SAN MATEO, Calif. Moribund end markets for electronics, a surplus
of competent design teams looking for funding, rapidly escalating design
costs. It's a brew would appear poisonous to any hopes an IC entrepreneur
might have for starting a company. Yet some funds are showing some renewed
interest in fabless semiconductor startups, and the number of fundings
is starting slowly to grow. It's a new world, with very different rules,
but it's not a dead planet.
There is money to be had. It is cautious money money that insists
on understanding the end-market the chips will serve and on active involvement
in some important decisions, such as outsourcing. But it is also money
that goes in for the long term.
So suggests William Quigley, managing director at Clearstone Venture
Partners (Menlo Park, Calif.). "We see 2003 as a sluggish year, and
we basically have no visibility into 2004," he said. "So, lacking
anything to the contrary, we assume that 2004 will bring a return to reasonable
conditions for the electronics industry."
That timing, Quigley explained, is great news for some companies and
dire news for others. "There is definite activity in funding early-stage
companies," he said. "These are companies who, if they get started
now, will have their chip completed just about in time to catch the 2004
recovery, should it occur."
But for older startups the outlook is far darker. "The older companies
tend to have taken a huge amount of money in their early rounds, back
during the bubble. They have burned through that and are now sitting on
"Worse, their technology is old now," Quigley said. "It
matured when there was no market, and so they have never been able to
ship anything. Even a socket win without any volume is worthless today.
So to get another round of funding, these older companies would have to
go out and compete against companies with much more recent designs for
sockets that will actually result in shipments."
Thousands of failures
To emphasize the seriousness of the situation, Quigley estimated that
the number of failing ventures in the coming year could reach the thousands.
"It will be painful, but it's probable that most of these investments
have already been written down in their investors' portfolios," he
In particular Quigley pointed to the coming debacle in the 802.11 system-on-chip
sector. "You've got maybe 50 or 60 multimode chip designs out there,
all chasing one market. Granted, that market will have some attractive
volume, but not that much. These guys don't have a prayer. It's in some
respects like the situation in network processors."
So is there any hope? Definitely yes, Quigley maintained. But the rules
for successful new ventures would be very different than in the bubble
To start with, Clearstone doesn't just fund neat ideas or hot design
teams. Rather, the company, like many experienced investment teams today,
systematically looks for end-user markets with significant volume potential
and significant entry barriers in the time frame in which the startup
could realistically complete a design. That means, for example, that Quigley
is moderately interested in some aspects of consumer electronics and networking,
and not particularly excited about optical communications.
"In optical, you have Nortel and Lucent with several years of inventory
on hand," he observed. "But in some other areas, where the end-customer
needs are shifting, the inventory on hand is sitting there growing obsolete.
It will have to be replaced."
There will not be much more blue-sky investing in chip startups, Quigley
predicted. For one thing, venture capital firms are syndicated now, and
they tend to watch how many ideas have been funded in a given area. "If
you are not one of the first four companies to get funded, you are going
to have to be something really special to get any interest," Quigley
Further, having been once burned, most venture funds not only conduct
careful investigations of the proposed team but also undertake due diligence
on their proposed market. The days when you could pass off a list of founders
as a business plan are long gone.
Finally, with risks now better appreciated and with design costs spiraling,
Quigley said that the new-venture investments would have clear strategies
not just for launching a design effort but for carrying a company through
to volume production. "You can't assume that anyone but insiders
will be in for subsequent rounds," he said. "In that regard,
venture investing is starting to look like the leveraged-buyout market.
The investment has to have a path all the way to break-even, not just
to Series B."
Part of that plan, Quigley said, is much more active involvement in how
the design is conducted and, specifically, how much it costs. The problem
is not the much-touted million-dollar mask set but the whole design cost.
It is commonly estimated that the total design cost for a 130-nanometer
system-on-chip will be in the neighborhood of $20 million. That won't
cut it, according to Quigley. And the key to beating that number is design
"We don't fund chip designs that don't outsource to India,"
Quigley stated flatly. If you rely on Indian contractors for the things
they do well, you can get a chip out for under $10 million. If you don't,
you can't, and you won't be competitive. It's that simple. Even major
Japanese semiconductor players, Quigley said, are starting to use the
"That means that we can possibly hit break-even with a company for
a total investment of $25 million," Quigley said. "That is an
attractive enough possibility to make us quite determined about this.
"We like to see a design team that has experience with outsourcing
significant parts of the design. We like to see a vice president of engineering
who has existing relationships with Indian contractors. There's a set
of skills there that is, we feel, necessary for success now."
There, in Quigley's view, is the new environment. Investors now expect
to be with a company until volume production.
But they won't be jogging along blindly, the director added. "We
will be continually watching to be sure that our original assumptions
are still valid."