tupperware.

Tim Draper thought back to a case he had seen in business school about Tupperware’s marketing arrangement.⁸ Tupperware had learned to inexpensively distribute and market its products by circumventing expensive retail distribution channels by working with homemakers, who acted as the sales force and were paid on commission. A representative would invite friends to a “Tupperware party.” Friends would invite more friends, and so on. It involved no expensive retail outlets or ad campaigns, and it required minimal regular staff.

As Draper listened to the entrepreneurs, he thought about how e-mail could emulate Tupperware by circumventing standard distributional channels for computer software, and thereby cutting out a big potential expense while remaining very effective. What if a user of an e-mail service could invite another friend to use the same service, just as the Tupperware sales force could invite their friends? Draper voiced a way to implement it. He suggested placing the URL of the e-mail service in the footer of every electronic mail message, letting the social network of each user help spread the service. He suggested that the hyperlink at the bottom of every e-mail contain a message, “P.S. I love you. Get your free Web-based email at HoTMaiL.” Clicking on the hyperlink would take the user to HoTMaiL’s site to sign up for the service.

Initially Bhatia and Smith pushed back, liking neither the idea nor the specific phrasing. They had many reasons to resist. For example, the idea mixed elements from modular layers of the Internet by crossing the content of the specific electronic mail message with the language of the World Wide Web. Mixing modular layers stepped across a line that good programmers knew instinctively not to cross.

After many arguments the founders eventually relented, agreeing to include a link, but only if they dropped “P.S. I love you.” That left the phrase “Get your free Web-based email at HoTMaiL” as a hot link on every e-mail. It took another day to strike the final deal after some brinkmanship over the ownership stakes. DFJ wanted 30 percent ownership for a $300,000 investment, while Bhatia and Smith wanted them to have 15 percent, effectively doubling the valuation of the company. The entrepreneurs eventually got the terms they wanted.

There was not much of a marketing budget for HoTMaiL, and all would learn quickly that it would not matter. As the service opened, it initially received uptake at addresses within the edu domain, with thousands of students signing up at universities. It then began to spread to other universifies, and registration on the service mushroomed. Within a few weeks it had spread to India, reaching one hundred thousand Indian users three weeks after the first registration within that country. Less than six months into the launch, HoTMaiL had one million registered users, and by the following year, at the end of 1997, it had twelve million.

The rate of growth was astonishing, and even more astonishing was the cost. It had cost almost nothing to acquire new users. On December 29, 1997, a few days after reaching twelve million registrants, founders Sabeer Bhatia and Jack Smith sold the company to Microsoft for $400 million in Microsoft stock.⁹ Bhatia became the new general manager of strategic business development for MSN.com, a position he stayed at for one year.

Two inventions occurred at that meeting with DFJ. HoTMaiL invented a business, and it would become one of many businesses oriented toward delivering a browser-oriented e-mail service.¹⁰ A second invention emerged too, a method for acquiring new users for a new web service. Observers said HoTMaiL spread like a viral infection moving between people via a handshake,¹¹ so the process eventually went by several labels, among them “viral marketing” and “word of mouse.”¹² In the next half decade viral marketing showed up in thousands of entrepreneurial pitches.

A little myth making was invented that day as well. In popular imagination, HoTMaiL’s invention emerged miraculously, as if out of the air. Its arrival and spread seemed especially fast to those who had not followed the development of the privatized Internet.

Looking more deeply, this was inventive specialization at work, and as part of a commercial network. A new application had simply employed existing infrastructure by working through a browser. Needing only to add some basic programming, the costs of developing the application were therefore low. Due to the modular structure, an inventive specialist could confine his or her effort and attention to the factors shaping an application, such as its features, operation, and marketing.

Inventive specialization was so powerful because so much lay behind the surface of what a user saw. An inventor could give the user a simple set of functions. As Brian Kahin, founding director, Harvard Information Infrastructure Project, said,¹³

Digitization enables modularization with documented interfaces at one level while it also makes it possible to integrate functions, hide them from users, and make them look like a unified whole with no points of access other than a single user interface.

Demand conditions also had changed in a fundamental way. Privatization of the Internet had widened the set of potential users. No longer would inventive specialists depend on the assessment of administrators at universities, or the aesthetic preferences of the users from the Internet’s academic community. What had replaced these preferences? That was an open question to contemporaries at the time. Nobody knew for certain what services nonacademic users would want, what services they would pay for, or how much they would pay for these new services.

Even the most optimistic venture capitalist did not forecast how inventive specialization would play out within commercial markets over the next half decade. Inventive specialists of all stripes would tap into a scale of demand far greater than any forecaster had dared utter at the outset of privatization. That would motivate an enormous and impatient explosion of economic activity.

A Boom at Its Adolescence

In his 1996 book Only the Paranoid Survive, the CEO of Intel, Andy Grove, remarked about the difficulty in foreseeing a fundamental change in business until those changes are right on top of the management.

Most strategic inflexion points, instead of coming in with a bang, approach on little cat feet. They are often not clear until you can look at events in retrospect.¹⁴

The Internet had crept up on the computing market on little cat’s feet, and the perception noticeably changed in 1995. The Internet market became vested with a degree of excitement and energy that insiders could sense. Netscape’s founding and the release of its new browser began to generate excitement about the new era, but no doubts were left after Netscape’s IPO in August 1995 and Microsoft’s announcement in December 1995 that it would enter into making browsers as well. A considerable number of new businesses were launched in the ensuing year, further adding to the sense of anticipation and urgency.

Contemporaries called it a boom, which continued unabated for five years until the decline in traded prices for many Internet-related stocks began in the spring of 2000. Those prices rebounded for a short time, but subsequently reached a nadir in the fall of 2001 after the 9/11 World Trade Center terrorist attack sent much activity in the US economy into shock. After the fact, contemporaries called the decline a bust.

The boom coincided with something very real and substantive—namely, changes in Internet participation, both in its breadth and intensity. During this era the Internet became available for use in a variety of contexts, and many decision makers faced a choice about whether to make use of the Internet or not. Unlike prior experience in universities and libraries, where users typically participated without personal expense, in the new era, participation was a commercial choice and a cost was attached to that participation. Many potential participants—households and businesses—chose to pay the price and participate, and some chose to invest heavily. The vast majority of these decisions were made after 1995 and, as will be discussed subsequently, only a select few of them were reversed after 2001.

Contemporaries needed an explanation for the increasing participation on the web. Several labels emerged. One of those labels, killer app, or killer application, eventually stuck. This label focused on the deployment of the browser and was a term borrowed from commercial computing. In personal computing a killer app delivered valuable functions and motivated a user to buy the application and all the complementary components (for example, a printer and PC). In the early history of computing, the killer apps had been word processors and spreadsheets.

By analogy the browser appeared to be the killer application of the web, just as electronic mail had been the killer app of the Internet prior to privatization. Many users were attracted to both applications, and their functionality motivated PC users to purchase modems, Internet-access service, and related software. For some non-PC users the suite of applications motivated them to purchase the PC and its affiliated complements. Thus, as will be discussed here, the killer app analogy captures a grain of truth about increased participation at home and in business.

Applications for the browser came from an entrant generically known as the dot-com, which led to the next label for this era, the dot-com boom. The label came from a feature of the domain name system for the Internet, which initially designated five types of domain names: gov, net, org, edu, and com.¹⁵ Gov was for government entities and edu for educational institutions. Net, org, and com were designated for nonprofit and private entities, organizations, and networks. Com became the most popular by far among commercial firms, even for firms not based in the United States. The domain name, .com, had not received much attention prior to privatization because it had been reserved for commercial firms. In short, the label, dot-com boom, also contained a grain of truth, as it stressed the commercial orientation of the many start-ups offering such changes.

Several popular labels stressed the macroscale of events, as entrepreneurs and established businesses engaged in a wide range of new thinking across many business activities—in back-office computing, home computing, and information retrieval activities in information-intensive industries, such as finance, warehousing logistics, news, entertainment, and more. Applications emerged for shopping for goods, planning for vacations, updating accounts, finding romantic companions, conducting transactions, experiencing erotic pleasures, investigating leisurely diversions, sharing experience with others, ticketing for traveling, understanding financial investing, and updating on the latest news. Thus because of the unprecedented nature of the rate of investment of new applications designed to enhance business processes, as well as investment in the infrastructure to support that growth, there was yet another grain of truth in labeling the Internet market as a macroscale of events.

Among such labels, entrepreneurs stressed the use of aggressive tactics in support of fostering these goods, such as viral marketing and the low expense for acquiring new users (as in HoTMaiL, for example). They also stressed maximum leverage of the existing Internet infrastructure for fast entry and quick growth. Deferring attempts to increase revenue until later, their emphasis focused on the strategies that aimed to gain users quickly. With prominent venture money to put such actions into practice, these strategies were sometimes labeled as “get big fast.”¹⁶ In the hands of the most enthusiastic advocates, get big fast was one example among many. It became interpreted in light of a broad principle: Internet business operated according to an unprecedented set of rules and norms, distinct from the fundamental benchmarks found in normal businesses up until that time. Those looking for a broad label called the entire movement “the new economy.”

While founded on a substantive observation about investment behavior and strategies to support it, labels were nonetheless merely shorthand. Their meaning changed. One change was not profound. It resulted from competition between firms, and the broad pattern was familiar from other software markets. Something akin to an arms race between competitors emerged over features that appealed to users. This type of arms race arose in browser markets, in many applications of electronic commerce to retailing, and in many applications of electronic commerce to publishing. This and other chapters will describe some of these arms races in detail. Technical insiders were acutely aware of the progress, and debated the market value of each technical advance. Insiders debated the strategic value of clustering different technical advances together in expensive and complex services, and, for example, compared that with the merits of removing expensive features to make a low-cost service appeal to a specific set of users. This type of arms race does not evolve in a linear direction, but, instead by fits and starts, as new proposals are tried, improved, or withdrawn. Winners emerge, losers leave, and during a boom the attention of insiders flits from one event to another.

If the new economy had involved only arms races, it would have been unusual for the scale of the market and number of participants, but not otherwise distinctive. A second change emerged, however, and it gave the boom its unusual flavor. The meaning of the new economy mutated over the course of the boom. The mutation had its origins in blurred boundaries between the (actual) substance of investment and the (as yet unrealized) forecast of its consequence. While early advocates for business in the commercial Internet were careful not to take the outcome for granted, later advocates eschewed such caution. They began to presume a large payoff was inevitable without questioning deeply the evidence for or against such a presumption.

This presumption was most evident in the strategies of many firms that aspired to get big fast. Often their business plan took a cavalier approach to generating revenue. Their business plan would aspire to grow an “installed base” of loyal users and generate revenue later, typically with unspecified plans for selling ads or charging loyal users a fee to retain service. It was as if generating the revenue was regarded as easy, while attracting loyal users required the greatest explanation.

Early on, a number of analysts for the new economy sprang up to help others navigate investing in and partnering with entrepreneurs. For example, one of the first systematic studies of the value chain for Internet services took place at Morgan Stanley, a Wall-Street-based investment firm with a wide range of clients. Morgan Stanley had managed the Netscape IPO and looked to step from that activity into leadership in a range of Internet investing. A forward-looking and ambitious analyst of software firms, Mary Meeker, led the analysis. Along with her colleague, Chris DePuy, an engineer turned analyst of networking companies, they published an early and widely read study for investors called The Internet Report.¹⁷ The report analyzed many firms with the ability to take advantage of the growing adoption of the web-based Internet. The report followed conventional norms for financial analysis, clarified the value chain and its leaders, and showed care about the links between investment and returns.

A new group of analysts and pundits also showed up and grabbed attention from the established line of analysts, such as Meeker and those similarly positioned, who were accustomed to speaking to a small group of professional investors. This new group did not have experience and considerable in-house investigative resources to build and enhance a reputation. Instead of aiming for the small audience of professional investors, the new group aspired to reach a new and growing audience—individual investors, hobbyists, entrepreneurs, and executives at existing firms who sought to understand the new economy. Analysts such as Henry Blodget appealed to this wider audience by speaking the language of the entrepreneurs and gaining notoriety for making predictions out of sync with the established analysts.¹⁸ Many futurists, such as George Gilder,¹⁹ also grabbed prominence, in this case, for offering a vision consistent with the enthusiasm of many entrepreneurs. As the Internet became a magnet for all manner of futurist, pundit, consultant, and analyst, the most enthusiastic advocates for the new economy found their spokesmen among a group that contemporaries called the digerati.²⁰ Later chapters will discuss their role at greater length.

Internet exceptionalism grew from these roots because it served the interest of entrepreneurs and many digerati, and it played to the hopes of many of them that they could succeed through emersion in the present setting. Internet exceptionalism argued that businesses based on web services could develop quickly, using existing Internet infrastructure and new web tools. It also argued that Internet businesses would grow online businesses superior to those followed by established firms, once again, playing to the hopes of many entrepreneurs that established firm were encumbered by old ideas and existing ways of doing business. This view presumed that the business rules of established businesses were outmoded, obsolete, and destined to fail in confrontation with the best start-ups from the new economy. It also argued that executives in established firms had either outmoded views and inappropriate technology or organizational structures that prevented extending their business activities usefully into Internet-oriented activities.

Internet exceptionalism also began rejecting the conventional benchmarks of financial capitalism, such as referencing price/earnings ratios within a specific range. This presumption was, at best, grounded in the view that Internet start-ups would grow to extraordinary levels of revenue, replacing established firms. Accordingly, the financial value of leading publicly traded firms in the new economy began to rise far above levels that conventional analysis would support. During the boom the stocks for a number of firms benefited from this rise, including Cisco, Amazon, Netscape for a short while, Yahoo!, AOL, and many others.

The digerati began to use their own terms and vocabulary in sync with entrepreneurial ambitions. Presentations and business plans would stress their aspiration to “get big fast” through the use of processes affiliated with the new economy, such as viral marketing in the pursuit of “disruptive technology”²¹ and as part of a “pure play”²² in an act of “disintermediation.”²³ A common public occurrence in this era had the digerati diagramming flows of economic activity that the Internet and web could support, usually arguing that the Internet would replace a flow of activities that previously had been accomplished entirely with tangible activities. Enormous numbers of entrepreneurs with these ambitions raised funds and aspired to implement their ideas in tangible processes. Entire conferences became devoted to these aspirations, and the trend acquired many of the trappings of a large social movement in which the members derided critics as outsiders who “did not get it.”

Geography deserves a share of responsibility in fostering this movement. Participants in the commercial Internet shared their interests with one another, interacted with one another in a self-referential way, and, like normal humans, tended not to converse with skeptics or others that held contrary views. Enthusiasm for new Internet businesses was no different than the environmental movement or PC enthusiasts of an earlier era. The concentration of so many of these participants in a small number of places—Silicon Valley; Seattle; New York; Boston; Washington, DC; and the Research Triangle, North Carolina—positively reinforced one another’s sense of belonging to something greater than themselves. To its participants the movement felt large enough to be self-sustaining, and in the absence of any large and visible shock, participants had every reason to continue their conversation.

Events eventually contradicted the premises underlying Internet exceptionalism, but the understanding that Internet exceptionalism was based on faulty logic and wildly optimistic expectations only emerged in retrospect. This revisionism eventually heaped ridicule on those who wasted resources under the guise of pursuing the new economy, but that was far down the road.²⁴ That retrospective bias should not shape our understanding of the phenomenon as it emerged.

Internet exceptionalism survived in the face of skeptical remarks from leaders with considerable authority and public stature. Here are two examples of skeptics who could not dent the tide. First, Andy Grove, president of Intel, continually (and incorrectly) publicly forecast that the end of the boom was near. He became known as the boy who cried wolf.²⁵ Before the bust his warnings had little public resonance. As another example, consider Carl Shapiro and Hal Varian—respectively, professor from the Haas School of Business, and dean of the Information School, both at University of California, Berkeley. They published a book in 1999 with Harvard Business School Press titled Information Rules: A Strategic Guide to the Network Economy. They openly criticized many new-economy philosophical tenets by arguing that mainstream economic models provided the basis for analytical insight. Addressing managers, they state in the introduction:

The thesis of this book is that durable economics principles can guide you in today’s frenetic business environment. Technology changes. Economic laws do not. If you are struggling to comprehend what the Internet means for you and your business, you can learn a great deal from the advent of the telephone system a hundred years ago.

The book became a best seller, and it is often credited with motivating many conservative business executives to take action. Yet in spite of wide circulation, as well as the authority of the opinions, it did not dent the overarching tide of enthusiasm for Internet exceptionalism among its supporters.

It was as if a fever had gripped many decision makers and entrepreneurs. To those not under its power this fever appeared like a lovesickness that clouded judgment. The warnings of authorities were treated selectively, like the wise musings of an older uncle whose insights about his generation seemed distant from the present. To those under its spell, the combination of resources and perception (as well as, perhaps, misperception) provided business owners, investors, and entrepreneurs with the courage and justification to take risky actions, to take the leap of faith needed at the start of a new entrepreneurial venture.

Seen on its own terms, the new economy philosophy offered a self-reinforcing system of beliefs. This system argued that investors traded effort in the short run for riches eventually, and the confident rhetoric reduced the perceived risk to entrepreneurial start-ups. Prior generations of entrepreneurs had faced stiff competition with existing firms that had many comparative advantages, such as established brands, operations, and investments, as well as loyal customers. Prior generations of start-ups had faced a steep challenge, trying to attract customers with their new service and operations. The new economy arguments argued that these challenges were less salient to the experience in the new economy. They convinced many investors to give money to start-ups—many of whom went public or sold out to existing firms. It also convinced many managers, engineers, and executives to change their careers in the hope of starting such firms, or joining one.

Later chapters will return to Internet exceptionalism in the context of describing particular application markets, such as the slow evolution of events leading to the dot-com bust, or the rise of advertising to support portals and search engines, and these examples will provide details about the decline of this system of beliefs. The remainder of the chapter turns to the real side of the economy and provides illustrations and outlines of three types of investment activities that were composed in the new economy:

1.  Entrepreneurial entry;

2.  Adoption of browser-based Internet services by homes;

3.  Adoption of browser-based Internet services by businesses.

These three groups were big investors, but not the only investors. As that adoption accelerated it motivated two types of investment that worked with the increasing number of Internet users, translating it into routine and pervasive processes at a large scale. Those investments are the subject of the next chapter. In other words, two additional investors join the above three, and these two focused on:

4.  Enhancement of business processes by existing businesses; and

5.  Capital deepening in the infrastructure to support all these activities.

Although uncoordinated by any single entity, it was as if all five of these disparate economic actors simultaneously agreed to build the commercial Internet at a scale that would make it ubiquitous for mass markets, and would do so with a pervasive sense of urgency.

Entrepreneurial Entry and the Start-Up Boom

As the Internet built up, it began to take on a great size, one that quickly began to remind venture capitalists of the PC boom, which had been the largest previous entrepreneurial boom. Eventually the scope of aspirations exceeded any prior boom. The breadth of opportunity was too big for a brief race among several dozen firms to develop new components and related systems. There was open discussion about changing the entire chain of actions supporting the delivery of any valuable information-intensive activity, such as music, publishing, news, financial activities, and entertainment.

Who would meet the demand for these new services? Perceiving gaps between what was possible and what they and established firms could do, droves of entrepreneurial firms jumped into the fray. Particularly motivated by notions of Internet exceptionalism, many entrepreneurs sought to establish “pure-play” businesses, firms that operated solely over the Internet, and, in many cases, without physical presence.

Much of the entry was geographically concentrated. As ISPs sprouted in many locations, and as browsers began to diffuse, software entrants began exploring new businesses. Many of them were founded in the San Francisco Bay Area, and in the area along the peninsula known as Silicon Valley, named for the integrated circuit firms founded there generations earlier. This location became a focal point as a direct outgrowth of the entrepreneurial orientation of the venture-funded community, and where the preexisting labor market for technically skilled employees was becoming privy to the latest developments for the commercial web. These entrants included Yahoo!, HoTMaiL, Excite, eBay, Vermeer, and many others that anticipated building businesses on browser-based computing.²⁶

The entrepreneurial movement at the outset extended beyond the San Francisco boundaries. For example, Bill Gates’s memo of May 1995 titled “The Internet Tidal Wave” advises other executives to examine web pages by Lycos, Yahoo!, Oracle, Symantec, Borland, Adobe, Lotus, the San Jose Mercury News, Novell, Real Audio, Disney, Paramount, MCI, Sony, ESPN, and many other sites.²⁷ Although the first eight of these companies were located in the Bay Area, the rest were not.²⁸ Nevertheless, as the growth continued, the venture-based entrepreneurial community based near Silicon Valley became the major source of funding, and a majority of start-ups located there. Other prominent locations were Seattle, Boston, New York, and Washington, DC, but all were much smaller in scale.

The scale of entry was enormous, whether measured by the number of new businesses proposed and founded or by the amount of investment dedicated to this activity. There were, in fact, two phases to the growth. The first phase could be labeled the Internet gold rush, and it occurred between 1995 and 1997, when the first wave of enthusiastic entry into entrepreneurship began. Because the first wave of businesses had success in the IPO market and in selling out to larger firms, another wave began, which provided additional momentum, lasting from 1998 to the middle of 2000. This second wave included many of the examples that later became infamous for the descriptive phrase, “Too much money chasing too few good ideas.”²⁹

An examination of funded entrants provides a sense of timing of the boom, as well as some sense for the difference in scale. Figure 9.3 shows the amount of money that was funded to start-ups in ten areas of computing, communications, and software. The first eight had their fortunes tied directly to prospects in the Internet, while the latter two—wireless telecommunications and computer peripherals—provide comparisons with the wider market.³⁰

All the graphs show the same pattern—growth in and around 1995, followed by decline in and around 2001. Similarly, the scale of the peak years exceeds anything that had come before it, and generally the peak comes in 1999 or 2000.³¹ This is so of any area, even those without direct relationship to the Internet. The non-Internet areas differ from those directly related to the Internet in only one respect, though they reach a peak in 1999 or 2000, many new start-ups continued to receive funding in periods past the peak. Examination of the number of funded entrants (instead of the amount of funding) provides a similar insight.

The magnitude of investments also obtained extraordinary levels during the peak. Investment in data communications exceeded $50 billion at its peak in 2000, and in Internet communications exceeded $45 billion. Internet software and Internet content each exceeded $20 billion at their peak. Wireless telecommunications exceeded $15 billion, fiber optics $13 billion, electronic-commercial technology $11 billion, and commercial communications $8 billion.

This only represents the funded entrants. By some estimates, these defined the tip of the iceberg. When less prominent businesses are included, one study estimates that less than 15 percent of the start-ups that sought funding during the boom followed get big fast approaches requiring large financial support, with only a tiny fraction subsequently having an IPO. Most had business ideas that did not require much funding.³²

Although dot-com entrants covered a wide range of new businesses and used a wide set of models, three categories received special attention: (1) electronic commerce, (2) advertising-supported media, and (3) portals. Consider first electronic commerce. An entirely distinct group went into electronic commerce in various forms. Some sought to sell and distribute goods and services, such as books, travel services, or apparel. Others sought to sell subscriptions to services, such as the New York Times crossword puzzle or regular updates of industry news. Still others tried to assemble groups of buyers and sellers, either in open auctions or for more tailored matching purposes.

Ticket purchases illustrate the last trend. It did not take long to remove travel agents from the vast majority of trains, planes, autos, and hotel purchases. Web technology allowed users and vendors to directly deal with each other. A related innovation arrived quickly, and it might be called the inexpensive automated intermediary. After a firm set it up, a user (or purchaser) could easily query a database for seats or rooms (or just about any object). These services extended into many other areas, such as vacations, concerts, theaters, and restaurant reservations.

Another area of electronic commerce that assembled groups of buyers and sellers had a web presence but needed a substantial operational component to support it. Amazon.com was a leader in that category. Amazon began selling books online early in the history of household web use, and its experience is well known. Online offerings of CDs, toys, plane tickets, greeting cards, cheese, and tons of other merchandise followed quickly, often by other firms. These entrepreneurs deliberately mimicked the practices of catalogs, and minimized building physical facilities in most states. Thus they avoided legal obligations to pay sales taxes, which made their products cheaper in many states and municipalities than those bought in physical stores.³³

These entrepreneurs generally attempted seven different approaches to starting a pure-play Internet firm.³⁴ Surveys of business plans³⁵ showed that many dot-com entrepreneurs listed more than one potential source of revenue. The most common revenue source listed was fee for service. Over three out of five (62 percent) had such a plan. Another popular source of revenue was advertising, with over a third relying on it (36 percent). Similar rates were found in the use of production (26 percent), subscription (25 percent), and commission (23 percent). Less common were businesses relying on markup (18 percent), and referrals (13 percent).³⁶

Now consider advertising-supported media. Initially a large number of entrants went into advertising-supported websites or sites with no usage fees.³⁷ A number of sites covered directory and search activities, while others specialized in supporting conversations and information sharing for particular topics or groups.³⁸

Much of the advertising-supported Internet initially resembled the related offline magazine industries, with users spreading themselves among a range of topics. Early surveys³⁹ showed that the most popular (after general portals and electronic mail) were community gatherings (for example, Geocities), financial information (Motely Fool), news (CNN.com), games (Boxerjam), streaming media (Real.com), sports (ESPN.com), technology (ZDNet), specialized portals (Women.com), music (MP3.com), weather (Weather.com), and so on.⁴⁰ Throughout the decade the list of categories continued to expand.

A tension emerged between start-ups that contracted for all their services using copyright licenses, and those who adopted pieces of the hacker ethic found among ISPs. Like the ISPs, even if they did not hold this attitude for ideological reasons, many small start-ups turned a blind eye to piracy of content among their users because it helped business. They had no incentive to police it, and many were too impatient to do so. This tension was largely held in check as long as larger firms who licensed their content, such as AOL, dominated commerce.

It was further held in check by a related important innovative offering called portals. Portals became some of the most popular websites among Internet users, accounting for the opening pages for most users, acting as a tool for organizing web surfing, and accounting for the highest share of an individual’s time online. There were two approaches to providing portal services at first. One approach made it an extension of another service, and virtually all of these firms aimed their product at mainstream users. For example, Netscape and Microsoft made the opening web page a default setting on their browser, directing considerable traffic to netscape.com and msn.com, respectively. Both were far down in use to AOL, which made its opening page an extension of its dial-up Internet service and organized its contents in an easy-to-use proprietary format—a strategy that became known as a walled garden.

Another approach was of stand-alone portals, and these differed in their appeal to technical and nontechnical users. Most early (and successful) portals in the mid- to late 1990s provided directory services for the vast amount of content on the web. Yahoo, Lycos, Excite, and others took this approach. Yahoo grew into a vast organizer of content and a popular destination. Excite sold itself to @home as part of a strategy to help cable Internet users.⁴¹

During this time many firms offered alternatives to portals, focusing on searching the web, which chapter 13 will discuss. In the late 1990s several different firms offered popular search engines, including Inktomi and Alta Vista. In 1998 a newly founded firm, Google, entered with a new approach to searching—namely, by ranking web pages on the basis of the number of links to those pages and, eventually, by supporting itself with advertising. This turned out to be popular with users; after the turn of the millennium, Google continued to grow this business and eventually surpassed the early leading portal, Yahoo, in usage, revenue, and capitalized value. At this stage, however, many of these search engines were pieces of functionality for the portals, acting in the background.

Internet Adoption in Homes

Many households viewed the increase in the web’s functionality as a good reason to acquire Internet access.⁴² Not all households adopted right away. The different timing of adoption arose from a mix of factors. Prices for Internet access fell perceptibly in 1995 and 1996, but steadied thereafter.⁴³ Meanwhile, its availability from a local phone call spread into virtually every small urban area by 1997. Households also differed in their willingness to adopt due to their perceptions of how affordable and valuable they found it. As the Internet improved after 1995, with the entry of many new services, many households found the benefit worth the expense.

The Internet’s adoption was closely linked to the PC’s. The Internet was much easier to adopt if the PC already existed at a household. In that sense, the diffusion of the Internet depended heavily on the diffusion process of PCs prior to 1995, and, for similar reasons, the Internet’s diffusion thereafter depended on the PC’s subsequent diffusion. On top of this interdependence, within the class of PC users there were also differences in their willingness to experiment and the intensity of their use.⁴⁴

The importance of the PC and browser can be seen in US government surveys of Internet adoption. The government had completed a large-scale sampling of PC use across the country in 1995, the fourth such survey. At the beginning of 1995, just under 25 percent of US households had a computer, up a couple percentage points from the previous survey done a year earlier, and up from 15 percent of households in the fall of 1989.⁴⁵ The Internet was so new in 1995 that the government survey did not directly ask about its use until the next survey in the fall of 1997, when 36.6 percent of US households had a PC. In that survey just over 50 percent of the PC users, or 18.6 percent of households, had an Internet connection. Growth was rapid from then on, with 26.2 percent of households using the Internet in the fall of 1998, and 41.5 percent in the winter of 2000.⁴⁶ As of 2001, approximately 56.2 percent of American homes owned a PC, with Internet participation rates at 50.3 percent in 2001.⁴⁷ Over 80 percent of US households with an Internet connection used dial-up connections in 2001. This is illustrated in figure 9.4.

After triangulating from multiple sources to fill in many missing observations and estimate adoption in those years, Greenstein and McDevitt (2012) put household adoption at slightly higher levels, as shown in table 9.1.

By 2001, Internet usage correlated with higher household income, employment status, and educational attainment. Much of this disparity in Internet usage can be attributed to observable differences in education and income. For example, at the highest levels of income and education there were no significant differences in adoption and use across ethnicities.⁴⁸

Since the vast majority (87.6 percent) of PC owners had home Internet access, the marginal Internet adopter looked similar to the marginal PC adopter. For households, therefore, PC demand had two distinct populations: (1) Those already owning a PC, and (2) those that were first-time purchasers of a PC. In the middle of the 1990s, two distinct Internet adoption patterns mirrored the two populations of PC demand. That pattern ceased to exist by the end of the period, when either existing PC adopters converted to the Internet, or households bought PCs and converted to the Internet. By 2001 to 2002, virtually all existing PC adopters had experience with the Internet at home.⁴⁹

Quite a different set of factors, therefore, shaped later adopters than did those for earlier adopters. As the diffusion of the PC moved deeper into mainstream use, the marginal PC and Internet adopter became a household with low marginal value for PC quality, high start-up costs, significant price sensitivity, and potential difficulty in determining when (not necessarily if) to buy.⁵⁰

The form of basic access at households also changed over time. Netscape tried building electronic mail functionality into its browser during 1995 and 1996. Many independent and small ISPs liked this function because it allowed them to offer electronic mail to their customers simply by supporting the browser. Many sophisticated users of basic services were satisfied with this option, or used one like that offered by HoTMaiL or Yahoo Mail. In 1998, AOL bought CompuServe,⁵¹ and, with that merger and organic growth, by the end of 2000 AOL was the most widely adopted provider of Internet access in the United States.

Internet Adoption in Businesses

The diffusion of basic Internet connections to businesses had many of the same determinants as in households. Implementation for minimal applications, such as e-mail and browsing, was rather straightforward in the latter half of the 1990s. It involved a PC, a LAN or modem, a contract with an ISP, and some appropriate software. The presence of internal IT expertise at a larger business establishment permitted a wider range of implementations, including Lotus Notes and cc:mail, and eventually Microsoft Outlook and Exchange.

In 1995 it would not have been a spurious conjecture to assume that some industries would adopt basic access sooner than others. Firms that moved earliest were those in industries with a history of being more information intensive or making more intensive use of new IT developments in operations. The latter group would include research industries, where the Internet was already familiar. Heavy computer-technology-user industries included finance, utilities, electronic equipment, insurance, motor vehicles, petroleum refining, petroleum pipeline transport, printing and publishing, pulp and paper, railroads, steel, telephone communications, and tires.⁵²

Such a forecast never faced a direct test because at that time no comprehensive survey equivalent to the surveys presented to households examined business use of the Internet. One of the earliest attempts to comprehensively measure national Internet adoption rates took place at the end of 2000 and confined attention to medium and large establishments.⁵³ These studies showed why it was essential to distinguish between two purposes for adopting, one simple and the other complex. The first purpose, here labeled basic participation, relates to activities such as e-mail and web browsing. This represents minimal use of the Internet for basic communications. The second purpose, here labeled enhancement, relates to investment in frontier Internet technologies linked to computing facilities. These latter applications were called e-commerce and involved complementary changes to (1) internal business computing processes, such as part procurement, or (2) customer-facing distribution channels, such as electronic retailing.

The economic costs and benefits of basic participation and enhancement were distinct. The determinants of the diffusion of each type of use differed, so adoption proceeded at separate and distinct paces.

Let’s begin with basic participation. Adoption costs for basic participation were low, and at its smallest scale, were similar to households—involving simple $20 per month contracts. Basic participation was economically feasible in virtually any business in almost any location. Because nearly every business experienced some benefit from adoption, it was widespread and was almost a necessity for US business by the end of the millennium. Thus by the year 2000, adoption of the Internet for purposes of participation approached near saturation in most medium- and large-scale establishments in most industries. Hence, the forecast for information-intensive industries was wrong, by being too pessimistic about those firms that had not been information-intensive in the past. Virtually all firms in all industries adopted.⁵⁴

The rate of adoption was remarkably fast in historical perspective. The closest historical analogue came from several software applications, which industry wisdom suggests were killer aps for PCs. In this sense, there was a large grain of truth to the claim of advocates of the new economy about the unprecedented nature of the rate of improvement. The diffusion of e-mail and the browser had few precedents except the diffusion of the spreadsheet (that is, Visicalc and Lotus 1-2-3), second-generation word processing (that is, WordPerfect), or electronic financial planning (that is, Quicken), which were some of the fastest-diffusing software applications.

E-mail played an important role in this experience, because basic service supported e-mail and browsing. As it turned out, the widespread diffusion of the Internet did render moot the advantage held by established firms, such as MCImail, or Lotus Notes.⁵⁵ Instead, it became an opportunity for the growth of web-based e-mail services, and at this time, Microsoft and Yahoo bought the two leaders of e-mail providers, HoTMaiL and Rocketmail, respectively. The widespread diffusion also made it particularly easy to operate e-mail software at the level of an establishment and connect servers between establishments using the Internet’s gateway.⁵⁶ Coincident with this opportunity, Microsoft began selling Exchange server, bundling the client software, Outlook, for free. Outlook also included its calendar and contact manager and could be purchased as a part of the Office suite bundle. The firm also began pricing at discounts to convert an entire enterprise. It eventually became the dominant electronic mail provider in businesses.

There were good reasons the investments in basic participation occurred at the same time as investments in enhancement. Increasing participation increased the gains to enhancing business processes. This is a big topic, and the next chapter will discuss its characteristics.

The New Normal

The commercial Internet could not be built without simultaneous investments from many distinct participants—households, business, carriers, software developers, ISPs, retailers of all stripes, and more. Consistent with the rhetoric to support it, many different participants in the economy, in fact, simultaneously adopted the Internet and made investments to make it useful. Virtually all of it leveraged off the existing infrastructure in the network, and then extended its scale. An innovation that had lived outside of the perception of many was quickly becoming a mainstream activity.

That accumulation of investment began to accelerate in the mid- to late 1990s, as all participants saw how pervasive the commercial Internet had become, and all forecast it would become more pervasive. Investment by one type of participant raised the gains to investing by another, which accelerated the real economic incentive to continue to invest after adoption, which generated incentives for the laggards to become involved.

For those from the edges this situation represented the realization of a long-desired dream. For entrepreneurs and investors in their venture-funded businesses this represented an enormous potential market. Many perceived it as a once-in-a-lifetime opportunity to change the world. Students dropped out of college to found businesses. Engineers gave up stable work to take a chance on a fast-growing start-up. News outlets reported extensively and widely on such episodes, reinforcing the public perceptions that the zeitgeist had reached an unusual and unique point in history.

¹ http://www.linuxjournal.com/article/2608, accessed October 2012.

² The name HoTMaiL emerged from the business plan. The name signified that the service was built on HTML, so it could be used anywhere on the web.

³ In time AOL would offer e-mail accounts that did not require downloading their software to the desktop, but at this point it was not available.

⁴ For the long account, see Bronson (1998).

⁵ The most widely publicized effort was undertaken by Lotus for its cc:mail service offering. It too involved using a web browser to access e-mail from any location. See InfoWorld (1995a, 1995b).

⁶ Grigory Potemkin led the military conquest of Crimea, and allegedly created false villages to impress Catherine II and her entourage during her visit to Crimea in 1787. Accordingly, a Potemkin village refers to a village of false facades.

⁷ The business concerned an Internet-based personal database they called JavaSoft.

⁸ Draper also thought of Tupperware in the context of chess strategy. One approach to chess strategy involves thinking multiple steps ahead. As Draper considered how one person would contact another in HoTMaiL, he began to think multiple steps ahead and envisioned that one user would contact another, who would then contact another, and so on. Tim Draper, private communication, August 2008.

⁹ Ransdell (1999).

¹⁰ For example, DFJ was one investor in another successful web-based e-mail service, RocketMail, founded by Four11 Corporation, which competed directly with HoTMaiL. Managed with an internal firewall, RocketMail was acquired by Yahoo! in October 1997, after they passed on buying HoTMaiL. It eventually became the basis for Yahoo Mail. See Pelline (1997b).

¹¹ Greenstein (1999).

¹² Greenstein (1999).

¹³ Brian Kahin, private communication, October 4, 2013.

¹⁴ Grove (1996). For a time Grove’s phrase about little cat’s feet was widely quoted. Where did it come from? It is borrowed from a Carl Sandburg poem about fog. “The fog comes / on little cat feet. / It sits looking / over harbor and city / on silent haunches / and then moves on.” Grove would have been familiar with this poem from living on the peninsula of the San Francisco Bay Area. Almost every summer evening Pacific fog crawls over the hills west of Intel’s headquarters and moves east on its way to the California Central Valley.

¹⁵ Every country was assigned control over the allocation of domain names underneath its two-letter country code. The country code for the United States is US, but com, net, edu, gov, and org presume US sovereignty.

¹⁶ Kirsch and Goldfarb (2008).

¹⁷ Meeker and DePuy (1996).

¹⁸ Blodget first received widespread attention in the fall of 1998 for forecasting the rise of Amazon’s stock to (then an unheard of price of) $400. This is described in chapter 12.

¹⁹ George Gilder writes on a wide set of topics, including information technology, for many prominent publications. This is described in chapter 12.

²⁰ The term combined digital and literati. The origin of the term predates the boom, arising first in 1992, and originally stressed writers that promoted their vision of digital technology. During the boom digirati took on additional meanings, denoting celebrity status among many involved in the dot-com boom. See, e.g., Markoff (1992), and Brockman (1996).

²¹ This term was first popularized in 1995 in Bower and Christensen (1995), then in an expanded description in Christensen (1997). The original publications offered a very specific mechanism—when a new technology with lower capabilities grows over time in functionality, eventually competing with firms who initially underestimated the potential for change. The term then mutated to encompass a much wider set of phenomena—almost any change with technical underpinnings that threatened to lower the value of services offered by an established firm, much more than analyzed within the original books.

²² By this is meant the business provides only an Internet-oriented face to potential customers. These businesses leveraged other aspects of the Internet to achieve lower overhead expenses and then they competed with large brands.

²³ This term generally refers to removal of some firm or other commercial participant in a value chain in order to save cost. In the context of the times, this term often referred to, for example, Internet-enabled processes substituting travel agents with the use of online ticketing.

²⁴ A particularly sardonic account of the wasted resource includes Kaplan (2002).

²⁵ See Heilemann (2001).

²⁶ With twenty-twenty hindsight it is too easy to treat this outcome as obvious. For insights into the uncertain experience of an entrepreneur in this early era, see Ferguson (1999).

²⁷ See Gates (1995). At the end of the memo Gates provided a list of websites to give other executives examples of sites Gates found exemplar in various respects.

²⁸ As another illustration of the extent of geographic dispersion consider this example from a slightly later time, Meeker and DePuy (1996) highlights young Internet companies from California, such as Silicon Graphics, SUN, Cisco, Excite, Netcom, and Netscape, but also features plenty from all over the country, such as AOL, CompuServe, UUNET, Dell, Compaq, US Robotics, IBM, and others.

²⁹ Lerner and Gompers (2003).

³⁰ All data come from Venture Expert. Commercial communications (peak number: twenty-four funded start-ups in 1999) includes several categories of entrants from media, broadcasting, CATV, cable services, and pay TV. Data communications (peak of ninety-five funded start-ups in 2000) include communications processors, local area networks, wide area networks, protocol converters, modems, hubs, switches, and routers. Internet communications (peak of 114 in 1999) includes Internet access, Internet infrastructure, and Internet multimedia services. Electronic commerce technology (peak of seventy in 1999) includes a variety of Internet security software, e-commerce software, and e-commerce services. Internet software (peak of 107 in 1999) includes Internet system software, site development software, search software, web-service software, web language and web authoring tools. Fiber optics (peak of forty-nine in 2000) includes cables, couplers, and other systems related to fiber optics. Internet content (peak of 178 in 1999) includes products for businesses, consumers, retailing, publishing, agriculture, transportation, manufacturing, and recreation. Internet services (peak of twenty-four in 1999) include services for medical health, and services aimed at communications, education, reference, scientific, and legal services. Wireless services (peak of eighty-seven in 2000) include wireless services, paging services, and components. Computer peripherals (peak at eleven in 2000) includes a myriad of products, including terminals, printers, I/O devices, hard-drive storage, and cards for PCs.

³¹ Six of the eight Internet areas reach their peak number of entrants in 1999, while two reach their peak in 2000. However, all reach their peak in funding in 2000, which includes funding for firms founded prior to 2000. These data are in nominal terms, but adjusting for inflation would not change the insight.

³² See, for example, the analysis of 50K business plans from this era, as analyzed in Kirsch and Goldfarb (2008). They argue that a large number of businesses were founded outside a fast-growth strategy backed by a venture capitalist, and that a large percentage of these firms (close to half) survived just fine.

³³ Why did political actors leave these local sales taxes uncollected, and not set up a system to do so? This is an interesting question about the political economy of the time. The nonaction implicitly subsidized online retailing as well as avoided local/federal jurisdictional disputes over collecting taxes.

³⁴ See Afuah and Tucci (2003).

³⁵ See as well, Goldfarb, Kirsch, and Miller (2007).

³⁶ Fee for service. This type of business charges for professional service in relation to use of it. Advertising. Especially used in media business, this online business attracts readers, and shows them banner ads, pop ups, etc., and receives fees for the ads. Production. A manufacturer sells directly over the Internet, which cuts out distribution expenses. Subscription. A company charges for use of a service for a certain period of time, such as month or year. Commission-based. A fee was imposed on a transaction by a third party, usually an intermediary. Mark-up based. Employed by resale sites acting as middlemen, a business charges a markup over wholesale cost. Referral fees. A company steers customers to another company, often on fee per click-through.

³⁷ See Goldfarb (2004).

³⁸ See Haigh (2008) for a discussion of this segment, and see Goldfarb, Kirsch, and Pfarrer (2005) for discussion of the variety of entrants.

³⁹ See Goldfarb (2004).

⁴⁰ The remainder were chat, television, electronic cards, entertainment, directories, health information, maps, jobs, city guides, movies, genealogy, classifieds, and real estate.

⁴¹ See the account of Rosston (2009).

⁴² Any household was considered connected to the Internet if it had the capability of communicating with other entities via the physical structure of the Internet.

⁴³ The price drop coincided with the rush of new entrants into the market for ISPs. See Stranger and Greenstein (2007).

⁴⁴ For more on the diffusion of PCs, see Goolsbee and Klenow (1999), US Department of Agriculture (2000), or NTIA (1995, 1997, 1998, 2002).

⁴⁵ See NTIA (2000), figure 1, page 1.

⁴⁶ NTIA (2004).

⁴⁷ See NTIA (2002).

⁴⁸ With regard to age, the highest participation rates were among teenagers, while Americans in their prime working ages (twenty to fifty years of age) also were well connected (about 70 percent). Although there did not appear to be a gender gap in Internet usage, there did appear to be a significant gap in usage between two widely defined racial groups: (1) whites, Asian Americans, Pacific Islanders (approximately 70 percent); and (2) blacks and Hispanics (less than 40 percent).

⁴⁹ In his paper, Prince (2008) describes three main determinants of the “divide” in PC ownership. See also Goolsbee and Klenow (2002), where the role of local network effects in motivating early adoption is emphasized.

⁵⁰ For a general explanation of browser adoption strategies and their interplay with the diffusion to early and late adopters, see Bresnahan and Yin (2007).

⁵¹ The deal was reached in the fall of 1997 and completed in February of 1998. The network services and infrastructure part of CompuServe was sold to Worldcom, while the Information Services part was sold to AOL. See Pelline (1997a).

⁵² See Cortada (1996) and Forman, Goldfarb, and Greenstein (2003 a, b).

⁵³ See Forman, Goldfarb, and Greenstein (2003 a, b). The US census also attempted to understand adoption, and took a different approach. It examined established businesses of all sizes—small, medium, and large—but only in manufacturing. See McElheran (2011).

⁵⁴ See Forman, Goldfarb, and Greenstein (2003 a, b). Medium and large establishments employ approximately two-thirds of the employment for the labor force, so this was substantial. However, this record was not replicated in many small establishments, which had a much more mixed set of outcomes, especially in rural areas.

⁵⁵ Its attempt to build an e-mail client software, cc:mail, did not make up for this problem.

⁵⁶ Todd Warren, private communication, June 2009.