Risk & Innovation: Small Companies in Six Industries: Background Papers Prepared for the NAE Risk and Innovation Study (1996)

Simon Glynn

Despite the increasing number of very large software companies, an extraordinary number of new companies continue to enter the software industry. Indeed, the level of concentration in software has declined in the last decade, as the number of small companies has increased (see Table 1). These new companies flourish because of closely related changes in technology and markets: (1) the rapid pace of change in computer and software technologies; and (2) the high degree of fragmentation in markets as electronics and software seep into almost every aspect of production and consumption in the economy. This is reflected in one recent estimate that software is now critical to commercial technologies that account for 12 million jobs and $1 trillion in revenues in the United States (Catalanotto, 1994). This fragmented, rapidly growing marketplace is ideal for start-ups.

As well as the opportunities created by new technologies and new markets, the following factors seem to influence the emergence of new software companies:

• Relative ease of entry—It only takes a computer to make a skilled programmer into a new “software business.”

• There have been enough, highly public business successes that have started from humble beginnings to encourage entrepreneurs to enter the software industry.

• Software companies can have substantial value if sold, even though they are not necessarily long-term survivors.

• A public market exists for the shares of software companies at relatively high valuations, compared with other industries, on the basis of revenues or profitability.

Emerging software companies often exploit technologies or markets deemed too small or too risky for established players. Thus, new markets and narrow, niche markets that sometimes lead to considerably larger markets let new software companies develop the revenue stream, product, and core competencies of valuable new software businesses. On the other hand, once such new companies are established in a market niche, they in turn become vulnerable to new players with better ideas. Since the development cycle of sophisticated software is lengthy and requires highly focused skills, reacting to a competitive threat is usually not an easy task. As a result, software companies tend to be divided into three groups. The first group consists of the few that become large and develop the internal resources to have long-term staying power and to stay on the advancing technology curve. But these are exceptions. The overwhelming majority of software start-ups are in the second group that develops niche-market products, reaching revenues perhaps in the $5 million to $15 million range. The life cycle of these companies is also quite short-typically, they will either fail when their product life cycle has run its course, or be acquired by or

merge with other players to reach sustaining capabilities. The third group includes those software companies that for a variety of reasons are not successful, and fail.

This economic organization of the software industry also has consequences for new companies. It is increasingly difficult for smaller companies to grow to become very large as the cost of market access increases in a crowded environment. Small companies may identify new markets or new technologies, of course, but they are extremely dependent on access to the capital, markets, and management expertise required to expand these opportunities. This is already a constraint for start-ups and smaller companies and can be expected to become an even more important factor in the next few years, as the technological and strategic uncertainties in the software sector begin to disappear. How small software companies deal with these developments will largely determine their growth potential.

Spending for externally developed software may be divided into two types: prepackaged software (SIC code 7372) and customized software and services (including computer programming services, SIC 7371, and computer systems integration, SIC 7373). Data on these sectors are presented in Tables 2 and 3.

It is important to understand that internal development is not included in these data. In economic terms, the majority of software continues to be developed for internal use by companies—this spending for internal development of software code may be as high as $150 billion to $200 billion annually (National Research Council, 1990). But the preponderance of these internal development efforts is for maintenance and incremental improvement to existing software, not new systems (Organization for Economic Cooperation and Development, 1985).

Global revenues for customized software and services by U.S. companies in 1993 were estimated to be $38.7 billion and are expected to exceed $40 billion in 1994 (see Table 2). More than 40 percent of these revenues are from markets outside the United States, where experience in the technologically advanced U.S. market provides a competitive advantage. The majority of these firms tend to be independent and entrepreneurial. Indeed, many start-ups that are formed to develop prepackaged software also (initially) develop software on a contract or fee basis to raise capital for their internal development efforts (U.S. Department of Commerce, 1987). Several large U.S. players also compete in these markets, including Electronic Data Systems (1992 revenues of $7 billion) and several consultancies such as Andersen Consulting and SHL-Systemhouse (Ferne and Quintas, 1991; U.S. Department of Commerce, 1994). More recently, large defense contractors, including Boeing and McDonnell Douglas, have entered these markets because of their expertise in developing large systems (National Research Council, 1992b).

Larger computer makers such as IBM are also focusing on systems integration services as their customers migrate from high-end hardware to a complex environment of mainframe and distributed desktop systems.

Global spending in 1993 for prepackaged software (including operating systems) is estimated at $71.9 billion (U.S. Department of Commerce, 1994). The United Sates is by far the largest geographic market for prepackaged software, representing 45 percent of spending ($32 billion) in that year (see Table 3). Japan was second, representing 9.6 percent ($7 billion). The markets of western Europe as a group represented $25.7 billion, or 36 percent, of global spending in 1993 (U.S. Department of Commerce, 1994).

U.S. software companies dominate this sector. According to International Data Corporation (IDC), revenues to U.S. companies were nearly $50 billion in 1992, or more than 70 percent of global spending (U.S. Department of Commerce, 1994), Accurate estimates of revenues for PC-based applications software are not widely available; however, they may be estimated from 6-month data published by the Software Publishers Association (SPA)¹ (U.S. Department of Commerce, 1994). Using these estimates, spending for PC-based applications software in the United States and Canada totaled more than $6.6 billion in 1993.

These estimates for the different sectors of software development are summarized below:

Opportunities in software depend on innovation in computer technologies. In this sense, the development of networked personal computers and workstations marks the transition to a profoundly different environment for software development. IBM is currently the world’s largest supplier of software, despite its current difficulties; IBM’s revenues from software (including operating systems) in 1992 were $11.1 billion (Hodges and Melewski, 1993). Several trends have affected IBM’s thinking about the software side of their business (and by extension, the thinking of larger computer makers). First, software has developed as an opportunity that is quite distinct from computers. Operating systems and enterprise-scale applications software have become very expensive and complex to develop. For example, IBM’s OS/2 operating system is estimated to have required at least five years and 400 programmers, and cost perhaps $1 billion. As well, in recent years independent software companies have pushed advances in several areas in operating systems and applications software.

Second, as computer hardware is increasingly commoditized, differentiation is less on the physical performance of the electronics, and increasingly on the performance of the systems software, and the collection of applications software and services available to users. For example, the success of Apple’s Macintosh computer (whose development was mainly in sophisticated operating systems software) depended on the commercial availability of software designed and marketed by start-ups and smaller software companies. Consequently computer makers have learned to encourage independent software companies to develop applications based on their architectures.

These dynamics create opportunities for software companies. In contrast to customized software and systems integration, the personal computer and, more recently networked computing, are radically changing the demand for software by creating very high-volume markets. Indeed, by 1984 the installed base of PCs was 23 million machines, compared to less than 200,000 for large- and medium-sized systems (Steinmueller, 1996). These high-volume opportunities easily absorb the fixed costs of software development. Also, standardization of personal computer architectures in the United States has enabled software companies to create software and operating systems that can be incorporated by different computer makers. This is in marked contrast to Japan, for example, where six of the top 10 software companies are tied through industrial groups to a computer maker (the top four are NEC, IBM Japan, Hitachi, and Fujitsu), each using a proprietary operating system. In the United States, by comparison, none of the top 10 software companies is tied to a computer maker (Friedland, 1993).

In sharp contrast to the few remaining computer makers, most of the independent software companies are relatively small, entrepreneurial companies. Indeed, the concentration of the software sector has decreased in recent years. Using 1987 census data, the top 50 software companies (by revenues) represented 35 percent of all revenues from software in 1987 (U.S. Department of Commerce, 1987). This compares to 42 percent of revenues from the top 50 software companies measured 5 years earlier (see Table 1) (U.S. Department of Commerce, 1982).

In this sense, the larger software companies that many people think of as representative of the software sector, for example Microsoft or Lotus, are exceptions³ (See Table 4; Hodges and Melewski, 1993). In Utah’s “software valley,” for example, three-quarters of the more than 1,120 technologyintensive companies have fewer than 25 employees, and 50 percent have revenues of $200,000 or less. Similarly, more than 70 percent of the software firms in Massachusetts employ fewer than 25 people (Economist, April 3rd, 1994). A detailed description of these smaller companies emerges from data in the Massachusetts Computer Software Council report, Software Industry 1993 Business Practices Survey. According to these data, 31 percent of software companies have between 1 and 5 employees, and 71 percent have 30 or fewer employees. In contrast, fewer than 10 percent of responding software companies reported 100 or more employees (Massachusetts Computer Software Council, 1994). Table 5 presents a profile of the respondents.

Several characteristics—specifically, very low barriers to entry to new markets and technology cycles that create opportunities—shape the opportunities for these new companies in software. First, the initial capital requirements to start in software are extremely low (with the exception, of course, of intellectual capital). This is likely to be as true in the future as it has been in the past, as Larry Ellison, CEO of Oracle observes: “The weird thing about software is that it really is creating something out of nothing; a couple of programmers can really start in their garage. In fact, you don’t even need a garage” (Wiegner, 1992). These extremely low barriers to entry, especially in the decentralized, software-intensive low end of the hardware spectrum, limit the amount of risk an entrepreneur is required to accept in software.

Second, because of fragmentation and rapid growth in the market, start-ups in the software sector have an opportunity to identify new markets that are not yet attractive to larger competitors. The opportunities in educational software, for example, are increasing quickly as the installed base of computers in schools and homes increases and the sophistication of software improves to include full-motion video and digital speech. By recognizing and exploiting these trends, start-up software companies, for example Broderbund Software, Learning Co., and Davidson and Associates, have created valuable software product “franchises.” Broderbund has sold 8 million copies in 7 years of its best-known titles, the Carmen Sandiego line of geography and history games and the Print Shop, a line of printing and publishing programs (Selz, 1993). Similarly, Maxis, with its SimCity titles has created a new entertainment software genre.

Smaller companies also exploit the opportunities created by technological change. Quite apart from the question of whether new technologies have made software more (or less) difficult to develop,

current technology cycles continue to create the opportunities that create small software companies. Client-server architectures that allow individuals using PCs to access information and programs on “servers” are an example of this type of opportunity. The demand for software for these new client-server architectures was initially exploited almost exclusively by start-ups and small software companies. Powersoft, for example, develops software “tools” that make it easier for companies to write applications software for client-server architectures. By exploiting these opportunities quickly, smaller companies can achieve impressive size and financial results. For example, Powersoft’s revenues increased 351 percent in 1992, to $21.1 million. These opportunities are also attractive to investors: Powersoft’s initial public offering in early 1993 resulted in a price-to-earnings ratio of more than 90 (Hoff, 1993).

The opportunities for small software companies are bounded, however. The valuable strategic positions in software (and hardware) are already controlled by proprietary architectures or are currently contested by several architectures, all controlled by larger software companies (Torres, 1993). In this context, launching a competitive challenge is very difficult, as IBM’s difficulties with OS/2 demonstrate. Instead, the overwhelming majority of start-ups and smaller companies in software play within an architecture defined by a larger competitor. For this reason, they are vulnerable both to the extraordinarily

rapid advances in software that absorb new functionality or surpass it, and to imitation by competitors (Ferguson and Morris, 1993; Torres, 1993). In client-server architectures, for example, mainframe software companies are now quickly bringing client-server applications to market and have excellent access to the installed base (Hoff, 1993). As these markets become increasingly crowded, the cost of entry and capital required to expand these opportunities markedly increases the risk for new companies. This danger is increased by uncertain intellectual property rights (see below).

As participants in the National Academy of Engineering workshop observed, different sectors in software have radically different business models, depending on the complexity of the application, the maturity of the market, and the application platform. And yet despite this complexity, several issues appear to largely determine the success of new companies.

A smaller company cannot wait for market pull but must instead push the technology to create lead time. Consequently, vision for where the market will be is essential and perhaps even more important than technical programming ability (Wiegner, 1992). Of course, pushing the technology in advance of user needs is also risky. The surprisingly unimpressive performance of many software entrepreneurs who try to repeat an early success illustrates this risk and points to the importance of serendipity (Pitta, 1991). Observes Andrew Grove, Chief Executive Officer of Intel: “With all due respect to Microsoft and Intel, there’s no substitute for being in the right place at the right time…That’s not the whole story, but it’s a lot of the story” (Sherman, 1993).

First, it is critically important for new software companies to realize that the opportunity lies with the market or customer, not just the technology. Identifying this convergence between technological opportunity and customer need is essential for success, because the technology is in many instances well in advance of user sophistication. Indeed, a major reason Nintendo and Sega have succeeded so spectacularly, in marked contrast to other Japanese software companies, has been their success in identifying this convergence (Schrage, 1993b). As the penetration of PCs increases, the opportunity in software is proceeding down a curve toward a user who is probably less technically skilled but who has high expectations for functionality and reliability. Understanding the needs of these users is critical if the opportunity created by new technologies is to be exploited.

Second, exploiting this convergence of technological opportunity and customer needs requires speed. Smaller software companies tend to exploit opportunities in niche markets or new markets with low barriers to entry, where an incremental edge in technology or expertise counts. Because of this, the

ability to move quickly is the single largest advantage a smaller company has. As the development progresses, though, there is a tension between the desire to squeeze more functionality into the software and the imperative of launching a new version immediately (Babcock, 1992). Managing this tension between technology and opportunity is a major challenge for new software companies and is critical for success.

It is imperative for new software companies to build competencies quickly beyond the core software development group. Software companies typically start with a core design and development team, and everything else except this group is viewed as peripheral. Software is different in this respect from other design and development processes in that production costs are essentially zero (Zraket, 1992). Consequently, the questions of whom to choose for the management team, and when to choose them, are very important questions for the success of the company. These questions also depend, in part, on the capabilities of the founders. Small companies can be “bound” by their technical founders and need to build their competencies in areas such as finance, marketing and sales, and general management. Building these competencies is usually a prerequisite for significant venture capital financing (Moore, 1993; Pope, 1993).

The economics of opportunities in software are also changing, in response to the increasing importance of prepackaged software. For smaller software companies, achieving large volumes in prepackaged software is increasingly important as markets expand. In core application markets, for example, software companies are using high unit volumes to cut prices. The price of software in these markets increasingly reflects the actual marginal cost of each unit, rather than the perceived value to the user (Schrage, 1993a; Sherman, 1993). As a result, gross profit margins decline.

This emphasis on increasing unit volume has several implications for smaller companies. First, the trend to marginal-cost pricing has forced small companies to build an installed base quickly in newer markets in order to recover their costs. To do this, these firms are forced to seek additional capital for marketing and sales, as well as development, increasing the economic risk of entering new markets. Second, marketing and sales expenses for companies in prepackaged software are increasing dramatically. According to Datamation, only two software companies in the largest 100—Novell and Microsoft—have cost-of-sales that are not the largest item on the balance sheet. For every other company analyzed by Datamation in 1992, selling and marketing costs were the largest single expense (Semich, 1993). On average, these selling and marketing expenses in software are almost 50 percent of revenues (versus 15 percent in PC hardware), giving a cost-per-sale as high as 20 percent in excess of the actual price of the product.

These costs, and the emphasis on high unit volumes, affect the profit model for smaller companies in prepackaged software (Massachusetts Computer Software Council, 1994). Start-ups and smaller

companies launching products in these markets require increasing capital investments. As well, profits now are expected largely from upgrades, maintenance, and service, demanding increased sophistication in these functions.

Alliances are extremely important to start-ups, for several reasons. First, a new software company is often judged by its relationships with larger players. In this context, the name recognition that a strategic partner provides is very useful. Second, tactical agreements can occasionally be used creatively to finance a product launch. For example, small companies that commit to port successful applications to a new computer usually receive press coverage, hardware support, and occasionally advance orders for the new version by the computer maker. Finally, as smaller software companies seek to expand into global markets, they require strategic partnerships with international distributors. Similarly, strategic partnerships with computer makers or larger software companies can be a prerequisite to expanding the market opportunity targeted by smaller software companies to include risk-averse, larger customers reluctant to buy from new companies.

Two observations deserve special attention with respect to innovation in software, and the importance of new companies. First, the development of the software sector in the United States has been shaped by very large federal funding of research and development in computers and software. The United States enjoys a “first-mover” advantage in software because it is very difficult to displace a successful first-mover in software, and because demand for new computer and software technologies developed first in the United States (Mowery, 1996). These first-mover advantages were created not only by commercial activity, but also by federal funding for research and development and the early development of computer science in U.S. universities (Steinmueller, 1996).

Second, developing the software sector has involved the transfer of learning and technology beyond institutional boundaries. Innovation in computers and in software has depended on the opportunity for individuals to move between academia, the federal labs, and technology-intensive companies, for example IBM. In this respect, new, technologically-innovative software companies— and the environment that encourages them—represent an increasingly important mechanism for exploiting intellectual and technological advances in software.

Innovative new software companies tend to be distributed according to a specific geography. That is, several geographic areas that have reached “critical mass” in software tend to be self-perpetuating.

Boston and the San Francisco Bay area, especially, are at the center of extensive networks of universities and research facilities with some of the best students, faculty, and researchers from around the world. Software companies are beneficiaries of the talent that is concentrated in these areas.

This creation of a new academic discipline that is exceedingly “instrument-dependent” has been shaped by large public sector investments. NSF is the second largest funder of computer science research, spending $122.7 million in 1991. Almost all of this spending is for universities— Indeed, NSF is largest funder of individuals in academic research in computer science (as opposed to departments or universities). These funds tend to emphasize basic research (National Research Council, 1992a). That said, NSF-funded research has contributed enormously to software development—the development of the BASIC and PASCAL programming languages were funded by NSF, for example, as well as software engineering and early object-oriented languages, for example, CLU. NSF spending also funds a large computing infrastructure. The most important components of this infrastructure are the four NSF supercomputing centers, and the NSFNET that supports the Internet.

Academic computer science is also funded by the Department of Defense’s Advanced Research Projects Agency (ARPA). Among federal agencies, the DOD continues to be the largest funder of computer science research, spending 418.7 million in 1991; typically, about one-third of this is for academic computer science (National Research Council, 1992a). In this context, ARPA has had an extraordinary influence in defining the research agenda for academic computer science. In contrast to NSF support using individual grants, the objective of ARPA funding has been to develop a basic research infrastructure in computer science using the leading U.S. research universities—Carnegie-Mellon, MIT, Stanford, and UC Berkeley (Mowery and Langlois, 1996). This goal includes funding for education as well as research: In 1990 one-quarter of faculty in the 40 leading US departments of computer science had received their computer science Ph.D. degree from one of the three major universities supported by ARPA—Carnegie-Mellon, MIT, and Stanford (Mowery and Langlois, 1996).

The High Performance Computing and Communications (HPCC) program is currently a large component of this funding for academic computer science. Started in 1992, the HPCC program is coordinated across all federal agencies, as well as large funding from NSF and ARPA. HPCC is defined in the context of specific applications of computing—a series of “grand challenges” in science and engineering, for example, modeling global climate change and weather, that can only be solved using powerful computing. The HPCC program areas are: high performance computing systems; advanced software technology and algorithms; networking; and human resources and basic research. If fully funded over the proposed five years, the HPCC program will represent about $2.0 billion over five years in addition to baseline spending for 1991 (National Research Council, 1992a).

New software concepts developed in the context of these programs are frequently commercialized by small companies spinning out of the university or established companies. Interestingly, participants at

the NAE workshop observed that small software companies will occasionally also “spin-in” to universities (as well as spinning out) to access this thinking. For example, they may send employees back to university to enable continued (relevant) research in a university environment. For these reasons, universities are seen by new companies to be critical for access to academic research, as well as for building intellectual capital.

As well as universities, relationships with other, larger technology companies are very important to small software companies, for several reasons. First, the software industry is marked by a large number of spin-off companies established by entrepreneurs who leave larger software companies (or hardware companies). These new software companies tend to concentrate in areas that include larger, technology-based companies where such spin-offs are common. DEC, for example, has spawned numerous spin-offs, including Data General, which in turn also led to other spin-offs. During a four-year period in the early 1980s, the Boston Globe identified 17 spin-offs involving Data General employees, including Lotus Development (cited in Lampe and Rosegrant, 1992). The spin-offs are less well documented in software but occur with the same frequency (if not more so). Lotus Development, for example, led to at least three spin-offs during the first 3 years after its founding, including Iris Associates (which developed the very successful Notes program using venture capital from Lotus).

As well as providing a source of entrepreneurs, large, technology-based companies also provide a critical base of new technology. New software companies often exploit research and ideas developed elsewhere—usually in universities or large, technology-intensive companies. The laboratories of IBM and AT&T Bell Labs especially, and also Xerox PARC have developed software technologies that have been successfully commercialized by new software companies. Bell Labs, for example, invented the language C without any involvement or direction by the business side of AT&T. C, in turn, was involved in AT&T’s efforts to develop UNIX, also invented by Bell Labs. Researchers at Xerox PARC originally developed the intuitive, icon-based user interface popularized by Apple’s Macintosh (as well as, in some form, networking, laser printers, and page-description languages) (Ferguson and Morris, 1993). All of these innovations were eventually commercialized with spectacular success by new companies that were quick to see the commercial opportunities.

The dynamics of opportunities for new companies in software is very appealing to venture capital. Indeed, software and services attracted more venture capital financing than any other sector of the economy in 1992, including biotechnology. About 200 software and services companies received 22 percent of venture capital invested in 1992, or $562 million. Biotechnology, in comparison, received $261 million of venture capital, invested in 95 companies.

As new software companies demonstrate the viability of new technologies or markets, the risk is less and these opportunities then become valuable to larger companies, creating liquidity by acquisition. Compared to other sectors, the valuations are also relatively high, encouraging the formation of new companies—For example, Microsoft’s bid to acquire Intuit for $1.5 billion represented a breathtaking 40 percent premium over the (then current) market value.

As well as these factors shaping innovation in software, opportunities for small companies are influenced by a set of external factors in which government policies play an important role. And yet, surprisingly, the perception of the importance of start-ups and smaller companies for innovation and technological advance in software in these policies is incomplete.

As well as funding basic research in U.S. universities, many of the initiatives funded by ARPA to develop new technologies have had surprising “spin-on” effects for commercial use. What is surprising is that these research projects, selected to complement the defense community, have had an enormous impact on commercial use. Examples of this research include timesharing, parallel processing, computer-enabled graphics modeling, artificial intelligence, and the Internet.

These surprising effects of the ARPA research initiatives illuminate a related point: research and development are relatively “closer” in computer science than in other disciplines. US universities in this sense have provided important channels for the dissemination and diffusion of these innovations in software between academia, and the defense and civilian research efforts in software. Frequently, the mechanism for exploiting these technology flows is new, technologically-innovative software and computer companies. Digital Equipment Corporation (DEC) is an example of this. DEC’s founder, Ken Olsen, developed many of his ground-breaking ideas for the minicomputer while working as a research assistant at MIT on Project Whirlwind—funded by the DOD—that was the precursor to a massive programming effort to develop the Semi-Automatic Ground Environment (SAGE) air defense system (Lampe and Rosegrant, 1992).

Finding new ways to distribute software could also help new software companies. Now, national distributors, including Ingram Micro and Merisel, exert enormous influence on the economics of software development, paying about 45 percent of list price to developers and distributing the product to resellers at 10–15 percent margins. New software companies that are just establishing a sales channel typically receive

about 35 percent of list price; these margins do not include the cost to the developer of perhaps $15 per package for packaging, manuals and diskettes.

Electronic distribution of software eliminates many of these costs. Smaller companies already use the Internet and electronic bulletin boards to considerable advantage in a number of areas, especially product support and recruiting. As well, LAN-based distribution of software has been successfully exploited by several companies, including McAfee Associates and Symantec. To the extent that the National Infrastructure Initiative (NII) enables new companies to market and distribute software to users electronically, their competitive position will improve. McAfee Associates, for example, derives 90 percent of its revenues from electronic distribution of virus detection software. McAfee Associates’ cost of goods sold is less than 6 percent of sales, and operating margins are 46 percent, compared with an industry average of 18 to 22 percent (Microsoft’s margins are 32 percent).

A related policy issue in this sense is the debate over the use of “strong” encryption technology to encode information and data. The National Security Agency has attempted to prevent the spread of strong commercial encryption technology by controlling licenses to export encryption technology. Currently, U.S. companies may offer encryption software for domestic use but are prohibited from including encryption technology in exports. These efforts to discourage encryption probably also limit the usefulness of electronic communications and potentially the NII.

The federal government is also a prodigious consumer of information technology. DOD programs to develop new, very complex computer systems have had a tremendous influence on the development of U.S. software and computer competencies. Perhaps the most conspicuous example of this is the development of the Semi-Automatic Ground Environment (SAGE) air defense system in the 1950s. SAGE was developed from the Whirlwind project at MIT, to coordinate the control of radar installations into a national air-defense system. Development of SAGE far exceeded previous programming efforts. Indeed, in the 1950s there were virtually no formal programs in computer science (Mowery and Langlois, 1996). By 1963, the SAGE development effort had seeded the emerging software and computer industry with more than 6,000 programmers (Mowery and Langlois, 1996). SAGE’s legacy also includes IBM’s development of transistorized computers. In 1955 IBM delivered the XD-1 to MIT, patterned after Whirlwind—that also inspired Digital Equipment Corporation and the minicomputer—to serve as the “brain” of SAGE. Critical to its performance was a new memory architecture called “magnetic core memory”, that later would appear in IBM’s enormously successful System 360 computer—as well as magnetic tape drives, and flexible software architectures that were all developed under government funding, and adapted almost immediately for commercial use (Ferguson and Morris, 1993).

Department of Defense (DOD) acquisitions continue to play a major role in the software sector— although increasingly it is commercial demand, and not defense-specific requirements, that drive the pace and direction of technological advance in software. According to one estimate, DOD spending for software applications (including development and maintenance) exceeded $30 billion in 1990, or more than 10 percent of all military spending and more than 20 percent of all U.S. spending on software. Numerous consulting firms are involved in developing these applications, although the number of companies that DOD considers qualified for large-scale, mission-critical projects is remarkably small: the Defense Science Board Task Force identified only 24 qualified suppliers in 1987, and International Resource Development identified 50 in 1988 (Zraket, 1992).

Gaining access to substantial capital is never easy for new software companies. The Software Industry 1993 Business Practices Survey collected data on the sources of financing for smaller software companies that raised capital in 1992 (Massachusetts Computer Software Council, 1994). Start-ups and smaller software companies (1 to 15 employees) appear to depend almost exclusively on self-financing, or self-financing with private loans for their initial capital requirements. Beyond this critical phase, and once the viability of the business and its growth potential are demonstrated, capital is available from professional venture capital and corporate partners. Nevertheless, the demand for capital greatly exceeds the supply for high-risk, early-stage companies.

As well as these sources of capital, public (government) financing for projects is available to small software companies through Small Business Innovation and Research (SBIR) grants. This source of funding is generally seen as less useful in software than it is in various computer technologies identified for federal investment (flat panel displays, for example). SBIR funding is also highly competitive and unpredictable. In particular, the protracted delay between the relatively small phase-one financing and the much larger phase-two probably precludes SBIR grants as a dependable source of support. To the extent that SBIR and newer mechanisms—the Technology Re-investment Program (TRP) and the Advanced Technology Program (ATP)—tend to be bureaucratic and slow, they are probably not useful to small software companies that need funding to respond quickly to market opportunities.

Several unresolved legal questions shape these opportunities for new companies in software. Concern to prevent domination by IBM’s extraordinarily successful System/360 architecture led the U.S. Justice Department to assert that this success represented an illegal monopolization. In response, IBM decided to “unbundle” the pricing of its systems instead of including the software in the pricing of the computer, essentially creating the opportunity for independent software companies

to sell competing software. Recently, similar questions have been raised by the Federal Trade Commission and the U.S. Justice Department in connection to the extraordinary successes of Microsoft Corporation in personal computer software.

Uncertain intellectual property rights (IPRs) are a second problem. This perception reflects the unique set of issues that software represents for IPRs. Existing IPRs assume that something is either an expression of ideas (in which case, the expression of these ideas may be protected by copyright law, but not the ideas themselves), or a patentable process (in which case it may be protected by patent law). But software is both an expression of ideas as lines of code, and also the process that the algorithm describes. For this reason, IPRs are an imperfect mechanism (at best) for protecting innovations in software (Barton, 1993; Samuelson, 1993).

As well as this difficulty, participants expressed the view that uncertainty in the area of intellectual property rights may disadvantage start-ups and smaller software companies. Patents, especially, present special problems in software. Many software companies are using patents to compensate for recent legal decisions denying them the copyright protection they see as needed. But patents are costly to obtain and difficult to enforce and defend. Large companies are consequently more likely to be able to threaten litigation and to defend against litigation. There is also ambiguity about what is and is not patentable (Samuelson, 1993). These problems have consequences for innovation, because small companies and start-ups are disadvantaged by the costs and uncertainties of litigation. Also, because larger companies and universities are usually the sources of the technology for spin-offs and smaller companies in software, stronger IPRs for software may actually impede innovation as patent portfolios grow but their value remains ill-defined.⁴

Of course, intellectual property rights are useless unless they are enforced. According to estimates by the Business Software Alliance (BSA), annual losses to U.S. software publishers and distributors are as high as $12 billion globally from illegal copying of software. Enforcing IPR is especially problematic in Asia, where according BSA more than 90 percent of all software is believed to be illegal copies, representing more than $5.4 billion in losses to U.S. companies in 1992. In the United States, perhaps 40 percent of business software is believed to consist of illegal copies (Business Software Alliance, 1993).

Currently, the enforcement of IPRs is directed largely by U.S. companies, the Software Publishers Association (SPA) in the United States, and the Business Software Alliance (BSA). The BSA also recently cracked down on the illegal use of electronic bulletin boards to distribute copyrighted programs. Internationally, the United States pursues violations of intellectual property rights using the U.S. Trade Representative’s Special 301 review of IPR policies to identify “foreign countries that deny adequate and effective protection of intellectual property rights, or deny fair and equitable market access to U.S. persons that rely on intellectual property protection” (U.S. Department of Commerce, 1994, p.27–5). But so far, no country identified under the provisions of this review has been penalized. Enforcing IPRs is also seen to be at odds with recent U.S. foreign policy initiatives, for example, to extend most-favored-nation status to China and implement the General Agreement on Tariffs and Trade.

Small companies will continue to develop new technologies and markets in software. For large companies, nurturing creativity and innovation has often proved difficult, and the risk-reward equation dictating product development is typically very demanding (Hooper, 1993). As a result, successful startups will continue to spin out of larger companies led by entrepreneurs with a riskier agenda.

But creating large software companies (those over $100 million in annual revenues) will be increasingly challenging. In prepackaged software, for example, the trend to marginal-cost pricing has forced small companies to build an installed base quickly in newer markets in order to recover their costs. Capital requirements and management expertise required to expand these opportunities faster than the competition and sustain a market position are already a constraint for small companies, encouraging mergers to achieve critical mass. This trend can be expected to increase significantly in the next few years as markets become increasingly crowded.

A second trend in prepackaged software is continued competition for control over various technologies that comprise the computer architecture. As a consequence, new software companies are more often being required to provide support for multiple software and hardware technologies (Massachusetts Computer Software Council, 1994). This requirement for multiplatform support, in turn, requires small software companies to form a variety of strategic relationships with hardware and software vendors that control various standards and elements of the computer architecture. The ability of certain new software companies to manage these relationships successfully is potentially a source of competitive advantage, but it is also an expense and burden for small companies.

As software programs (including prepackaged software) have become larger and more complex, software developers have started to run into the problems of quality and reliability previously faced only by systems integrators. This trend is referred to in the literature as the “software bottleneck” (Ferne and Quintas, 1991). Major delays in product releases, including 1–2–3 (Lotus), dBase IV (Ashton-Tate), OS/2 (IBM), and Windows NT (Microsoft) are examples of this trend (Brandt, 1991). Small companies may not

be sophisticated enough (or liquid enough) to compete and expand as the complexity of these systems increases.

Attempts to address these quality and productivity problems replace the current approach to software development with a more rigorous, engineering-based approach (Ferne and Quintas, 1991). Japan and Europe, especially, have put a premium on developing process innovations in software design and automation (U.S. Department of Commerce, 1994). These innovations (if real) may present an opportunity to compete with currently dominant U.S. software companies, especially in code re-use and object-oriented designs (Brandt, 1991; Ferguson and Morris, 1993). While new technologies present opportunities for new entrants, established smaller companies may not have the resources to adopt these innovations and remain competitive.

If correct, these trends represent a dilemma. Previously, small companies and start-ups have developed most of the new technologies and markets in software, and have competed successfully with larger players. In the future, smaller software companies may need to focus on opportunities that minimize capital investments. Small software companies will depend increasingly on relationships with larger companies, for example to act as publishers and distributors for the software developed by smaller companies. This trend is appearing already: Larger competitors, for example IBM and Microsoft, increasingly see new companies less as threats and more as opportunities to identify promising new technologies and markets.

Babcock,C. 1992. Software tar pits. Computerworld. Dec. 28th. P.75.

Barton, J. 1993. Adapting the Intellectual Property System to New Technologies. Pp. 256–283 in Global Dimensions of Intellectual Property Rights in Science and Technology, M.B.Wallerstein, M.E.Mogee, and R.A.Schoen, eds. Washington, D.C.: National Academy Press.

Brandt, R. Can the U.S. stay ahead in software? Business Week. March 11th. P.98.

Business Software Alliance. 1993. The U.S. software industry: economic contribution in the United States and world markets. Prepared for the Business Software Alliance by S.Siwek and H.Furchtgott-Roth, Economists Inc.

Catalanotto, K. 1994. Computer software: Federal policy for critical technology. IEEE Technology and Society Magazine. Winter. Pp.7–12.

Economist. April 3rd, 1994. Technology in Utah: software valley. P.69.

Ferguson, C. and C.Morris. 1993. Computer Wars: The Fall of IBM and the Future of Global Technology. New York: Times Books.

Ferne, G. and P.Quintas. 1991. Software Engineering: The Policy Challenge. Paris: Organization for Economic Cooperation and Development.

Friedland, J. 1993. Japan’s soft spot. Far Eastern Economic Review. August 5th. P.54.

Hodges, J. and D.Melewski. 1993. Top 100. Software Magazine. July. P.75.

Hoff, R. 1993. Investors hug software’s new babies. Business Week. March 15th. P.98.

Hooper, L. 1993. The creative edge: nurturing high-tech talent requires a delicate balancing act; but the payoff can be huge. Wall Street Journal. May 24th. R6.

Lampe, D. and S.Rosegrant. 1992. Route 128: Lessons from Boston’s High-Tech Community. New York: Basic Books.

Massachusetts Computer Software Council. 1994. Software Industry 1993 Business Practices Survey: A Current View of the Industry. Boston: Massachusetts Computer Software Council.

Moore, J. 1993. Manager’s journal: the tough interface between execs and techs. Wall Street Journal. August 16th. A14.

Mowery, D. 1996. Introduction. Pp.3–14 in The International Computer Software Industry: A Comparative Study of Industrial Evolution and Structure, D.Mowery, ed. New York: Oxford University Press.

Mowery, D. and R.Langlois. 1996. The federal government role in the development of the U.S. software industry. Pp.53–85 in The International Computer Software Industry: A Comparative Study of Industrial Evolution and Structure, D.Mowery, ed. New York: Oxford University Press.

National Academy of Engineering. 1995. Revolution in the U.S. Information Infrastructure. Washington, D.C.: National Academy Press.

National Research Council. 1990. Keeping the U.S. Computer Industry Competitive: Defining the Agenda: A Colloquium Report. Washington, D.C.: National Academy Press.

National Research Council. 1992a. Computing the Future: A Broader Agenda for Computer Science and Engineering. Washington, D.C.: National Academy Press.

National Research Council. 1992b. Keeping the U.S. Computer Industry Competitive: Systems Integration: A Colloquium Report. Washington, D.C.: National Academy Press.

Office of Science and Technology Policy. 1994. High Performance Computing and Communications: Towards a National Infrastructure. Washington, D.C.: Office of Science and Technology Policy.

Office of Technology Assessment. 1993. Background Paper: Advanced Network Technology. Washington, D.C.: Office of Technology Assessment.

Organization for Economic Co-operation and Development. 1985. Software: An Emerging Industry. Paris: Organization for Economic Co-operation and Development.

Pitta, J. 1991. Second time too rich. Forbes. August 5th. P.94.

Pope, K. 1993. Avoiding risk: venture capitalists aren’t looking for geniuses these days; they’re looking for results. Wall Street Journal. May 5th. R11.

Samuelson, P. 1993. A case study in computer programs. Pp.284–318 in Global Dimensions of Intellectual Property Rights in Science and Technology, M.B.Wallerstein, M.E.Mogee, and R.A.Schoen, eds. Washington, D.C.: National Academy Press.

Schrage, M. 1993a. Good news for software consumers: the future is going to be cheap. The Washington Post September 24th. B3.

___. 1993b. Japan’s software industry needs customers who can fuel innovation. The Washington Post. May 14th. D3.

Selz, M. 1993. Small software companies crack the educational market. Wall Street Journal. March 10th. B1.

Semich, J. 1993. Software’s big 100. Datamation. September 15th. P.38.

Sherman, S. 1993. The new computer revolution. Fortune. June 14th. P.56.

Steinmueller, W.E. 1996. The U.S. software industry: An analysis and interpretive history. Pp. 15–52 in The International Computer Software Industry: A Comparative Study of Industrial Evolution and Structure, D. Mowery, ed. New York: Oxford University Press.

Torres, C. 1993. Heard on the street: Red flags fly for one-product tech companies. Wall Street Journal. February 5th. C1.

U.S. Department of Commerce. 1982. Census of Service Industries. Washington, D.C.: U.S. Department of Commerce.

U.S. Department of Commerce. 1987. Census of Service Industries. Washington, D.C.: U.S. Department of Commerce.

U.S. Department of Commerce. 1994. U.S. Industrial Outlook. Washington, D.C.: U.S. Department of Commerce.

Wiegner, K. 1992. Nerds, not Forbes. October 19th. P.272.

Young, J. 1993. From Star Trek to desktop. Forbes. July 19th. P.204.

Zraket, C. 1992. Software: productivity puzzles, policy challenges. Pp.283–313 in Beyond Spinoff: Military and Commercial Technologies in a Changing World, J.Alic, L.Branscomb, H.Brooks, A.Carter, and G.Epstein, eds. Boston: Harvard Business School Press.