Calling into question the long-term viability of the IP industry, a number of core developers are beginning to seek a more immediate revenue payoff by going the standard-product route.
Despite the efforts of industry groups like the Virtual Component Exchange and Rapid to standardize licensing models and best business practices, companies are still struggling to make money in the IP market. At the same time, inadequate design-in support has soured many OEMs on third-party IP, sources said.
As a result, several privately held IP companies are now taking on the role of a fabless chip supplier to maximize their income, while continuing to license low-grade cores to customers that require little in the way of design support.
This week, for example, CoreEl MicroSystems Inc. will unveil an application-specific- standard-product (ASSP) strategy it hopes will account for more than half of its revenue within two years.
EnThink Inc. is adopting a similar initiative, although its first chips won't be released until early next year.
To many in the business, such moves are simply evolutionary. "One of the early misconceptions in the IP business was that anyone who could develop a piece of IP could be an IP company," said Robert Nalesnik, vice president of Phoenix Technologies Inc. "I've seen several companies get into the market, then quickly shift their focus when they realized it was harder than they thought."
Much as the fabless-semiconductor model has shifted back toward fab ownership to varying degrees, the IP industry is going through some soul-searching, according to Robert Chaplinsky, a partner at Mohr, Davidow Ventures in Menlo Park, Calif., which funds several IP start-ups.
"IP has definitely lost a lot of its glamour," Chaplinsky said. "Two or three years ago, it was the model of the day. But system-on-a-chip has been this Holy Grail promise that still hasn't been delivered. I think that's impacting the excitement level over IP."
Industry executives said they have felt the chill from the financial community. IP developers are being told to produce a positive cash flow quickly or look elsewhere for funding, sources said. But for many companies, revenue is tied up in royalties on designs that may never see production.
While everyone wants to be the next Rambus-which is addressing a $35 billion market with a single interface-the fact is that few proprietary technologies will find as lucrative a niche. Most proprietary cores have a very limited application base, Chaplinsky said.
"If the market you're addressing isn't big enough to sustain a meaningful revenue stream, you may as well go fabless," he said. "Why give away your IP for 50 cents when you can sell the same thing as a chip for $50?"
Often, survival is a matter of size, according to Nalesnik of San Jose-based Phoenix, which has become a powerful player in the industry by acquiring other companies with related technologies.
"There's a financial barrier to small IP companies that essentially forces them to get to a run rate-which I put in the range of $10 million-where they can support a reasonable-size direct-sales channel," Nalesnik said. "But if they put it in a chip, now they can run that through an indirect- sales channel."
Yet even vendors that have managed to generate a reasonably steady revenue stream are finding that customer support requirements are too intensive for small engineering firms to handle, said Sudip Nandy, vice president of marketing at Santa Clara, Calif.-based EnThink.
"It turns out we're spending as much as 40% of our effort in creating the first IC for a customer because everybody uses a different on-chip system bus, different tools, different foundries," Nandy said. "We came to the realization that with the same amount of effort, we could do a standard part for the same customer-or several customers at one time-while reserving some portion of the design for customization."
EnThink, which develops bus-interface IP for Internet appliances, is readying an ASSP launch for early next year, the details of which are being closely guarded for now. The company is not abandoning the IP business; rather it is expanding its offering to address the continuum of customer requirements between custom and standard designs, according to Nandy.
CoreEl MicroSystems is also revising its IP strategy. Beginning today, the networking-core developer will go to market with three separate business models, each aimed at reducing customer-support requirements.
"It's a question of how do I get a return on my investment," said Yu Hao Lin, president of CoreEl, based in Fremont, Calif. "For every dollar I put into engineering, I want to get $20 back within five years. If I'm spending so much of my resources supporting a customer's design, there's no way I'm going to get that 20 to 1 return."
Under one model, CoreEl will license its commodity-oriented bus-interface cores to the general population, a model that currently represents 100% of its business.
A second model will target system companies only, offering higher- value IP for a minimum upfront, nonrecurring engineering fee. In some cases, CoreEl will support the design through to production to ensure its royalty stream, Lin said.
Under the third model, CoreEl will keep its high-end, "strategic" IP to offer as application-specific standard products. In this way, the company will avoid risk while gaining further insight into the future IP needs of system companies. According to Lin, ASSPs will represent more than half of CoreEl's revenue within two years.
In a related move, CoreEl has sold its Bangalore, India-based semiconductor- distribution business to the Memec group, a VEBA Electronics company. The operation is a major distributor in the region for Xilinx and PMC-Sierra.