INDIAN WELLS, Calif. The Intellectual Property (IP) block business is in a state of "chaos" now, but the design technique will grow and evolve over time, giving programmable logic a run for its money. That was one conclusion of analysts here at the annual Dataquest Inc. semiconductors conference, where IP emerged as a major issue in the semiconductor outlook in the next 10 years.
Analysts in separate presentations agreed that low-end "commodity" IP, such as PCI cores and the like, will remain a brutal business with few winners. High-end, differentiated IP will succeed on its own (or enable its provider to be acquired), while a new class of IP downloaded by consumers will emerge in the next decade, analysts said.
Amid several bold predictions, analysts Jordan Selburn argued that FPGAs will never become a major IP delivery approach, which will cause the PLD business to decline at that inflection point.
"There's always going to be a pla ce for PLDs, but by 2005, they will start a never-ending revenue decline," he said.
Both Selburn and fellow Dataquest analyst Jim Tully said that standards, verification and legal issues will all be resolved in the next few years and that the main issues to be addressed will be which IP is best for which solution.
"VSIA, VCX, Rapid will have, for the most part, done their jobs," Selburn said. "System on a chip is the reality today, but if it is to become pervasive, it (IP) needs to be standardized."
Despite the importance of the cores approach to increasingly complicated SOC design, there won't be "an appreciably greater number" of successful IP companies in 10 years, Selburn said. The business, which burst onto the scene several years ago, has seen prices plummet as more vendors offer similar cores on the market for sale. The top five semiconductor vendors will delivery 75 percent of the IP in 2010, and 95 percent of the average SOC design will consist of IP and embedded memory, he predicted. < P> Selburn was quick to highlight a demarcation between value-added IP and "cost-plus" IP, where differentiators are few and most OEMs can make the IP easily, rather than have to purchase it.
"Most IP does not differentiate the end user," he said. "It limits value, price and margin."
Valuable IP, however, will command a price in the mergers-and-acquisitions market, where a large company stands to gain from not just licensing the IP but owning it exclusively, he added.
Tully took a slightly different view, arguing that if today's embedded functions are fixed for the lifetime of the product, tomorrow's will be configurable and updatable thanks to the Internet. Tully argued that end users will purchase and download blocks of IP that will change the functionality of a given device-in much the same way a consumer can change options from his cable provider. "It's good for manufacturers because it opens up revenue streams. It's good for consumers," he said.
Tully also noted that in the next 10 yea rs the percentage of IP blocks purchased from ASIC vendors will decline from 60 percent to nearly nothing, while the percentage from OEMs will rise dramatically in the next few years before falling off around 2007. Third-party IP vendors will fluctuate in their popularity until taking off around 2006 with consumer demand, he added.
"No one values the current IP," he said. "It'll only be recognized when a direct channel to the consumers is realized. We're talking about millions of customers per IP block." He added that he expects such revenue to be $3 to $10 per customer.