(FORTUNE Magazine) – In this age of outsourcing, many companies don't actually produce the goods they are known for. Nike doesn't make its sneakers. Snapple doesn't squeeze its juice. Schwinn bikes aren't assembled by Schwinn. Now, half a century after the invention of the transistor, here come semiconductor companies that don't produce semiconductors or even sell them. Instead, they design circuits for others to manufacture and market. Call them chipless chip companies or, as they would prefer, intellectual-property semiconductor companies--this latest wave of electronics startups may herald a significant new segment of the industry.

Five such companies already are public--ARM Holdings, of Cambridge, England; Artisan Components of Sunnyvale, Calif.; Aspec Technology, also of Sunnyvale; MIPS Technologies of Mountain View, Calif.; and Rambus, also of Mountain View. In the past year, venture capitalists have invested a total of $90 million in 14 more, and there may be as many as 100 firms in the business.

These intellectual-property (IP) semiconductor companies are part of a decades-long specialization that's been taking place in the world of microelectronics. Think of it as industrial dis-integration. Early on, the typical semiconductor company had to do everything itself: design and manufacture its products and build the tools used in manufacturing. Then, in the 1960s, independent tool companies arose. In the 1970s, companies were formed to automate the process of semiconductor design. The 1980s brought so-called fabless semiconductor companies, which design and sell integrated circuits actually fabricated by others. Today all three are multibillion-dollar industry segments. IP companies are simply the logical next step.

Their rise is due in part to Intel founder Gordon Moore's now famous law that the number of transistors on a chip doubles every 18 months. Today's chips have tens of millions of transistors--so many that few companies other than memory-chip makers can effectively use all that power. Also thanks to Moore's Law, a single circuit can take the place of electronic devices that not long ago required separate circuits for such functions as the central processor, communications, graphics, and memory. As a result, electronics manufacturers are developing so-called systems on a chip that marry designs from several suppliers on a single piece of silicon. The older approach of building a digital system, such as a CD player or hand-held computer, required several chips. For example, Nintendo's very successful Nintendo 64 videogame player combines circuit designs from three companies--Artisan, MIPS, and Rambus--into a single integrated circuit. The design was key to achieving child-pleasing performance at a mass-market price.

The system-on-a-chip trend could alter the balance of power among the giant microprocessor and memory manufacturers. Asian companies that dominate the low-margin computer-memory business may be able to capture a larger share of worldwide electronics profits by combining memory with high-margin microprocessors in a single component. The rise of semiconductor IP companies even poses a threat to Intel. As digital technology moves beyond the PC to the consumer-electronics business, compatibility with Intel's microprocessors becomes almost unimportant. To an engineer developing, say, a digital camera or an intelligent dishwasher, compatibility with Intel doesn't matter--all that does matter are the price and performance of the integrated circuit he employs.

Most of the semiconductor IP companies design circuits--from microprocessors and graphics controllers to digital signal processors and networking processors--for others to manufacture and sell under their own labels or to blend with other designs into a combination chip. For example, NEC's new microprocessors are based on plans licensed from SandCraft of Santa Clara, Calif. Broadcom plans to incorporate microprocessor designs from MIPS in its one-chip systems for television set-top boxes, cable modems, and satellite communications. ARM designs are at the heart of forthcoming Texas Instruments microprocessors for the automobile and consumer markets.

Other companies such as Chrysalis Symbolic Design of North Billerica, Mass., Duet Technologies of San Jose, and Verisity Design of Mountain View sell software tools for creating or making use of IP semiconductor designs.

Yet another category of IP companies is focusing on proprietary ways of connecting the diverse elements of an integrated circuit or computer system. That's the business of such companies as Palmchip of San Jose and Rambus.

The future of chipless chip companies looks promising. Four of the five public companies already are profitable--ARM, Artisan, MIPS, and Rambus. Small IP companies are attracting large partners. Recently, for example, Lucent Technologies, National Semiconductor, NEC, and Qualcomm started working with ARM. In addition, some IP companies are starting to get the bulk of their revenue from royalties, as intended. Five years ago, 80% of MIPS's revenue came from payments for helping other companies incorporate its designs; today almost all of its revenue is from royalties.

Still, this is a nascent industry. Artisan and Aspec have almost no royalty revenue, and Rambus seems to be losing ground. In its most recent quarter, royalties accounted for only 18% of revenue, down from 26% in the same quarter a year earlier.

Moreover, royalties for semiconductor intellectual property are small. Only big markets, such as consumer electronics, can provide enough royalty revenue to sustain the chipless companies. Large makers of electronic products are understandably reluctant to depend on young, unproven suppliers for essential technologies. Furthermore, creating and testing single-chip systems with intellectual property from several companies is difficult; the individual designs may operate at different voltage levels or frequencies, or may be intended for use with incompatible semiconductor manufacturing processes or materials. Worse yet, big-name chip companies are becoming IP competitors. Recently, for example, Lucent and Motorola agreed to share circuit designs, as did IBM and STMicroelectronics.

The young IP companies, however, are working to overcome these obstacles. They've formed a consortium called the Virtual Socket Interface Alliance to create mix-and-match standards. There's a trade group for IP companies (it's called Rapid), and even a magazine, Silicon Strategies.

More encouraging for investors, I think, is the example of an earlier generation of semiconductor upstarts, the fabless companies. For this column, I looked over several years of financial history and found that a total of $3.6 billion had been invested to create 30 public U.S. fabless semiconductor companies. Those companies, however, have generated a mere $1.6 billion in retained earnings. In other words, only 44 cents has been earned for every dollar invested in a fabless enterprise. By comparison, every dollar invested in a public semiconductor company with its own fabrication facilities has returned $2.40 in earnings.

Yet the fabless companies have been good to early investors. The $3.6 billion they put up has resulted in a market capitalization today of $14.7 billion, or a gain of a 308%. My guess is that the IP companies now aborning have at least as much potential for wealth creation.

RICHARD A. SHAFFER is founder of Technologic Partners, an information company focused on emerging technology. Except as noted, Shaffer has no financial interest in the companies mentioned. For an expanded version of Watch This Space, visit www.tpsite.com/tp/fortune/. If you have comments, please send them to shaffer@technologicp.com.