Since the recession in 2009, the semiconductor industry has led the recovery and is projected to see between a 10 to 15 percent growth for 2010-2012, far exceeding what’s projected for the global gross domestic product (GDP) growth.
Here, in the Silicon Valley, the semiconductor industry is very much yesterday’s news, ignored by venture capitalists, talented employees and trade publications. That makes industry pundits wonder what’s happening to the U.S. semiconductor industry, especially the startups who aspire to become another Atheros, Marvell or NVIDIA.
In my opinion, 2011 will be the year that transforms the U.S. semiconductor business model from fabless companies designing chips and building them off-shore, to focusing on just licensing the core technologies to chip companies worldwide. This model is formally called the semiconductor intellectual property (SIP) licensing business, and after its meteoric rise in the mid-1990s and much neglected due to a lack of success, is finally making a comeback again.
In many ways, the semiconductor industry has been the fastest to globalize. Companies large and small are all multi-national; most of them outsource manufacturing, packaging and other logistics. And, for most of the 2000s, entrepreneurs believe that Silicon Valley can be the control center of a chip company, conceiving ingenious product concepts, attracting funding and expertise and executing globally.
Unfortunately, the model is broken for several reasons. As emerging markets account for the bulk of the growth and now volume, product concept and management move with it. As bio-science, on-line media and, recently, clean tech soak up bulk of the venture investment, there is also little interest to fund $50-$100 million for a chip startup to be successful in the U.S.
As a result, few new startups are funded, and most of the early-2000s funded companies are crawling along on their own, most of them cut off from further funding to make the leap to success.
If the end to globalization is ultimately efficiency in resource utilization, then it’s important to focus on what the U.S. does well: freedom to innovate and a diversity of talent. With these two unique ingredients, the inventions often deliver disruptive benefits to consumers.
Funding a company focused on inventions is a small fraction of what it takes to fund a full-scale semiconductor chip company from ground zero to profitability. For example, it often costs more than $20 million for a chip company to wait idly for 18 months from an engineering sample to full production. After that, substantial more dollars are needed to fund the ramp-up in production, a process already perfected by established companies worldwide.
Instead of spending big portions of the $100 million re-inventing wheels, new startups will be focused on delivering the invention only as a small portion of the chip designs that make big impacts. Companies such as Kilopass and SuVolta are following the success models of ARM, Rambus and Tessera to license their technologies to their semiconductor companies. So far, so good. What’s new, and why 2011?
Well, there are, by estimate, still more than 500 semiconductor-related private with strong patent inventions but no hope of attracting the rest of the $100 million needed to get to market. The only way for them to make a market impact, and return profits to investors, is to focus on becoming IP suppliers instead of chip suppliers. Undoubtedly, most of them have thought of this idea, but were reluctant because, until recently, semiconductor IP exits –– that is, a public offering or a merger or acquisition –– have been few to none, until now.
Long time Electronic Design Automation (EDA) software power-houses Cadence and Synopsys have decided to pay premiums to acquire two companies focused on licensing semiconductor IP. Cadence, for example, paid 5X of revenue to acquire Denali Software, a DRAM IP supplier. The semiconductor industry is $250 billion, and with cost of goods sold (COGS) being about half of that, there is plenty of room for innovative technologies to help reduce $125 billion worth of cost. Or, with the same $125 billion, deliver more than $250 billion of value to the end-user.
Does the math work? Assume that about 20 percent of COGS have been delivered through the semiconductor IP business model. This translates to about $25 billion in revenue, or 20 percent of $125 billion in COGS. At 5X revenue, this translates to about $125 billion in valuation. Spreading this wealth evenly across the 500 now reshaped semiconductor IP companies, an investor would be looking at $250 million in average return.
Even at an average investment of $50 million –– half of a fully developed chip company’s funding needs –– each investment would average 5X return.
That’s a very respectable average return indeed. And, that’s why I’m optimistic about 2011. About the author: Charlie Cheng joined Kilopass Technologies Inc. (Santa Clara, Calif.) as CEO in 2008 with overall responsibility for leading the company and its business growth worldwide. His career spans more than 10 years of experience in the semiconductor IP industry and an additional 15 years in the global technology marketplace. Prior to Kilopass, he served as vice president of marketing and international business at Faraday Technology. Previously, he was co-founder and CEO of Lexra Inc., an embedded RISC CPU IP company.
USA is giving up the semi industry,just like the textile industry in the past. The IP model won't work for majority of the semi start-ups since the rev is one hundred fold less than making a product to sell. A lot of Talents are flocking to the emerging market in Asia , like Morning Star and Ralink in Taiwan. They are all founded by the Veterans from the Silicon Valley.
Another one is the Sequoia Microelectronics, they sepcialize in the LED lighting ACDC offline drivers SQ9910 series.
Check out the website.
The real issue seems to be lack of funding. Don't you think that the VCs would put up their $ if they believed a good return was possible? There seems to be a lack of good examples to show a big return is possible. Until we get a few funding may continue to be diverted to those industries that show the VCs the money...
As others have pointed out, design and manufacturing complement each other. Companies with the strongest IP portfolios are not the pure play IP companies but the ones with their own manufacturing or at least have their own products. By moving up the value chains, you start to lose more and more controls to others. Not only you become a smaller player but your quality also suffers especially in the advanced technologies. Being smaller also means you become more dependent on and have less leverage over the big players. The US semiconductor industry will become irrelevant if this is the direction to go. Regarding why investors are not hot on the IP start ups, I personally believe that they would rather be looking for the next Intel or Marvell who has the potential to make it big. If it is not in the semiconductor industry, they’ll try other industries.
I agree with you no.name_#4, the US has what it takes to be strong at all steps of the value chain. Smaller countries with small economies and markets should specialise in one or few steps and divest from others, but the US does not have to IMO.
This debate seems to have the U.S. conceding its significant and growing semiconductor manufacturing base with state of the art 300mm fabs (Intel, SAMSUNG and GF).
Yes ARM and IBM both do indeed have a significant global chip market share and business in intellectual property (SIP) licensing. However, there seems to be some overlooked elephants in this room. Also it’s not certain if cutting edge chip manufacturing (with latest chip IP) can truly be economically ported to any foundry with out significant yield impacts.
When it comes to world’s best manufacturing and much envied IP, Intel has it. They have it because they never brought wholesale into off-shoring or conceded world class quality. Their Chip design feeds into their manufacturing which feeds back into their design. Only Intel is doing this well, it seems.
There was a point when Intel’s chip design had fallen behind AMD; but, they bounce back way stronger and opened up another market front in netbooks.
Intel’s manufacturing die yields and throughput has never lagged behind any fab!
SAMSUNG and GF both are spending billions in CapEx to equip new US 300mm fabs and so is Intel.
As Intel, SAMSUNG, GF and TI show; IDM, FAB less/light and Foundry models are continuing to grow a larger significant US manufacturing base with state of the art 300mm fabs.
Another way to look at it, its a move higher up the margin chain.
Most IP companies enjoy +70% GPM, even 90% GPM is reachable.
My concern, silicon valley investors are not so hot on IP start ups. We need more execs like Charlie to show the way.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.