Richard Goering - May 02, 2007
Panelists, drawn mostly from IP vendors, generally agreed that royalties represent the best revenue model for silicon IP providers. But they clashed over such issues as IP patent protection and the viability of pure-play IP providers.
The one user representative on the panel was Ken Tallo, managing director of the mobility group at Intel Corp. He opened the panel discussion by citing a recent EE Times article, " IP providers cast wary eye on TSMC."
"As I read the article, I had one thought I really like Home Depot," Tallo said. Home Depot isn't just lumber and nails, he noted; customers can buy pre-made cabinets, snap-in-place flooring, and entire bathrooms. And that's the kind of service Intel wants to see with third-party IP, Tallo said.
"We want everything complete and ready to go. We need a complete solution, but now we just can't get it," Tallo said. The IP industry, he said, needs to continue to increase the value proposition of third-party silicon IP. "You're doing it, but there are still gaps," he said.
Bill Martin, general manager of the IP division at Mentor Graphics, noted his company's recent release of a complete USB subsystem that includes a digital controller, a physical-layer component and middleware. It's a complete solution that leveraged both Mentor's embedded software and IP development groups, he said.
"It's raising up the value proposition, and we need to keep doing that," Martin said. "Right now there's too much IP that doesn't work together." The biggest problem Mentor sees, according to Martin, is incompatibility between software drivers and the core IP.
The IP issue is "very simple," said venture capitalist Lucio Lanza, managing director of Lanza TechVentures. "There is IP that is needed and IP that is useless." If you produce IP that's easy to copy and make, Lanza said, you won't make money. The kind of IP that has value, he said, is hard-to-develop IP that's on a "treadmill" that requires continuous improvement.
"Most silicon IP has a very limited value," Lanza said. Most of it is based on RTL code, he said, and it might save a few weeks or reduce the design cycle by a few percent. "Most of these companies have non-viable business. I don't want to invest in them." Lanza was chairman of Artisan Components before its 2004 purchase by ARM.
Mike Kaskowitz, senior vice president for semiconductor IP at Mosaid Technologies, said there are five IP business models. They include embedded software providers, verification software IP providers, "failed semiconductor companies," design consultants, and true third-party IP companies. For the latter, he said, "the best thing is to charge royalties. The industry doesn't like royalties, but that's how to make a go of it long term unless you're really large."
Brian Gardner, vice president of the IP products group at Denali Software, noted that the total available market (TAM) for most types of IP is in the $50 to $100 million range and that there may be 10 or so competitors. "At the end of the day, the only way you can succeed is to out-integrate the other guy," he said.
Denali mostly builds memory controllers, and one challenge there is that "it's hard to have one size that fits all," he said. Thus, the IP needs to be highly configurable. His prescription for IP success: "figure out the balance between how to build things that are configurable enough for 80 to 90 percent of the market, and still be a product."
ARM Ltd. wants to create the "Home Depot" of silicon IP, said Brent Dichter, general manager for physical IP at that company. As system on chip (SoC) complexity increases, he said, customers need IP for CPUs, busses, peripherals, software, and libraries. What customers want, he said, is "complete platforms where they can build their devices."
According to Jack Browne, vice president for marketing at MIPS Technologies, there are three criteria for IP success. He said IP vendors need a large market opportunity, characterization by standards, and enough difficulty to prevent being "commoditized."
Browne observed that SoC success rates haven't changed in 20 years. Half of chip starts don't finish, and half of those who finish don't reach volume production, he said. IP and EDA vendors, and customers, must realize that "perfect is the enemy of good." The key for IP providers, he said, is to "give customers a good enough solution to solve their problem, in as aggressive a time frame as we can."
Debating IP protection
In response to a question about "patent trolling," Mosaid's Kaskowitz noted that Mosaid derives much of its revenues from IP patent litigation, and that his semiconductor IP division is up for sale. Mosaid indicated it might sell its IP product business during a management shakeup in April.
"People who are using IP to enforce patents to make money are damaging the industry," Lanza said. "They should not be profitable. I hope they all disappear." Lanza said that IP patent protection has "gone too far" and that the time frames should be closer to the lifecycles of products.
Sparks flew when Mentor's Martin said that "pure play IP people have a problem. They'll have to expand their solutions, and they're competing with other people who have a wider breadth of a solution."
"I disagree completely. I think pure play IP has the best opportunity," Lanza said. "There is far more demand for pure-play IP, especially when the main processor is becoming overloaded." EDA companies selling IP, he said, are in the "wrong business model" and will end up using IP to sell tools.
ARM's Dichter said pure-play IP providers have a great advantage in their independence. Pure-play IP, he said, can work with any vendor's tool and any foundry. Mentor's Martin countered that ARM itself sells development tools.
"As a customer, I'm happy to find it wherever it comes from, so long as it works," said Intel's Tallo.