Everyone,
Any comments:
Intellectual-property cores biz: Harder than it
looks
-- Sat, 20 Mar 1999 01:36 EST
Mar. 19, 1999 (Electronic Engineering Times - CMP via COMTEX) -- The
intellectual-property cores industry is still the talk of Silicon
Valley, but the talk this year is turning sour. The dream a year ago
was that myriad startups would spin semiconductor cores that larger
companies would knit together with ease into systems on chips.
But the IP cores business is turning out to be harder than anyone
thought, in terms of how to sell the cores, how much support is
required and even which technologies can make a viable IP cores
business.
"The dream is to go surfing on the Web and grab this core here and
that core there. That's not going to happen," said Gary Smith, who
tracks the cores industry for market researcher Dataquest Inc.
Investment interest has waned, too, and even the success of public
cores suppliers such as Advanced Risc Machines Ltd. (ARM) isn't being
taken as proof that IP can build a viable business. While most industry
observers agree that an IP cores industry will eventually exist, it's
becoming apparent that the model won't be as easy to build as was
originally thought.
"Once we know how to productize IP, it will be of huge value. The
business model is still a question mark," said Sam Lee, a partner with
Information Technology Ventures.
The IP model says you remove the major costs of semiconductor
development-manufacturing, primarily-and concentrate on designing
circuitry. It also says you're able to sell that circuitry to multiple
industry players using multiple fabs and to collect ongoing revenues,
usually in the form of royalties.
Expectations for that kind of business remain lofty. DMG estimates
that the merchant IP business will hit $1 billion next year. "I've seen
estimates that are four times that," said John Bourgoin, chief
executive of MIPS Technologies Inc.
Last year, with a few IP companies going public and valuations
rising, "everybody was kind of bullish," said Bruce Graham of Bessemer
Venture Partners. But despite a few early successes with companies such
as MIPS and Rambus Inc., problems erupted. Some designs just didn't
work inside a customer's chip. Prices for commodity cores plummeted.
The wreckage has left many venture capitalists gun-shy about IP
companies.
"Sand Hill Road has shut down," said Mark Bowles, president of
DSP-core company Billions of Operations Per Second Inc. (BOPS).
"They're just not funding IP deals because the model's not proven."
Sand Hill Road is the address of many of Silicon Valley's venture
capitalists.
In this environment, even Rambus and ARM aren't considered to have
proved themselves yet, because they haven't weathered the public stock
market for very long. "
Part of the problem is that IP hasn't become the kind of gold strike
that attracts technology investors. Compared with an e-commerce
company, or even a fabless semiconductor house, the revenue growth of
an IP play isn't fast enough for venture capitalists' needs.
"ARM just reported $69 million revenues. That's not a bad fabless
company, that's an out-of-business fabless company," Bowles said. "Even
if you're wildly successful, you end up with these low numbers that
aren't too exciting to a venture capitalist."
In many cases, investors believe in the IP concept but are waiting to
see if the business model works long term. Despite his reticence
against IP companies for now, Lee thinks the market is a shoo-in for
the future. "The fundamentals are certainly there-design reuse, the
ability to buy off-the-shelf components-it makes a lot of sense," Lee
said. For Bill Elmore, a partner at venture firm Foundation Capital
(Menlo Park, Calif.), the question is how large an IP company can
become, even if it's viable. "How many Rambuses are there going to be?
Two or 200?" Elmore said.
Others remain skeptical. Bruce Graham of Bessemer Venture Partners
said he has stuck to semiconductors, his original area of expertise,
and hasn't pursued an IP deal for the past six months. "Frankly, I
don't really believe in the model," Graham said. "I'm more comfortable
with IP as an adjunct to the fabless [semiconductor] model."
Possibly the most bullish of venture capitalists is Rob Chaplinsky of
Mohr, Davidow Ventures. As a stock analyst with Hambrecht & Quist,
Chaplinsky was an early champion of the IP concept and helped shepherd
Rambus and ARM into the public market. He remains upbeat about the
future of IP companies, but he also concedes that early struggles have
left some investors skeptical. "People are a lot more familiar with the
model, but I think people also recognizing the challenges," Chaplinsky
said.
In fact, some venture capitalists are trying to steer IP-cores
hopefuls toward a fabless-chip model. But some companies-BOPS, for
instance-are even less enticed by the competitive conditions they'd
face as fabless semiconductor companies.
Certainly the demand for silicon cores is real, as Graham has
discovered as a board member of Mips-device processor Quantum Effect
Design Inc. "Companies are coming to us saying, 'Is there some other
way we can use this technology?' We're scratching our heads trying to
figure out if we can do that."
Because not every IP play has flopped, it's apparent that the
business model can work in some cases. Dataquest's Smith, who divides
IP cores into three business models plus the traditional ASIC one, said
it's mainly the "independent" vendors-those trying to do scattered,
individual cores-who have disappointed. The remaining two business
models are doing rather well, he said.
Smith believes the "star" core providers-MIPS, ARM and DSP Group
Inc., in his book-have managed to build sustainable businesses. Also
promising, Smith said, are the "core stores" such as Mentor Graphics
Inc.'s Inventra group, which are acquiring IP cores and can therefore
sell several into one chip design.
Two elements seem to be essential to the success of the star core
providers: a link to a massive market and the fact that the companies
sell processors, Smith said. Processor vendors are particularly well
placed due to the emerging embedded market and "anything but Microsoft"
philosophy that's cropping up in pockets of the embedded market.
According to Smith, embedded designs are being developed without
allegiance to a particular microprocessor and standardized on a
real-time operating system from a small company, which opens the door
for a variety of processors to be used in a variety of designs.
The core store model, originated by Technical Data Freeway Inc., is
being perfected by Mentor Graphics Corp.'s Inventra group and by
Phoenix Technologies Ltd. "Phoenix and Mentor are buying up all the
real good independent core guys," Smith said. He believes there's room
for one or two more large core stores. "My suspicion is that Synopsys
will be there someday."
So, what is it exactly that IP companies have been missing? One quick
business lesson was that IP works best when it's specialized. Commodity
products attract too much competition. PCI and USB cores are prime
examples, as their prices have plummeted. But at the same time, that
specialized IP has to be a sustainable business, which means finding a
high-volume application to match.
IP companies also must take an approach that goes beyond providing
components. It's becoming clear that IP operations, like the rest of
the electronics industry, must follow a systems-minded approach.
"In an IP model, you'd better be service-oriented. In a fabless
model, you have to be manufacturing-oriented. But in both those models,
you have to be systems-oriented-you have to understand your customer's
customer," Chaplinsky said.
Perhaps most discouraging to investors is that it took years to build
the successful IP core firms. Rambus "had a lot of ups and downs until
it got pushed over by Intel and others," Graham said. ARM took seven
years to break through; "I'm not sure you could keep engineers around
for seven years in Silicon Valley to build a company that way," he
said.
But it's really the unforeseen technical issues, such as verification
and customer support, that have doused hopes for a fast-paced, flexible
IP marketplace.
"The thing that's killing off these IP guys is they're finding that
support engineering is costing four times what it cost to build the
cores in the first place," Smith said. "Really, what ASIC guys are
looking for now is silicon-proven cores. There's just too many
disasters out there. It's a shame, because a lot of the time it was the
engineer's fault because they were messing with the core."
"What separates the real companies from the new, young companies is
the companies that have a verification procedure built in," Chaplinsky
said.
Legal concerns also must figure prominently into the business plan,
because an IP company lives or dies by its patents. Although not yet a
pressing problem it's possible-indeed likely-that legal questions will
be the next hurdle for the IP industry.
-0-
By: Craig Matsumoto
Copyright 1999 CMP Media Inc.
|
