SIP Market to Grow in Every Direction

By Gale Morrison -- 1/1/2001

Research Triangle Park, N.C.-Physical libraries will have another foundation-building year in 2001, but for the most part the semiconductor intellectual property (SIP) market should fare very well this year.

SIP's good fortune this year will be the direct result of the overall long-term trend to outsourcing in the worldwide semiconductor and electronics industry. The year 2001 should be an exemplary year to find evidence that the top 20 global semiconductor companies are no longer hesitant to pay third parties for key design engineering.

Five years ago, the top companies were pouring hundreds of millions of dollars into engineering their own DRAM. Now, some have acquiesced to the licensing demands of Rambus Technologies Inc., Mountain View, Calif. Even as recently as two years ago, the top microprocessor companies held their architectures as sacred. Now, they have mainstream plans with their architecture licenses from MIPS Technologies Inc., also of Mountain View, and ARM Ltd., Cambridge, England, to name two. The same examples are found outside SIP: The top money makers invested hundreds of engineers and millions of dollars in packaging, now they go to West Chester, Pa.-based Amkor Inc.; they spent billions on new fabs, now they go to the Southeast Asian foundries.

The year 2001 will be the year that ARM, Rambus and MIPS think in terms of hundreds of millions of dollars in revenue and the year that many more SIP companies reach long-term viability. That viability might derive from the quality of their investments in SIP start-ups and not from longevity as stand-alone companies, as the inevitable acquisitions and mergers occur, but it is viability nonetheless.

Jim Tully, chief analyst in the semiconductor group at Dataquest U.K., says the revenue forecast looks good, but he does caution that vendors do not have enough of their customer base paying royalties.

"The IP market is moving ahead very strongly. We are forecasting a compound annual growth rate to 2004 of 46 percent," Tully said. "But right now, the bulk of this is license revenue with royalty taking slightly more than 20 percent. The number of design starts worldwide is falling as companies standardize on a platform that is common across several products," he warned.

"Those vendors charging a license fee per design will either need to sharply increase their fees or move to a royalty model," he said. Tully thinks SIP vendors are up to the task, though. "We are likely to see royalty revenues growing significantly over the next few years."

The number of transactions involving the licensing of third-party IP will more than double in 2001 compared to 2000, said Andy Travers, president and chief executive officer of the VCX IP exchange based in Edinburgh, Scotland. He expects 2,000 transactions in 2000, growing to 4,200 by the end of this year.

Max Baron, principal analyst at Cahners In-Stat Group, Scottsdale, Ariz., also expects solid growth for SIP revenues in 2001, but he cautions that the field of competitors will see a marked change.

In a recent report, Baron said that during 1999 a new type of IP competitor emerged: large, well-established semiconductor and system companies ready to license at least some of their SIP. And in fact, by the end of 2000, one could see that Philips Semiconductors with its Nexperia platform, Texas Instruments Inc. with its OMAP, Motorola Inc. with new ARM-based offerings, and even firms like Tality Corp. with its Bluetooth and wireless experience had big plans for the SIP game. Some of their offerings include third-party SIP, but it's surely their intention to earn the engineering services dollars that come with SIP.

Baron sees a few other factors affecting the IP industry. SIP trading and more reuse of designs will reduce the need for multiple suppliers of the same type of IP, he said. The year 2000 already saw the fallout from such a scenario as Mentor Graphics, Beaverton, Ore., left the library business at the beginning of the year, and then Synopsys Inc., Mountain View, also left at the end of the year by selling its library unit to Artisan Components Inc., Sunnyvale, Calif.

The physical library business, where Artisan is considered a leader, is looking at a forecast of sun and clouds in 2001, executives say. Both Artisan and competitor Virage Logic Inc., Fremont, Calif., said that their fiscal year 2001 ending in September would have them bringing in around $30 million in revenue and earning about $3 million. Virtual Silicon Technology Inc., Sunnyvale, is not yet a publicly traded company so it didn't give guidance to Wall Street last fall. But all three companies are competing with the same type of product for the same customers, which means excess capacity. This brought on Synopsys' exit from the business that as recently as February of 2000 it had planned to own.

"We are heading into a period of excess capacity. I expect the resultant price pressures to strain the business models of the library vendors," said Steve Carlson, CEO of Tharas Systems Inc., Santa Clara, Calif. Over the course of 2000, library vendors got closer to the Southeast Asian foundries and established partnerships in which foundry customers get libraries for their designs and the foundry itself pays the library vendor when a given customer has successfully come through the fab.

Executives said that model would continue to work because the foundries are growing so well, but on the other hand, it's problematic because the royalty payment can be delayed or not come at all if a customer does not get into production for other reasons.

"As far as the foundry-sponsored royalty model is concerned," said Sanjay Srivastava, CEO of memory compiler firm Denali Inc., Palo Alto, Calif., "companies like Artisan, who pioneered it, have still not been rewarded either by the equity market or in the balance sheet. So the jury is out."

"I believe it is a flawed strategy if only because the customers do not pay anything if their design does not hit production even though they utilized the IP and the services," he said. "Also, it would be impossible to build long-term customer loyalty and maintain differentiation from one process to the next," he added. Baron at In-Stat also is concerned that these "foundry giveaway" programs may expand into the SIP business as a whole, and that could spell bigger trouble.

The Software Base

The lion's share of the SIP market is in control, meaning either microprocessor or DSP cores. Mark Bowles, president and chief operating officer of BOPS Inc., Mountain View, Calif., says that is sure to remain true in 2001.

"Programmable IP (as CPUs and DSPs) is fairing better than other types of IP. The implication is that programmable is best suited of all for the IP model because the value of the IP is perpetually regenerated by the software base," Bowles said. "Because of this, vendors - ARM, MIPS, DSP Group, and soon ARC - are able to maintain a de facto standard foothold in a way that InSilicon, Rambus and others aren't," Bowles said.

In discussing his firm's broad licensing deal with Motorola in December, ARM's Robin Saxby said he could not overstate the importance of software compatibility for a microprocessor architecture, knowing that has been ARM's success.

"I worked at Motorola in the 1970s and 1980s, and at the time Fairchild came out with their own version of the (Motorola) 10K processor. But it was only 98 percent software-compatible, and for that reason it failed," Saxby said. Vendors across the semiconductor board learn-some say over and over again-that software engineering, if navigated poorly, can cripple a project.

Of Course, Communications

To those who have been living in a cave for the last 18 months, it might come as a surprise that SIP vendors with expertise to offer customers in the communications space are going to have a good year in 2001. Expertise in communications often means DSP expertise and SIP vendors have been coming to market with new DSP SIP faster than you can say Internet.

"For the IP market, year 2001 will be the year of the DSP," said Bowles of BOPS, albeit with more than a little self-interest. "There are now three public RISC CPU companies (maybe four soon with Tensilica) and only one public DSP IP company (DSP Group). There are lots of up-and-comer DSP players. BOPS, 3DSP, Massana, Improv, etc., have received the bulk of venture-capital funding for IP companies in the past 12 months, and the markets - both the customers and the equity markets - are screaming for them to come on."

All of these players are offering new DSP configurations and are devoting lots of their time to the systems know-how needed around DSP arrays to accomplish packet processing and other next-generation communications tasks.

MIPS, which is an elder statesman of SIP compared to these other fields, is heading into 2001 with a very positive outlook, said Casey Eichler, the company's chief financial officer. MIPS is the architecture inside the Playstation2, the Net-access-capable gaming box that consumers literally fought over for Christmas 2000. And MIPS' architecture licensees include next-gen communications first-stringers such as Broadcom Corp., PMC-Sierra Inc. and those hot smaller firms the two bought up last year including SiByte and Quantum Effect Devices, Eichler said.

© 2009, Reed Business Information, a division of Reed Elsevier Inc. All Rights Reserved.