Hi Doug, thanks for your response to my April 2 post...please give me your comments/opinions on this April 6 news article; actually I would appreciate the opinions from anybody on this Thread who wants to comment on it:
Hey Jeff Jordan, I especially want your comments! <VBG>
April 06, 1998, TechWeb News
Panel examines all aspects of business models built on cores -- Wall Street skeptical as IP vendors go public
By Craig Matsumoto
Santa Clara, Calif. - Successful intellectual-property (IP) companies may be proving the skeptics wrong, but the financial community still wants to know how these businesses plan to survive and how the IP industry will operate in the long run. With Rambus Inc. and Artisan Components Inc. having already gone public-and Advanced RISC Machines Ltd. and Aspec Technologies Inc. prepared to follow suit-customers and investors alike are pressing hard for answers.
Executives took a shot at addressing those issues at the recent IP98 conference here, in a panel discussion titled "Closing the Design Gap with Semiconductor Intellectual Property."
The chairman of the discussion, financial analyst Rob Chaplinsky of Hambrecht & Quist LLC (San Francisco), outlined Wall Street's skepticism about IP businesses. For starters, he said, investors think IP customers are unstable ("Otherwise, why would they need your IP?" is the complaint Chaplinsky reported). Others fret that royalties are too emotional an issue, and that IP fees will be too high to suit chip makers.
That last concern was called invalid by Bill Davidow, a general partner in venture-capital firm Mohr Davidow Ventures (Menlo Park, Calif.). Companies should look at what they'll save in R&D and realize that-in most cases-IP royalties will be far less, he said.
"They so frequently focus on what they pay for the technology rather than the benefits they get," said Davidow. As an example, he cited Rambus's model of asking for a royalty on every product using its IP. "I believe Rambus has saved and will save the semiconductor industry billions of dollars."
Davidow contended that Rambus (Mountain View, Calif.) has become a joint R&D lab for the industry. The company has spread its development costs among several partners, including Intel Corp. Davidow guessed that it's taken $100 million in investment to build Rambus so far, while Rambus itself has made very little money.
"Partners individually have each paid a fraction of what it would cost to develop that technology themselves," he said. For example, Davidow guessed that the sheer reduction in die size permitted by Rambus know-how is enough to offset the royalties paid.
The new IP industry also permits easier creation of standards, Davidow contended. Company-created standards cost R&D money and usually are decided after a standards war between competing ideas, all of which incurs costs that aren't recorded on the balance sheet. "I'm sure if we could capture all of those costs, we would all be supporters of Rambus," Davidow said.
A cloudy future
Still, questions linger about what a sustainable IP industry would look like, or whether IP business models will have to change (see March 30, page 1)
"We are at the beginning of a new industry," said Mark Templeton, chief executive of Artisan Components (Sunnyvale, Calif). "We are at a time when there's going to be a lot of experimentation," both in royalty models and IP delivery methods, he said.
Demonstrating that uncertainty, panelists were divided on the eventual fate of IP. Templeton predicted a wide scattering of IP vendors-"There's room for a good many more than there are," he said-while Davidow believed the industry will consolidate.
"You're going to see a few dominant suppliers of collections of intellectual property, which will make it possible for companies to come to them and buy a whole portfolio. That's the ultimate endgame," Davidow said.
Neither Artisan nor Rambus shows signs of wanting to create a "portfolio" business. Artisan isn't interested in "bundling" different types of IP to help encourage a sale, as the software industry does, Templeton said. Customers do purchase multiple products, but that's more because they like Artisan than because they're interested in a bundled deal, he said.
Geoff Tate, chief executive of Rambus, likewise didn't see his company branching out. "A company comes to us and they've got a particular problem: How do they get the highest possible performance out of their memory subsystem," he said. "How they design the rest of the system is up to them."
Royalties have also been a touchy subject. The key for Rambus has been to keep the royalty cheap, even if it feels like it's below market value, Tate said. "If we charged more, we couldn't have gotten critical mass to get people interested in using us in the first place."
But Tate admitted that Rambus has prospered under that model largely because its memory architecture is so widely applicable. "You've got to be addressing a very big market," he said.
On top of that, the IP business is going to be one of long relationships, Tate and Templeton said. Chip makers see a coming dependence on IP, while IP makers realize they've got to target large customers and/or long-lived cores to make any decent money off royalties, so both sides are interested in finding partners they can stick with for the long haul. In fact, as Tate pointed out, some of Rambus's licensees have been with the company since its founding in 1990.
And hand in hand with that, the IP industry had better get used to building perfection into its wares, Templeton said. "We're not selling software. We're not selling products with millions of permutations," he said. IP, in fact, can be 100 percent verified. "Customers will not have the tolerance for defects or bugs that they had with the EDA industry," he said.
It's also unclear what effect an IP industry would have on other businesses. Most of the concerns being aired revolve around ASIC vendors, whose business appears threatened by an industry that makes IC cores available to all comers.
But gathering up IC cores isn't the same as running an ASIC business, said John Daane, executive vice president of computing, communications and ASIC products at LSI Logic Corp. (Milpitas, Calif.). LSI Logic can stand out among the growing IP ranks because it offers a unified methodology that simplifies the integration and testing of blocks, Daane said. It also gives customers the ability to change a core without worrying that the new version won't knit properly into the ASIC, he said.
In addition, he said, IP firms don't have the money to fully verify a system-on-a-chip or indemnify the IP buyer in case particular cores don't interoperate properly-two things LSI Logic is able to do. Indemnity in particular is a huge advantage over small IP providers, Daane said, because without it, "you may find yourself in a situation where legally, the TIs and Intels and other large companies may be coming after you."
In addition, LSI Logic is doing a lot of design services, something not found in the IP business model or even in many ASIC firms, Daane said.
But just as LSI Logic is moving into services, the IP industry begs the question of whether service companies can encroach on ASIC turf. In particular, the industry has its eye on Cadence Design Systems Inc. (San Jose, Calif.), which some feel is coming perilously close to the ASIC business model.
The company's shift into design services came with a pledge that Cadence wouldn't start collecting IP, a move that would make it a de facto ASIC house. Given that pledge, Chaplinsky of Hambrecht & Quist asked Cadence director of strategic marketing Steve Glaser why the company recently hired microprocessor engineers in Palo Alto. Glaser said Cadence needed that expertise to be able to help with designs.
"We sell to a lot of semiconductor companies who don't see [us] as threatening but are starting to see . . . that no one company can do it all," Glaser said. Chip firms stick to their own IP and manufacturing as their specialties, while systems companies usually have expertise in architectures and software. In both cases, the companies find holes in their expertise that they might turn to Cadence to fill. What Cadence's customer base wants is "an applications expertise," and the company was trying to acquire it through those microprocessor engineers, he said.
That's a business angle that foundries won't touch, Glaser added. "And we wouldn't recommend it."
Daane likewise offered caution to any EDA companies that become more ASIC-like by offering cores. "What's going to be an issue there is if EDA companies become ASIC companies, their tools become captive tools," Daane said. "If ASIC companies felt EDA companies were getting into their space, they'd go somewhere else [for tools], maybe fund startups."
Glaser said Cadence's role for design services lies largely in a middle ground where customers are doing their own designs but adding choice bits of IP from outside vendors. And Cadence expects that area to stay fertile for a while. "The world doesn't start today at ASICs or custom logic and jump tomorrow" to systems-on-a-chip, Glaser said.
Another concern often aired is that semiconductor companies' own competitive edge might fade if design expertise can be bought for a small royalty. But venture capitalist Davidow said that fear was misplaced, because many chip makers' strengths lie not in chip design but in manufacturing-something the semiconductor firms themselves don't always understand, he said.
"IP doesn't change that equation very much," he said. "Too frequently, people who are in, say, the manufacturing business try to lever their position with a clever piece of IP, and then they forget how to manufacture. That is rampant in our industry." |
