By Don Clark
Moore's Law helped bring numerous startups into existence, as
cheaper chip technology drove developments like personal computers,
smartphones and the Internet. But the force that sparked so much
innovation lately has thrown cold water on semiconductor
startups.
In 2004, venture capitalists helped 62 chip startups get off the
ground with seed or first-round funding, according to Dow Jones
VentureSource. Since then, investors have become increasingly wary.
Just four startups received first-round infusions in 2013 and seven
in 2014, with no seed investments in chip companies either
year.
The money flowing to semiconductor companies in such early
rounds shrank from more than $400 million to $94 million over that
time frame.
"No one will take the risk," said Alex Lidow, chief executive of
Efficient Power Conversion Corp., who opted to fund his chip
startup with his own money and contributions from a manufacturing
partner in Taiwan.
The primary reason is cost. Moore's Law, as Intel's co-founder
amended it in 1975, states that chip makers will squeeze roughly
twice as many transistors on a chip every couple of years. That
progress brings many technical benefits, but designing products
with more transistors takes more time and money, as does testing
them.
International Business Strategies, a Silicon Valley consulting
firm, estimated that it costs $132 million to design and test a
typical chip at the current state-of-the-art width of chip
components, 14 nanometers, or billionths of a meter. At 65
nanometers, a component size introduced about a decade ago, a
comparable chip cost $16 million.
Consequently, chip ventures require more startup capital than
many types of early-stage company. They are also less likely to pay
off quickly or reliably. For instance, a chip made using the latest
14 nanometer process would need to bring $987 million in revenue to
meet a return benchmark of 7.5 times the cost of its design, IBS
estimates. A 64-nanometer chip would need to take in $123 million
to meet the same benchmark.
The story of Tabula Inc. illustrates the enormous pressures on
chip startups. The company was founded in 2003 to develop a novel
programmable chip. It used Intel's 14-nanometer process and
designed a product with more 5 billion transistors.
"The complexity and time and expense associated with doing
design at that level was tremendous," said Dennis Segers, who
served as Tabula's CEO since 2006. "It proved to be a challenge to
us."
The company also faced high expenses to develop software to
program its chip, and delays in getting customers to adopt those
new development tools, Mr. Segers said. As a result, despite
raising more than $200 million, Tabula's investors threw in the
towel. The company, which employed 120 people at its peak, recently
closed its doors and is selling off its remaining assets.
Big companies are still designing plenty of chips. And some
startups founded in the past decade continue to get more money.
Movidius, a company based in Ireland that makes chips used in
visual-processing applications, announced on Tuesday a $40 million
funding round, one of the biggest for a semiconductor startup in
two years.
One investor who has continued to invest in semiconductor
startups is Lip-Bu Tan, who is CEO of chip-design software company
Cadence Design Systems Inc. as well as chairman of the
venture-capital firm Walden International. Some of the recipients
are foreign, including India-based Ineda Systems Inc., which makes
chips for wearable devices.
"I decided to be a contrarian," Mr. Tan says. "I doubled,
tripled down on my bets."
He is the exception, though, and some people worry about the
dearth of new entries to the chip sector. The decline in startups
means fewer mature companies. Also, in the past, such ventures
often were acquired by larger chip companies, helping them develop
new products.
"If you look at the semiconductor business, the majority of the
innovation has been contributed by startups," said Andy Rappaport,
who recently retired from the venture firm August Capital, where he
concentrated largely on chip companies.
As venture capitalists shy away from chips, Mr. Rappaport
expects some large companies will step up to fill in the gap.
Meanwhile, startups that have emerged are devising new strategies
to survive. Some are developing products that exploit older
production processes.
Others are trying new materials. Mr. Lidow's EPC, for example,
is developing chips based on gallium nitride rather the silicon
that has been semiconductor industry's mainstay.
But most others will have to adapt to the harsh realities of
keeping up with the competition, which means greater spending to
stay in the technology race.
"Moore's Law is a discipline for all of us," said Randhir Thaku,
executive vice president and general manager of the Silicon Systems
group at Applied Materials Inc., the big maker of semiconductor
production tools, "to make sure that we behave."
Write to Don Clark at don.clark@wsj.com
Access Investor Kit for Intel Corp.
Visit
http://www.companyspotlight.com/partner?cp_code=P479&isin=US4581401001
Subscribe to WSJ: http://online.wsj.com?mod=djnwires