6

How Not to Start a Gold Rush

I’d say there was a fair amount of skepticism at the time about whether the Internet held any promise. And of course I felt that it did.

—Jim Clark¹

Just as the California gold rush of 1849 involved rigorous and concerted exploration of natural resources for commercial gain, a gold rush in a technology market can be an intensive exploration of a specific product’s commercial value. In mid-1995, especially after Netscape’s first browser and Netscape’s initial public offering (IPO), the commercial Internet turned into something akin to a gold rush, experiencing intense exploration by a large number and wide range of firms. Many established firms altered their investment plans to take advantage of the new business opportunities the commercial Internet offered. Many newly founded start-ups, attempting either to displace incumbent firms or establish service in areas incumbents had not addressed, also tried to take advantage of these opportunities by becoming dominant in specific activities.

Examining these events leads to two economic questions regarding the ways that firms invested their resources in the commercial Internet: Why did they explore and invest simultaneously? Why did the Internet gold rush not occur sooner?

The beginning of the explanation lies in Coloma, California, and in lessons learned from the history of the California gold rush. Coloma gets its name from a Native American word for beautiful. It deserves the label. Located on the southern fork of the American River, just over fifty miles east of Sacramento, it retains an abundance of oak trees and other flora. These dot the rolling foothills, which lie at the base of the Sierra Nevada mountain range. It is not an easily accessible place today and can be reached only by way of hilly roads off major highways. It was regarded as an even more remote location when it first gained notoriety, as the location for the discovery of gold in California in 1848.²

Contrary to popular romantic conceptions, gradual exploration is the norm for metal mining, and that is how most of gold in the Sierras eventually was mined. It is not what happened in Coloma. There are several seams of gold in the Sierra Nevada, put there hundreds of millions of years ago by natural forces. Millions of years ago plate tectonics formed the mountains and pushed several of the seams close to the surface. Miners eventually explored many of those seams using underground tunnels.³ However, glacial erosion did much work to expose one of those seams, combined with wind and rain. The southern fork of the American River happens to travel over one exposed seam, taking flakes from the rock, depositing it a little downriver.

Spanish settlers had been in California for more than seven decades. Evidence of gold occasionally surfaced in rock in the Central Valley—over hundreds of thousands of years rock easily could have moved down the hills in cataclysmic landslides and landed in the valley, far down from its sources in the Sierras. Yet the Spanish had never identified the origins in the mountains, many miles east and much higher up in elevation. Nobody had traveled to the exact place in the mountains because it was remote from the Spanish settlements in California, and it was not easy to find. The weight of gold also helped keep it hidden. The American River joins the Sacramento River upstream, but far downstream from the exposed gold seam over which the American River traverses. Gold flakes are too heavy to travel far in water, even in a fast-running river like the South Fork of the American during a snowmelt in spring.

An enterprising American commercial iconoclast and entrepreneur, John Sutter, decided to be the first person to go high up on the South Fork of the American River. He was seeking fresh timber, not gold. He established his mill in Coloma in an attempt to get timber nobody else had harvested. In 1848, in the midst of building that enterprise, one of his employees, John Marshall, discovered gold flakes in the water tracings of the mill, which had turned over sediment around the river. Weeks after Marshall’s discovery the word was out. The mill site was soon overrun with prospectors. The following year the hills contained as many prospectors as oak trees, the “49ers” of the California gold rush who panned for gold.

What are the lessons from this story for the Internet gold rush? Stated broadly, gold rushes possess four features:

1.  The value remains unknown for some time and is discovered to the surprise of the discoverer and all others.

2.  Once the value becomes known, knowledge about it spreads rapidly.

3.  Many prospectors explore and invest in the hope to gain later.

4.  Every prospector takes action quickly, believing that only the earliest movers will reap part of the newly discovered value.

There is another lesson embedded in this story, which later chapters will return to: a few of the earliest pioneers profit in the long run, but only a few. That last lesson, however, requires several steps to illustrate, and the first of which is explaining the rush. This chapter discusses the first step, why circumstances led to the Internet gold rush—namely, the simultaneous exploration of value by many market participants. That is also equivalent to asking why everyone waited until a specific moment in time, so an oversimplified statement can summarize the puzzle one must explain: if a gold rush begins when a great deal of activity follows a quiet moment, what triggers the change? To understand the timing of a rush it is necessary to explain what circumstances prevented the Internet rush from happening earlier. As the above story suggests, typically some new information acts as a catalyst, and the crucial question becomes a query about what action led to new information, and whose knowledge changed and why.

Looking at it from this angle, the puzzle requires stretching the metaphor about the gold rush to make it fit events in the Internet. There were, in fact, two overlapping periods of exploration. The first period involved many prospectors looking for Internet gold, but nobody shouting, “Eureka, I found it!” Stephen Wolff’s discussions to privatize the Internet started this period of exploration. It led to the founding of the for-profit data carriers PSINet and UUNET in late 1989 by William Schrader and Martin Schoffstall, and Rick Adams, respectively. It also led to a slow trickle of entry by commercial ISPs into the US market for such services, by firms such as Netcom. It also led a few intrepid BBSs to add an Internet service. It also led CompuServe, a very large information service using BBS technology, to offer Internet service. As earlier chapters discussed, these were interesting events among the cognoscenti of the Internet. To the core of the communications and computer industry, however, a new valuable market had not been found.

The introduction of Netscape’s first commercial browser eventually started the second and much noisier rush. By the end of 1994, many commercial computing and communications firms had seen a beta version of Netscape’s product, and many were on the way to altering their plans. By the end of the summer of 1995, after Netscape’s IPO, every market participant was evaluating the value chain for the commercial Internet and web. Many venture capitalists were beginning to assess the prospects for newly founded firms in a variety of applications. A few of the early pioneers also began to reap large rewards about this time. Like a gold rush, new information about the source of value triggered these actions.

To summarize, by the end of the summer of 1995 events were set in motion that would last for a considerable length of time, events that contemporaries would label as a boom in investment. Netscape’s IPO is, therefore, probably the best symbolic event for marking the birth date of the commercial web, and, accordingly, the beginning of events that resembled an Internet gold rush.

The above also suggests why the metaphor about the gold rush fails eventually. While the Netscape IPO did generate a short and intense period of prospecting by entrepreneurs, the boom was sustained by more than just the immediate and one-time reaction to new information. Different factors explain why the boom sustained itself long after the initial rush of 1995 and 1996.

This chapter, in contrast, focuses on understanding the sense in which the metaphor about a gold rush actually explains events. It focuses on commercial markets before the Internet rush, say, in early 1994, before participants were informed about the value of the Internet. Why had only a few major commercial firms prospected for value in the commercial Internet up until that point? This chapter explains why circumstances left many potential participants in doubt, and that illuminates why those who came from the edges would have such a large impact on catalyzing commercial markets.

The Collective Shrug of ’94

Looking back at the commercial computing market in 1994 requires a big change in perspective for an observer from the future. While years later it would be obvious that the privatization of the Internet resulted in an extraordinarily valuable gift to the US economy, most contemporaries in 1994 did not appreciate the value of that gift at the time it was given. Most contemporaries did not appreciate the NSF’s plan to privatize the Internet. Most of them just shrugged.

The marketplace primarily attracted prospectors from the edges. By early 1994 the catalytic pieces were all in place for motivating investment to meet the anticipated demand. The plans for privatization of the Internet were almost complete and the web was beginning to diffuse, available as shareware for anybody to use and assess. Several commercial iconoclasts had made investments in carrier markets and access markets, demonstrating the viability of providing commercial services.

The actions of those early pioneers did not generate a reaction. No prominent business magazine in early 1994 hailed PSINet or UUNET as leaders of an Internet revolution. No Wall Street analyst changed his or her recommendations in early 1994 from sell to buy for the stock of prospecting firms who had a great deal to gain from privatization. No venture capitalist made an explicit effort to round up the leading talented Internet entrepreneurs in early 1994 and fund a portfolio of the most sensible plans for new Internet services, even though later events illustrated beyond any doubt that it would have been extremely profitable to do so.

No three established firms could better illustrate the causes of the shrug than the three firms best positioned to take advantage of the commercial Internet. IBM, MCI, and Cisco each held a unique position. Each firm had experience serving the need of users in the research-oriented Internet. If anybody would invest heavily in anticipation of the coming Internet gold rush it would be these three firms.

Both IBM and Cisco focused a substantial fraction of their business on enterprise computing, and there was a straightforward explanation in 1994 for the lack of investment in Internet products and services. Most customers did not ask for it in 1994. Most customers were in the midst of converting large-scale computing to client-server architecture or any other form of local networking. Client-server computing was the revolution-du-jour among IT consultants and others in the enterprise computing market. It typically had a workstation act as server of data, while personal computers on a local area network acted as clients, or receivers of data. According to the standard mantra, the Internet contributed to client-server networks in some useful ways, since exchanging data between computing systems on different local area networks could be cumbersome in other formats.⁴ Some consultants were disdainful, however, and recommended that secure electronic commerce use another approach, something called electronic data interchange (EDI), not the protocols of the Internet.

Forward-looking management should have been able to see beyond short-term demand, nonetheless. The short-term focus of customers on client-server should not have deterred investments by forward-looking managers into TCP/IP services, which were about to grow in demand.

IBM would have had the best position if it had only acted on it. IBM had a unique position within the NSFNET backbone. IBM also had strategic advantages in enterprise equipment markets. Its involvement with the NSFNET gave its employees in the research division early insights into how to make TCP/IP-based equipment, especially routers. There was one obvious hitch. IBM’s research division handled the activity, and that needed to be translated into viable businesses in other divisions within IBM. Even though the wild ducks had a good idea this time around, their initial experience inside IBM would be mixed.

As it turned out, by 1994 most of the attempts to translate the research division’s insights into products did not pan out. One of these initiatives was touched on in prior chapters. For example, it had become rather clear by 1994 that IBM’s data carrier, ANS, did not and would not be able to assume a dominant position during the transition to privatization.

That was not the end of the commercial opportunities, however. ANS did not dominate the backbone, but IBM did operate one of the largest worldwide ISPs for its clients for the next few years. While ANS eventually was sold to AOL to enhance its consumer business, IBM’s access business for commercial clients did thrive for a few years, being regarded as one of the largest and most reliable access businesses for large business users.

The commercial experience with the equipment business has its roots elsewhere inside IBM. In the early 1990s, IBM’s enterprise division refused to be involved with any equipment using Internet protocol. Its division favored selling proprietary equipment, such as Token Ring and OS/2, which yielded higher margins for its sales force. Symptoms of the broad commercial problems affiliated with this practice were already apparent by 1994. While a few loyal buyers continued to use only IBM equipment, by the early 1990s many enterprises had made a deliberate choice not to, investing in much less expensive nonproprietary configurations of client server networks. These consisted of Ethernet-based local area networks, Microsoft operating systems in clone PCs, as well as servers coming from many firms, such as SUN Microsystems. By the early 1990s others firms—such as Cisco, 3Com, SUN, Compaq, and Novell—had taken advantage of the situation and established strong commercial positions.

Louis Gerstner was hired from outside as CEO in April 1993 with a mandate to change the financial situation in the organization. He was brought in because the company was in the midst of a financial crisis, ostensibly caused by declining demand for its mainframes (being replaced by client-server architectures), a crisis exacerbated by an almost willful lack of planning by Gerstner’s predecessors for the structural changes wrought by client-server architectures. As an outsider and former consultant for many of IBM’s clients, Gerstner perceived the crisis in broad terms, and also blamed the crisis on policies that placed too much emphasis on selling proprietary equipment instead of listening to customer preferences.⁵

Gerstner met with tremendous internal resistance trying to move the firm to support non-IBM technologies, such as Unix and Microsoft-client software. In that light it is hard to know just how much of IBM’s subsequent failures were collateral damage from the reorganization and how much were a by-product of prior myopia. In any event, IBM never became a strong router and switch supplier for enterprises.⁶

The silver lining of Gerstner’s strategy became more apparent after the Internet gold rush began. It positioned IBM for what turned out to be an enormous market opportunity. As IBM’s employees became more vendor neutral in their advice, their consulting business for enterprises would thrive. Later chapters will describe how this vendor-neutral approach became the foundation for the financial rejuvenation of IBM.

Cisco’s actions illustrate what commercial opportunity IBM lost in equipment markets. Cisco began as a supplier of routers and hubs for academic ISPs and operations. The company had specialized in the routers and hubs that stood between different computing systems, reading addresses on data packets and redirecting the packets from one computer to another. As late as 1993, it was justifiably obscure in comparison to the leading computing equipment firms in the world.

By 1994 Cisco’s management had adopted a strategy to expand far beyond just hubs and routers, positioning Cisco as the leading Internet equipment supplier for enterprises. Cisco’s strategy would have been a good idea whether or not the commercial Internet grew at a fast or slow rate, but it looked like a spectacular idea when the gold rush arrived. Cisco’s organizational change positioned the company for dramatic growth in response to increases in demand for its products. It also put Cisco in a good position to supply products it had not anticipated with appropriate research and development; Cisco planned to buy a leading product through acquisition of start-ups. Indeed, many years later it would be apparent that (perhaps) nobody profited more from the commercialization of the Internet than Cisco’s stockholders.

The most radical feature of this new strategy was its stated reliance on acquisitions, which left much of the risk for technical inventions to others and deferred the growth of Cisco’s product line until the company’s management needed it. The firm declared that acquisitions would become a deliberate piece of its strategy to grow, making this one of the firm’s core tenets. Normally a firm that acquires others faces risks related to overpaying, not getting sufficient value from the acquisition for their money. In addition to that risk, there were two primary strategic risks in Cisco’s declaration, and these would have sunk most firms: Cisco’s management had to absorb new acquisitions fast enough; firms had to supply acquisition targets fast enough. If Cisco failed at the former, it would drive up the internal cost within Cisco’s organization. If the latter did not occur, it would drive up the external price to Cisco’s stockholders, which would reduce the potential gains.

Realizing the ambition set Cisco on a path that no firm had ever pursued at such a large scale. Cisco would invest in lowering the transaction costs and frictions normally affiliated with so many acquisitions. It routinized processes for acquisitions and began educating others about its aspiration. Altogether Cisco made seventy-two acquisitions between July 1994 and December 2000, with sixty-five of them after January 1996, and twenty-three in the year 2000. During the Internet boom its market capitalization was valued at over $500 billion (in 1999), making it the most valuable company in the world for a short time. As later chapters will discuss, the stock market eventually came back down to earth, but for most of the late 1990s it valued Cisco highly, which helped it make stock-only deals for acquisitions, further lowering the price of an acquisition.

Cisco’s early repositioning contrasts with its other competitors in 1994. Most equipment firms took partial steps toward strong positions in markets related to the Internet, such as enterprise servers. This included companies like Nortel, Novell, Alcatel, and 3Com. None of them made such overt changes in strategy as early as Cisco. The equipment division of AT&T, which became Lucent in a few years, also did not take action to anticipate the Internet. In general, these firms did not consider changing their strategic direction until the winter of 1995, when such changes become part of conventional wisdom for everyone who had observed the functionality of Netscape’s first commercial browser. In that sense, Cisco was a rare early prospector.

MCI’s experience after the NSFNET also contrasts with IBM’s. First, MCI had developed skills and assets it could translate into a business for national data transport over its own fiber network; later acquisitions further enhanced that capability. Second, MCI could offer ISP service for both homes and businesses in the coming commercial data market. This business was further enhanced by MCI-mail, an Internet-compatible service. After the Internet gold rush arrived MCI gained a large amount of business, especially for service to homes.

MCI’s actions resembled that of two other early movers and incumbents, Sprint and BBNNET. Each had significant experience carrying international data traffic for the NSF. Each also managed to adjust to privatization, at least at first, positioning for further business.⁷

MCI, Sprint, and BBNNET can also be understood in contrast. They were the exceptions that identified the rule. There was no better illustration of their uniqueness than to contrast them with AT&T, the telephone company headquartered in New Jersey. Divestiture had spun the local telephone companies off as separate firms, leaving AT&T as a long-distance company, competitor to MCI and Sprint, as well as an equipment supplier, which would eventually be spun off as Lucent. AT&T did not move as aggressively into Internet services. It was only when the Internet began to enter widespread discussion in 1995 that AT&T moved into providing backbone and access services, both to businesses and to homes. In brief, AT&T was later than its most direct rivals, MCI and Sprint.

AT&T’s managers were not alone in their hesitance. Their closest cousins, the seven other local telephone companies created by divestiture, also did not invest in Internet service. Most had no immediate plans to provide Internet service in 1994. Instead, they spun their wheels on many projects that did not pan out.⁸

Why did so many firms such as IBM and AT&T ignore investing heavily in the Internet just prior to the boom? Overall their lack of action can be interpreted in three (overlapping) ways:

•  a situation in which the Internet lacked internal “champions,” perhaps because of the expectation that Internet services would cannibalize too many revenue streams at existing business;

•  a misunderstanding of the potential for the Internet, perhaps due to a commitment to an alternative technological vision or forecast;

•  a situation in which the Internet benefited many users at once, perhaps because no single firm had incentive to nurture adoption that seemingly did not directly contribute to their own bottom line.

In practice these three motives yielded similar outcomes, no investments in Internet technologies. It left market opportunity available to firms, such as Cisco and MCI, who had short-term market reasons to act, as they met immediate market demand.

Errors in the Consensus

In early 1994 Microsoft was one of the largest software companies in the world, and, by many accounts, possibly the best managed. By the early 1990s Microsoft had a track record of successfully forecasting the commercial value affiliated with new technical opportunities in personal computing. Management in Redmond believed its success partly arose from its systematic studying of the actions of many other firms before settling on strategic plans.

Due to this systematic approach, Microsoft’s actions reflected a critical analysis of the consensus of views elsewhere in the industry. Microsoft’s mistakes can illustrate the types of forecasting errors made elsewhere at the same time.

Microsoft’s decisions occurred within the context of several complex events. In early 1994 Microsoft was in the midst of developing a product that later would be called Windows 95. Since 1992, long and elaborate planning had gone into its design. This product was supposed to replace DOS (disk operating system) as the standard OS (operating system) in PCs. Windows had been an application built on top of DOS, and Windows 3.0 and then 3.1 were the versions of Windows most widely in use in 1994. Microsoft sought to make a better version of Windows the next widely used operating system in PCs, and, at the same time, cement itself as the central coordinator of application software development for the PC. By 1994, the company was late in executing these ambitious plans, but no viable competitive alternative had arisen, so it appeared it would succeed, nonetheless.⁹

At around the same time, management began making plans for and subsequently investing in MSN, a proprietary network that would have content for home users. The management in Redmond made plans that called for a gradual development cycle over many years, anticipating that the mass-market opportunity for MSN would emerge slowly, giving the company enough time to learn from its experience offering a proprietary dial-up service.¹⁰ In short, in early 1994 the architects of Windows 95 and MSN were confident that the main competitive issues affiliated with the Internet had been addressed, and the biggest concerns were far into the future.

Microsoft’s plans for servers reflected the same outlook. Reading TCP/IP compatible files had been a standard feature of Unix systems for some time, as it was an outgrowth of the military’s procurement requirement—all Unix systems had to be TCP/IP compatible. Unix was commonly used in enterprises as a server in a client-server architecture. To compete with Unix-based servers, Microsoft made Windows 95 and its server compatible with TCP/IP,¹¹ anticipating that others might build Internet applications on top of the OS. In these plans Microsoft accepted the conventional wisdom that the Internet would become one of several pieces of plumbing for servers, but largely reside in the background. Placing TCP/IP in its servers illustrates one additional aspect: the standard assessment relegated the Internet to status as internal plumbing for an IT system, not as a customer-facing aspect of software.

One of Microsoft’s most successful enterprise applications in the 1990s, e-mail for business enterprises, ended up benefiting from the Internet, but even this experience shows the way in which Microsoft misperceived the future. In the early 1990s Microsoft’s strategists (correctly) anticipated demand for business-oriented e-mail, and they had begun designing an application, later known as Exchange. Like many of Microsoft’s other products in the early 1990s, Exchange was not conceived as a part of a broad Internet strategy. Instead, Microsoft sought to develop applications for enterprise users of Windows NT and its other server software employed in mainstream client-server architectures. This software would reside on local area networks (LANs) inside enterprises.

Exchange’s design process followed the textbook for an exemplary product development. Not long after taking responsibility, Todd Warren, the group product manager for mail products from 1990 to 1993, led his team on an analysis of the strengths and weaknesses of all competitive LAN-based electronic mail products. In Warren’s words,

Exchange was an amalgam of all features of competitive products at the time, prioritized by customer wants and needs.¹²

What was the best design for electronic mail for business in the early 1990s? Microsoft’s team concluded that Lotus Notes, designed by Ray Ozzie and his colleagues at Lotus, offered an attractive model for users, and it embedded a vision of the future that Microsoft would have to match in functionality. From former IBM employees who had moved to Microsoft, the Microsoft design team also learned greatly, embedding Exchange with multiple features needed at enterprises, such as calendar systems, designed specifically to meet the needs of a senior executive. On Bill Gates’s insistence, Microsoft designed its standard implementations to work best with the SQL server and Microsoft’s other LAN products.

Lotus contained one crucial design feature, its reliance on a proprietary solution for internetworking. Lotus Notes had a tight and elegant integration, which worked optimally within one corporate network under one manager’s control, and Notes addressed internetworking issues with the same elegant solution. That solution came at a drawback: it could not integrate easily with other types of systems, and especially not with systems owned by others in different locations. It was potentially not a drawback; it was the largest enterprise e-mail system in use at the time, and as long as it retained its status as the largest, most systems would conform to it.

Microsoft’s designers anticipated the spread of some general standards for internetworking, such as those coming from OSI. They expected generic standards for internetworking would eventually play to Microsoft’s competitive advantage because it played to Lotus’s competitive disadvantage. They anticipated “loose coupling of networks,” giving Exchange a technically viable path for linking up geographically dispersed e-mail servers and clients provided by multiple firms. In other words, with a viable internetworking solution from a third party, Microsoft could be a late entrant into this market and viably compete. It did not have to engineer solutions to internetworking issues and could take for granted that establishments would be able to manage the internetworking functions without Microsoft’s help.

Consistent with this forecast, Microsoft designed Exchange with flexibility at the enterprise gateway. At this point they made their fateful—though, as it turned out, not fatal—choice and embedded the erroneous forecast in their design. Their analysis did not conclude that SMTP, the standard mailbox for Internet e-mail, would be as competitively important. Rather, they concluded the OSI had a strong likelihood of being widely adopted, and its electronic mail design, known as X-400, offered a useful roadmap for desirable features. Microsoft followed this road map, using the Internet gateways to hand off electronic mail flowing outside a corporate enterprise.

As earlier chapters have already hinted and the next few chapters will discuss, OSI did not become widely adopted in the 1990s. Like others who believed this consensus forecast, Microsoft’s forecast turned out to be wrong. However, sometimes it is better to be lucky than right, and, as it turned out, X-400 was sufficiently flexible to serve Microsoft’s strategic aims. Exchange could be tailored to fit TCP/IP traffic. In other words, in spite of the forecasting error, Microsoft’s system handed off traffic at the gateway, and the firm got the competitive advantage it sought, just not by the path it anticipated.¹³

What lesson does this illustrate? Like the prior examples, the consensus view about e-mail turned out to be mistaken, because it did not anticipate a central role for the Internet in the design of enterprise electronic mail. Even at one of the best-managed firms at the time, the newly designed electronic mail product reflected this small role for the Internet based on TCP/IP.

Heeding Cassandra’s Warning¹⁴

Microsoft’s approach to the Internet came straight from Gates’s assessment of the conventional wisdom, and after he did his homework. When Microsoft made an error it illustrated something unique about Bill Gates and something general about the prevailing view.

The unique part of Microsoft was Bill Gates’s managerial style. Bill Gates constantly sought to educate himself and use his learning in every aspect of decision making. He was technically skilled enough and energetic enough to debate virtually any employee about the strategic merits of a decision. He also had an extraordinary amount of personal authority, so his judgment carried the day during disputes.¹⁵ In short, employees and other executives took it for granted that little got done without Gates’s understanding and approval.

Gates’s reasoning about TCP/IP and the web arose from the consensus in commercial computing. In the spring of 1994 Microsoft had held a major strategic meeting to discuss the Internet, as part of potential design changes for what would become Windows 95. After long discussion Gates concluded—like much of the rest of the software industry at the time—that browsers could not be a profitable stand-alone business and, therefore, would not make much money in commercial markets. It was not difficult to understand the origins of this point of view. Mosaic had built functionality on nonproprietary protocols and standards, which rendered them vulnerable to imitation, widespread entry, and intense competitive rivalry, which inevitably would reduce the software’s price. So went this view, anything built on top of such common elements would become—in the lingo of the time—a “commodity,” and any such application could not and would not generate much revenue for long.

Most important, Gates made a decision that showed how certain he was that the consensus had it right. He did not concede to the Internet enthusiasts that there might be some set of market circumstances that could conspire to produce a profitable Internet service or application worthy of Microsoft’s attention. Rather, Gates concluded that no investment in Internet applications on the desktop had any potential payoff, not for Microsoft, nor for much of anyone else.¹⁶ Hence, he chose not to invest in any application development, not even development of working prototypes of web-based applications.

The forecasting and strategic flaw became apparent later, but not at the time. Had Gates had even a modicum of doubt, it would have been strategically sensible to authorize a few exploratory projects in April 1994. These would have involved a few employees, and would not cost much money. With the benefit of hindsight Gates certainty appears, at best, to be penny wise and pound foolish. A less flattering characterization would say he was—take your pick, as none of these are a compliment—rather unimaginative, too distracted to be thorough, overconfident in his judgment, or merely arrogant.

Microsoft employed a number of wild ducks who disagreed with Gates, however, and, to the eventual benefit of the firm, they had the courage of their convictions. In the summer and autumn of 1994, a small group inside Microsoft undertook a review of trends; Brad Silverberg, a comparatively senior manager who reported to members of the strategy team, organized this group.¹⁷ These employees ostensibly did something that was not unusual at Microsoft, thinking about new initiatives. What they actually did, however, was start an unauthorized skunk works devoted to examining the commercial potential of the Internet and web, and in depth.

A skunk works could be an organizational home for wild ducks, seemingly mixing metaphors from the animal kingdom.¹⁸ It is housed away from the main operations of an organization, sometimes in secret or with organizational barriers, and often with top management support for these barriers.¹⁹ Typically the development projects involve something of value to the future of the organization but are not directly connected to its present operational or service missions. Sometimes a skunk works has the approval of senior management, and sometimes, as in this case, it does not.

No one paid much attention to the team studying the Internet, and, by the same token, they received few resources. As events began to unfold in the fall of 1994, however, this team decided to come out of hiding and conduct many wide-ranging conversations with existing stakeholders inside Microsoft. They eventually received permission to license Mosaic, finishing negotiations in January 1995. They refined their vision about the future and Microsoft’s potential strategy for addressing it, and they internally publicized their views and efforts. A member of this team, Ben Slivka, articulated the vision in a widely circulated memo titled “The Web Is the Next Platform.”²⁰ Eventually this memo would go through redrafts, finishing on version 4.0, and would provide the beginnings of a framework for Microsoft’s Internet strategy. Slivka would later be rewarded for his efforts with an extraordinary responsibility, as lead programmer for Internet Explorer, and he stayed there through versions 1.0, 2.0, and 3.0.

The group eventually did change Gates’s mind, but much later than the change in the consensus. As later events would illustrate, this turned out to be early enough to allow Microsoft to participate in the Internet gold rush, though that was far from assured when Gates first announced the change in direction. The key event happened in April 1995, and it illustrates Bill Gates’s stubbornness and the opposite, his willingness to change his mind in the face of contrary evidence. It also illustrates where the common consensus went wrong.

The team organized an evening of surfing for Bill Gates, with instructions about where to go and what to look for. Gates had not tested the web extensively with Mosaic, and he had not gone back to the web after the explosion of the entry of Netscape’s browser in February/March 1995. He was unfamiliar with the range of capabilities the browser could perform by this point. Gates spent the better part of the night surfing, and it changed his view about the commercial potential of the software. A month later he issued a memo titled “The Internet Tidal Wave,”²¹ which other prior chapters have mentioned, and later chapters will return to. It was especially significant, because it announced the realignment of priorities for strategy inside Microsoft.

Having won the argument in principle, at the same time Slivka issued the fourth and final version of his memo. Slivka’s memo goes to sixteen single-spaced pages while Gates’s goes on for nine, so Slivka’s provided more detail about both the opportunities and potential competitive issues. Nonetheless, Gates was CEO, so his memo had authority as official policy. After May 1995, many executives inside Microsoft debated how to address “The Internet Tidal Wave,” as the memo took on canonical status inside the firm.

Referring to Netscape in his memo, Gates states, “A competitor was born on the Internet,”²² that is, a competitor to Microsoft. He goes on to outline why he interprets a browser company as a threat. It was not the company, per se, that posed any threat to Microsoft’s interest, but the potential for its product to support an entire ecosystem of applications. Its standards were becoming pervasive on virtually every PC, which other applications would use.²³ Gates worried about two scenarios in which Microsoft’s operating system could become “commoditized.” One scenario involves Netscape directly, where Gates asserts:

They are pursuing a multi-platform strategy where they move the key API²⁴ into the client to commoditize the underlying operating system.

Translation: Netscape helped to coordinate developers whom Microsoft would have been coordinating, and if those developers provided functionality equivalent to PC software, Netscape’s browser substituted for Microsoft’s operating system as the point of access for those applications, which reduced the value of Microsoft’s operating system.

The other scenario is less specific about who will take the action. It involves the combination of several component firms developing a new device. In this vein he says:

One scary possibility being discussed by Internet fans is whether they should get together and create something far less expensive than a PC which is powerful enough for Web browsing.²⁵

While the memo also remains rather unspecific about the timing for this particular forecast, the fear and paranoia comes through. Events could generate an outcome that threatened a very profitable business at Microsoft.

Gates makes it clear that the browser was the most urgent concern. Browser technology held the potential to radically change the way a mass market of users employed the PC, possibly leaving Microsoft outside its central standard-setting position, or leaving Microsoft in the central position of a less valuable PC.

Immediate events appeared consistent with the paranoia. Netscape had begun to make money from sales to businesses and employed a unique distribution mode involving “free” downloads by households and students, anticipating revenue from business licensees.²⁶ Importantly, Netscape had begun a program to invite third-party vendors to make applications compatible with the Netscape browser, practices aimed at influencing the rate and direction of innovation. Netscape had also begun to expand its product line into complements to browsers, such as products for servers and areas of related networking.²⁷ This market-development activity would bring the browser and the Internet into play as an effective way to achieve mass-market e-commerce and content. Responding to it all would eventually become a matter of competitive urgency at Microsoft.

What do these later actions say about Microsoft’s earlier errors? There was never any technical error at Microsoft. Both Microsoft’s strategy and conventional wisdom were based on up-to-date technical information that someone with sufficient technical skill could understand. Rather, Gates misinterpreted the value of the Internet’s commercial prospects. This error would take three interrelated forms in its conventional assessment:

1.  Underestimating the Internet’s value to users;

2.  Underestimating the myriad and clever ways entrepreneurs and established firms would employ Internet and web technologies to provide that value for users;

3.  Underestimating the ability of Internet firms to support applications that substituted for Microsoft’s in the marketplace.

These three errors were common at other firms in the spring and fall of 1994. Only two additional errors were unique to Microsoft. First, Bill Gates recognized months later than did most computing industry participants that the first three errors had occurred. In addition, Gates had not authorized any investments in the event that the conventional assessment was in error, so by the spring of 1995 Microsoft had limited short-term ability to follow through on the anticipated growth in demand for Internet services. In particular, the first browsers in Windows 95 were not designed to support HTML, so, as a result, software application firms got their tools and solutions in 1995 and 1996 from shareware and other firms, such as Netscape.²⁸

Microsoft’s unique problems would have consequences for others, however. If applications could come to users without using Microsoft’s OS then such applications held the potential to reduce the prices Microsoft could charge for its OS in the near-distant future. That made the Internet a threat to Microsoft’s interest in its biggest market. This was not an outcome Bill Gates would accept passively. He would resist it, let others know about his resistance, and try to realign the actions of other market participants. Later chapters will describe these actions in detail.

In contrast, in 1995 virtually every participant in commercial computing looked at prototypes for the commercial browser in a more positive light than Gates. Many participants in computing and communications first experienced the Mosaic browser or the Netscape browser with a sense of joy at the technical and commercial possibilities. The browser illustrated the potential to help create value that had not been as apparent to observers only a short time earlier.

For many of these observers, the Netscape IPO confirmed the unique status of Netscape’s browser. The IPO followed an atypical script.

At a typical initial public offering, or IPO, a young firm receives money from investors in exchange for shares of common stock. (Before the IPO, the shares are privately held by founders and venture capitalists. After the IPO, the shares can be traded on a stock exchange, such as the New York Stock Exchange or NASDAQ.) Typically, the number of shares to be issued and the price are decided by the company’s investment bankers (also known as underwriters) in consultation with management. The underwriters have two objectives that run in opposing directions. First, as agents of the company, they want to raise as much money as possible for a given number of shares. Second, to protect their own firm from loss and to maintain their reputation in the marketplace, the underwriters want to be sure that all the shares on offer are sold in a short amount of time (a few minutes after the offer “goes live”).

The bankers’ success in balancing these two objectives can be gauged by looking at the performance of the stock in the days and weeks after the issue date. On the one hand, if the stock price goes down significantly on the first day of trading (or after a short time, say, a couple weeks), then initial buyers will be faced with a loss; as a result, they may felt cheated and lose faith in the underwriter who sold them the stock. On the other hand, if the price increases significantly on the first day (or after a short time, say, a couple weeks), then the young firm’s management may feel cheated. They could have sold the same number of shares at a higher price or sold more shares at the same price. Either way, if the price rises, post IPO, the company has left money on the table. (Academic research indicates that the deck is stacked somewhat in favor of investors and against the companies. IPO stock prices typically rise on the first day of trading and stay at the higher level thereafter. Gains are typically on the order of 16 percent.²⁹)

Netscape’s IPO departed from the standard script, and it was never settled whether the departure was accident or deliberate. In the first day of public trading, the stock started around $28 a share and then shot up to over $75.³⁰ After the fact, nobody would admit publicly to a mistake. While the run-up conferred instant paper wealth to the founders and key management, the gain came at a seeming cost to the young organization. By setting a higher initial price, Netscape could have raised several hundred million dollars more than they did, while issuing the same number of shares (thus suffering no more dilution).

It gave the impression that Netscape was so confident it could cavalierly throw away wealth. Garnering attention from many corners, Netscape’s IPO became one of the most successful publicity stunts a young firm could have done.

This event had one other consequence, entirely unforeseen at the time. Later on, as more Internet and web businesses engineered their IPOs, many founders wanted to imitate Netscape’s pattern—namely, generate large run-ups in the stock price on the first day of trading. Such run-ups were viewed as a sign of success. Cooler heads could not prevail against the request in many later cases. Hence, many of the IPOs in what became known as “the dot-com boom” became lucrative for those who got access to new stock, who often profited merely by acquiring it at the IPO price, and holding it for a short period of time before selling.

The Cost of Oversight to Investors

The shrug of ’94 was costly for investors, because indifference led many investors not to invest in ways that would have been extraordinarily profitable. For example, foresighted investors who made propitious investments in 1993 and sold the shares between 1997 and 1999, or even 2003, would have made significant financial gains. Said another way, in retrospect two comparatively rare types of investors principally gained: those who invested in the equity of the right established firms, and those who invested in the right start-ups.

Specifically, investments in established companies that later directly benefited from the growth in the Internet—such as IBM, MCI, PSINet, UUNET, Cisco, and Intel—would have yielded huge returns on the dollar. To be sure, such investors would have needed considerable courage to follow their convictions. None of these firms, with the possible exception of Intel, were considered to have large growth potential in early 1993.

Table 6.1 gives a flavor for how large the gains could have been. Specifically, it answers the question: if an investor put $100 into company x on June 30, 1993, how much money would that investor get back on June 30 in 1994? What about the year after that? And so on. This is a standard financial exercise. The table calculates the date at which the maximum return would have been realized, and it shows these returns over a ten-year horizon, providing a flavor for both the up and down.

Table 6.2 shows that valuation of most of these companies between 1993 and 1994 did not change dramatically, reflecting the grip conventional wisdom had on investor assessment of these firms’ prospects. No major change in mainstream investing priorities occurred until 1995—and even then, the revaluations did not begin to accrue until a few months prior to (and then especially after) the Netscape IPO in August 1995.³¹ The first high-profile analysis of the Internet access market on Wall Street did not arrive until Mary Meeker’s at Morgan Stanley, which then resulted in a publication for general audiences in 1996, after her team had done related work for clients.³² By June 1996 the news had caused a change in views. Any general investor in the sector from June 1993 would have done rather well. That continued to hold through June 1997, ’98, ’99, and early 2000, and even later for some companies, such as IBM and Cisco.

A similar insight holds for venture capitalists. There was no noticeable change in the rate of first-stage investing by venture capitalists in computer and communications equipment companies in 1993 or 1994, nor in PC or networking software firms. Similarly, there was no significant change in the number or rate of initial public offerings by related firms. Later chapters will say more about this nonevent.

One example can illustrate how the consensus of conventional wisdom held back new entry. Indeed, this example illustrates the exception company that ran against the herd, and profited from it. In 1994, a few intrepid visionaries started Vermeer, a company seeking to develop server software tools and related Internet applications to take advantage of the newly developing World Wide Web. They later developed a software-authoring tool that ultimately became a Microsoft product called Frontpage.

This tale is well known because, several years after he sold out his shares, one of the company’s founders, Charles Ferguson, published an articulate memoir about his perspective on his start-up’s experience.³³ As Ferguson’s account makes abundantly clear, the commercial vision of his firm for the World Wide Web was not largely shared by others in the summer of 1994. He had a difficult time convincing anyone that a market need existed or that a firm started by outsiders had commercial potential, and he was met with considerable skepticism about the prospects for his firm.

We were dangerously ahead of the Internet curve; nobody knew what the hell the Internet was, and the Internet industry didn’t exist. And finally, we had a truly original product that didn’t fit into neat, established categories, so VCs perceived us as risky—particularly because a lot of them didn’t understand anything we said.³⁴

Being so unusual initially made it difficult for Ferguson and his business partners to raise money for Vermeer, but later that same uniqueness conferred enormous competitive advantages. It allowed Vermeer to position its product where few others could create equivalent value. It released its authoring tool in beta format in September 1995, not long after Netscape’s IPO.³⁵ Only a few months after its product release, Vermeer auctioned itself off in a bidding war between Netscape and Microsoft, making all the founders and investors quite wealthy.

Did that make Vermeer the canonical gold rush pioneer that made out well? This example illustrates another misleading aspect of the gold rush metaphor. In a gold rush the pioneers find the largest chunks of gold by virtue of being early. Vermeer is similar in that respect, in that they had the right idea in advance of the Internet rush. Yet, notice the key differences too: pioneers in a gold rush get their payoff comparatively quickly, if they get one at all. In contrast, Vermeer had little value at first, and really had no prospects to cash out early. Vermeer did not realize the gains from that idea until several years after the entrepreneurs hatched the idea. While comparatively quick, the risks also were enormous. A VC had to finance their firm in the interim, and share the risk with them over whether there would be a payout at all. In short, Vermeer did not arrive quickly and pan for gold in the hopes of a quick payout; it had pursued a unique path prior to the rush, built an enormous apparatus, and once the other gold diggers arrived, it sold its position.

Working Prototypes and the Consensus

Why did the near future remain elusive? The complexity of privatization made it difficult for any observers—both those close to it and those distant from events—to grasp the consequences of the NSF privatization. Even if Stephen Wolff had taken out a full-page ad in every major daily newspaper in the summer of 1994 and explained the NSF privatization in great detail, few participants in the computing or communications industry would have grasped the implications.

It was not impossible for an imaginative technologist to foresee much of what commercial markets might be able to do, and that yielded a few entrepreneurial prospectors. Management at PSINet, UUNET, MCI, Sprint, and Cisco invested in anticipation of the commercial Internet, and, to their credit, before the summer of 1994. Ben Slivka at Microsoft foresaw what was coming, and in detail, and in the fall of 1994. Jim Clark and Marc Andreessen and Charles Ferguson also saw opportunities and acted on their visions. There were a few others, and those few did not work in influential positions in private firms.

Everyone else in commercial markets was hesitant, even though later events would make it abundantly clear that early actions would have yielded enormous payoffs. What were they waiting for? A compelling working prototype from a commercial firm had to be sufficiently advanced in its operations to sway more than a visionary technologist. It had to be compelling enough to sway entrepreneurs and VCs who would invest time and money in new ventures. It had to be good enough to convince CEOs at established firms to gain board approval to commit to major investment projects. These people needed to observe an entire system operate with all its layers working together, delivering service to users, under reliable contracts, supported by established firms. In short, it had to demonstrate how the technology could generate revenues from applications.

The NSF-sponsored Internet produced plenty of the working prototypes, but not those needed to overcome the hesitance found within commercial markets. A commercially viable working prototype could not exist until the NSF finished announcing its privatization plan, as it had in the fall of 1994. That made the Internet ready for a catalytic working prototype, one that would generate a killer application in the market. Netscape walked into that situation, and its browser dazzled.

The same prototype was shown to just about everyone only a few months apart, so perceptions about the commercial opportunities changed nearly everywhere, and did so at just about the same time. Very quickly the consensus changed, and the shrug of ’94 disappeared. Many established firms recognized that they had to respond, and quickly. This realization initially would generate activity, and resemble an Internet gold rush. It would begin with a decided tilt toward entrepreneurial activity and innovation from firms who had previously been on the edges.

¹ BrainyQuote.com. August 4, 2010. http://www.brainyquote.com/quotes/quotes/j/jimclark406158.html. Accessed September 2013.

² Coloma today supports a well-functioning, if somewhat sleepy, national park. The area also supports an active white-water rafting industry.

³ After being discovered in the mid-nineteenth century, it took decades to recover the precious metal thousands of feet below the surface. Only a few of those mines operate today.

⁴ For a study of the challenges from that conversion, see Bresnahan and Greenstein (1997).

⁵ See, e.g., Gerstner’s (2002) account, which argued that, in retrospect, these changes were among those that were most beneficial for the long-term health of IBM.

⁶ Perhaps just as interesting, IBM’s close rival, Digital Equipment Corporation, faced a similar problem. Its own internal division committed to a proprietary local area network technology for enterprises, called DECNet, and put up considerable resistance to moving the firm to support other technologies. DEC’s equipment division for enterprises was, therefore, poorly positioned for the Internet gold rush.

⁷ For more on Sprint’s commercial actions, see the account in Hussain (2003c), or Meeker and DePuy (1996). The chief architect for BBNNET wrote about his lessons from that experience in Marcus (1999).

⁸ Ferguson (1999), pages 54–55, provides a sarcastic and caustic recounting of many of these business adventures, including deals with Ziff-Davis, 3DO, General Magic, and Lotus.

⁹ See Cusumano and Selby (1995) for extensive discussion.

¹⁰ See, e.g., Swisher (1998) for an accessible comparison of the AOL/MSN differences and similarities—especially their market positions and strategic approaches during the mid-1990s.

¹¹ Under the original design plans for Windows, there were two target markets, one aimed at PC clients and one at servers. TCP/IP compatibility had value for server software as a direct competitor to Unix systems. It also had value because it eased data exchange between server and client.

¹² Todd Warren, private conversation, May 29, 2009.

¹³ Todd Warren, private conversation, June 5, 2009.

¹⁴ In classic Greek mythology Cassandra can foresee the future but her prophecies are never believed, and her warnings never hinder her friends and family from taking tragic action.

¹⁵ This partially arose out of Gates’s driving personality and his smarts. It also arose from experience, as the leader of Microsoft when it displaced IBM as leader of the PC industry. See the discussion in Cusumano and Selby (1995), or Bresnahan, Greenstein, and Henderson (2012).

¹⁶ In this view, electronic mail would make its revenue through licensing the Exchange server, which supported Outlook on the client.

¹⁷ Ben Slivka, private communication, October 2008.

¹⁸ The phrase, skunk works, originated from the aeronautics industry. A project for the air force at a division of Lockheed in Burbank, California, had called itself the “Skonk Works” in a bit of salty humor about its own secrecy. The phrase came from Al Capp’s Lil’ Abner cartoon—the skonk works was a “secret laboratory” that operated a backwoods still. The label became well known throughout the industry, in part because it was considered humorous and saucy. Lil’ Abner’s publisher eventually asked Lockheed to change it, and “skunk works” emerged from there because the Lockheed facility was next to another plant that emitted fumes (Rich and Janos 1994).

¹⁹ The legendary leader of the skunk works at Lockheed was Clarence “Kelly” Johnson. He also helped establish these precedents. He was known as a technical leader who kept outside meddling and paperwork to a minimum, albeit, he would not be characterized as a wild duck.

²⁰ See Slivka (1995) for the fourth and final draft of this vision statement. A publicly available copy is at http://www.usdoj.gov/atr/cases/ms_exhibits.htm, exhibit 21.

²¹ A publicly available copy of Gates (1995) is at http://www.usdoj.gov/atr/cases/ms_exhibits.htm, government exhibit 20.

²² Gates (1995), 4.

²³ See Bresnahan (2003).

²⁴ An API is the application programming interface, that is, the frame seen by programmers when interacting with the program.

²⁵ Page 4 of Gates (1995). Attempts to realize this forecast were tried shortly after Gates predicted it, though the attempts largely failed in the marketplace. See Bresnahan (1999). They came to be realized much later, in the form of smart phones and tablets.

²⁶ As discussed in chapter 4, the browser was free, only for evaluation and educational purposes. They attempted to establish usage share through households by making it free, while collecting significant revenue from business licenses.

²⁷ Cusumano and Yoffie (1998) have an extensive description of how Netscape explored the commercial potential of many complementary service markets through site visitation of lead users and interaction with many user and vendor experiments.

²⁸ This is a necessary summary of a long set of complex events. See, e.g., Haigh (2008), Bresnahan and Yin (2007), and Bresnahan, Greenstein, and Henderson (2012).

²⁹ See Beaty and Ritter (1986), and Ritter (1991).

³⁰ See Cusumano and Yoffie (1998) for a fuller account.

³¹ Krol (1994) gives a good sense of movement prior to 1995. For a description of the type of resistance foresighted entrepreneurs encountered during this time, see, e.g., Ferguson (1999).

³² See, e.g., Meeker and DePuy (1996). This team of analysts was not the first one to organize a systematic analysis of the vendors in the market. The first-stage venture capital firms certainly were earlier. A few of these stories are described in later chapters.

³³ See Ferguson (1999).

³⁴ Ferguson (1999), 71.

³⁵ http://www.seoconsultants.com/frontpage/history/, accessed July 2009.