For almost 50 years venture capital and the semiconductor industry have been intimately linked. The stories of how the two have evolved together are the stuff of legend. How a letter from Eugene Kleiner found its way to the desk of Arthur Rock and led to the founding of Fairchild Semiconductor. How Jerry Sanders convinced Steven Merrill, who had never invested in semiconductors, to help fund the startup of AMD.
The stories continue to be written. If we had to predict who would be the central character 20 years from now, we’d put our money on Lip-Bu Tan, chairman and founder of Walden International, someone we’ve taken to calling “Mr. Chips.”
With 12 chip investments made between November 2003 and January 2005, Tan and his colleagues at Walden outpaced the likes of Intel Capital, 3i and JPMorgan Partners. And, Tan tells us, Walden is betting exclusively on semiconductors (and China) for the foreseeable future. It is one of the boldest statements by a venture capitalist that we can think of, but not the only one Tan made in a wide-ranging interview.
Q More than $3 billion went into chip investments in 2004, considerably more than 2003, and well above 2002 and 2003. Why?
A The founders of semiconductor startups are creating opportunities that make people want to invest. For a long time it was just handsets that were driving the market. Now there are startups in digital TV, home-gateways through set top boxes, projector televisions, communications such as WiMax and all of the consumer wireless space.
Was too much money invested in semiconductors in 2004?
Yes and no. There was a lot of money in a few areas that were over-funded and some areas where there was not enough.
Which areas have been over-funded/under-funded from among the sub-sectors fabless (including imaging and communications), design and manufacturing, memory, processors or novel technologies, and sensors?
Well, money that VCs put into process technologies is hard to imagine as successful. The IDMs [independent device manufacturers] have to do that investing. Memory, well there are some opportunities, but not a lot. Samsung is dominant there and there are only a few niches, like low power or high speed. In imaging there are some opportunities, making sensors from CMOS instead of using CCDs. Very high-resolution displays has some niches for investing. Design for manufacturing has opportunities as geometries are shrinking.
Which sub-sectors are overcrowded, most prone to failure by investors?
Process engineering. And within fabless models, WiFi, storage [processors] security, graphics and MP3.
Which of the sub-sectors still has room to invest?
It’s not that simple. This is a $200 billion per annum industry. But in 2005 it’s going to be flat. There is maybe 10% growth in the industry over the next five years and at the same time prices [for semiconductor devices] are declining. So there are not lots of growth opportunities. So if you want to invest, you have to go to areas where there are big technical problems to solve or areas where you want to be over the long term, like power devices and advanced technology like processors.
Which of the sub-sectors shows the most promise and why?
There are so many. Take any one, say automotive. There are so many uses of semiconductors today in a car: imaging, sensors, GPRS [general packet radio service], entertainment, transmissions. That one area presents a huge set of opportunities over the next five years, and there aren’t many investors there yet.
About 440 VCs invested in at least one chip deal in 2004 vs. about 150 in 2003. Are too many VCs jumping into this sector without really knowing what they’re doing?
The answer is yes. If you ask me, there are fewer than 10 really solid investors in this field. But because there is so much money raised by VCs a lot of it goes into semiconductors. Look at the over-investment in communications over the last five years. Look at areas like enterprise software where the big incumbents are consolidating and where the money that went into enterprise software will never be recovered.
What is the end result of more VCs participating in semiconductors?
Lots of startups are receiving $50 million to $100 million and that leaves very little opportunity to make a profit from these companies. As a result of that over-investing, you have to find ways to succeed. For example, by focusing on technically difficult problems or by exploiting opportunities for lower cost design and manufacturing in Taiwan, India or China, where you spend less in the development cost and leave room for your profit margin. You have to go overseas now. In the old days it cost $8 million to $10 million to design a product. As we moved to .13 microns, the cost went to $20 million. And now that we’re going to 90 nanometers, the cost to produce a chip in the United States is $50 million. And then there are lots of yield problems at 90 nanometers. So you have to go to emerging markets to make a profit.
Will the majority of semiconductor investments made in 2004 be successes or failures and why?
The majority of them will fail or will make too little money. The returns are going to be so poor. [Some of these VCs] are spending $80 million to produce a chip set and then they can exit for only $100 million. The only way to make 10x on your investments in semiconductors today is to decrease your cost of manufacturing or to go into consumer electronics where there is a huge market for your parts.
Are we in the middle of a semiconductor investment bubble?
We’re heading there. And if VCs are not careful it will be a big bubble. For example, in China there are 480 fabless design companies. About 90 will survive and, say, 10 will become world-class companies if they work closely with their customers, study carefully and work hard. Look at WiFi: Ninety firms put cash into that area, and maybe three or four will make any money. The rest will lose their money.
Is this the most competitive environment you’ve seen in semiconductor investing?
Very much so. Especially if you’re looking at a good company with strong technology or that’s addressing a big problem. The only way to get into a deal like that is to sell your value-add. I just finished closing a deal for such a company. There were eight other top-notch VCs bidding on the deal against me. I had to compete on the value that I could add to the deal.
Just how competitive is it?
VCs are so anxious to enter semi investing that a deal I just made six months ago went out for another round that was prompted by another top VC firm which was ready to pay 2x to 5x on my valuation in order to get into the deal. After just six months! Just to pre-empt any other VCs later.
So what happens when valuations are so unrealistic?
If valuations are too high, the margin for venture capital goes down. When too many VCs rush into a sector, margins [and returns] go down.
Are there variations in valuations across the various sectors?
Very much so. Fabless investments are rising from $30 million to $50 million for a startup. Leadis Technology, a company we IPO’ed last year only needed $12 million to be profitable. Now most deals cost $50 million. In fabs, it takes $1.5 billion to build a new fab that will have a profitable business, like SMIC. In equipment it costs $55 million to $70 million to build a company to profitability. Lots of variations. And if you’re doing a complex fabless wireless chip, that costs $80 million to $90 million to develop.
Where are valuations headed across various stages by year-end?
A typical series A round of $10 million is $6 million to $15 million pre-money valuation. For a series B, unless you are talking about a down round, is $45 million to $50 million pre-money, with a minimum of a 1.5x mark up for a good company, but a really good company commands 2x or 3x. And you raise $20 million in the B round, so post-money you’re at $60 million to $70 million in valuation. At a late tape-out or C stage company, with a high valuation, the C round leaves you with a $90 million to $100 million valuation. You have to make $500 million to $700 million in revenue to make your money back on that kind of an investment through an IPO. And there are not many such companies these days.
So if Walden or another VC wants to beat that scenario?
Most of us want to invest in an A round where we can still make 10x on our investments. There is just too little profit left in subsequent-round investing.
Of the 12 chip deals, Walden did last year, how many were new investments?
What was the round of each deal?
They “were all A rounds, with one exception, Anyka in China, which was a C round.
Why did Walden make so many chip investments in 2004? What’s the rush?
We’re in a push, and all of my investment focus is towards the 2007/2008 years when much of the consumer electronics we’re investing in now will come to maturity. There are a few things driving that. China is developing a digital TV standard. In Asia, in particular, that is going to be a big driver. The use of broadband wireless at the 2008 Olympics in China is a part of the big push for infrastructure development there for the delivery of video over cell phones. To take part in those events then, to benefit from them, we have to be investing four to five years earlier, so now is the time to do that.
What do you think your investment pace will be next year?
So far, we’re doing 10 to 12 investments per year, and I think we’ll maintain that over the next year or two years. The entire team works on that. There is Andrew Kau in the United States, Brian Chiang in Shanghai, Kee Lock Chua in Singapore and myself all doing semiconductor investments.
What was your favorite chip investment in 04 and why?
Two stand out in my mind. Beceem Communications, a complex fabless semiconductor startup, is working in 802.16 addressing media and mobility applications. The team there is from Centillium. They’re a really a top-notch design team. Mark Stevens at Sequoia is my co-investor in that. Then there is Telegent Systems, a fabless video infotainment application company, with one of the best analog mixed signal teams addressing a huge market, first in Japan, Korea and China, then later in Europe and the United States.
Of the 12 deals Walden did in 04, how many did you personally do?
How many board seats did you and/or Walden take out of all the 04 investments?
We have board seats on all of our investments.
And you sit on how many?
I don’t know. We aren’t a typical firm. All of our board seats are occupied by two team members so that there is always someone available to participate if someone else is busy. But I probably sit on more than 10 and less than 15. And some of those, like SMIC or Leadis, I’ll retire from now that they have gone public.
Are you stretching yourself too thin? Most VCs will do no more than one to two new deals per year and try to limit their board seats.
Not for now. I don’t miss any board meetings. But I am extremely active with all of my companies. For example, with Telegent, which I invested in during 2004. The CEO and I made seven trips to customers in Asia last year. The CEO told me the other day we had exchanged 300 emails during the year. And I did four trips to customers for Beceem in Asia last year.
How long have you been investing in semiconductor startups?
That’s a long time ago-20 years. I think my first investment was S3 [now Diamond Multimedia], a graphics chip company with Paul Wood.
That isn’t around anymore?
Yes, but it did very well. It went public and we made 9x or 10x on our investment. Eventually I left the board and it was sold to Diamond.
How did you decide to make that investment?
John Doerr of Kleiner Perkins invited me into that deal. There were lots of Taiwanese companies making personal computers at the time and I knew the Asian market well, so John asked me to participate. So I led the B round, brought several friends into the deal, companies like Mitek. The founder, Diosdado Banatao, is the best semiconductor entrepreneur that I know. He built S3, then Marvell, then SiRF Technology, in which we invested and took public last year. I invested in all of his companies.
What was your most successful chip startup investment?
I don’t think I could name just one. We had a couple of semiconductor deals that were only average, but in terms of returns we’ve had so many that returned 10x: Integrated Silicon Solution Inc., Centillium, S3, IC Works, Newave Semiconductor/IDT, Electronic Resources, Transmeta, Microchip, Leadis, Unisem.
What did you learn from all those successful deals?
Well, I continue to learn every day. Lately there have been two things that have driven the deals that I have done. First, I try to focus on startups that address big markets and I try to invest in companies that can be the No. 1 or No. 2 in their market. Like SMIC, it’s the No. 3 foundry in the world after just three years. After that, I’ve learned that today there are more opportunities for teams with strong technology backgrounds that can out-design their competitors. Diosdado Banatao just out-designed his competition to build Marvell. I try and take those lessons with me to China for my investments.
What has been your worst investment?
The only investment loss in all my semiconductor investments is Invox in which Walden International invested $2.8M. I sold the company with a $1.2M realized loss, so I recovered 60% of the capital.
Did that discourage your enthusiasm for semiconductors?
On the contrary, it’s the area where we have done the best. That’s why 100% of Walden’s investing is going into semiconductors now. I’ve returned to my core competency and I am applying the lessons that I’ve learned in semiconductors.
You’ve wound down your life sciences investments entirely?
Yes. All gone. And communications equipment continues to wind down. We still had some work in communications over the last two years, but the burn rate of communications and infrastructure equipment companies is from $1 million to $5 million a month, so those companies suck in $100 million to $150 million in investment before they get near break-even. So at least temporarily [we’re not investing there] while the incumbents are doing so little investing in new equipment, say for the next two or three years or until the communications companies come back to the market for equipment. Then we’ll reconsider.
What are the key areas in chips that you have focused on and why?
Well we’ve already reviewed the reasons but the areas include the analog mixed signal space, display and mobile computing, video, a few narrow areas of networking and communications. We selectively back foundries like SMIC. And we back equipment companies selectively.
Which areas in semiconductors have you avoided and why?
Memory. It’s very cyclical. Any areas in microprocessors. We can’t compete with Intel, although in the past we did Transmeta and made 10x on that investment. Basically anything that is capital intensive, like foundries, with the exception of SMIC where we think they can be one of the top three competitors.
What is your investment strategy going forward and why?
Well as I’ve said fabless is over-crowded. In the past we have done much the same as we did last year. But going forward we’re going to pursue deals that leverage cost structures or that will focus on big markets and technical leadership.
How does that differ from what you have been doing in the past?
I think in the past my focus was more on the quality of the management team.
Do you bet on technology, markets or people?
On all three.
Why are you directing Walden exclusively into semiconductors?
I know this area best. I have seldom lost dollars. It’s almost entirely a positive record of investment for us. In the past we were in areas like life sciences, but I don’t know biotech. And our LPs have told me, Invest where you can add value.’ So I am avoiding areas I don’t know. We did a few Internet deals, like Sina.com, the largest portal in China, and LookSmart. We made 100x on those deals, but those were really just lucky. I don’t know that area. We invest in semiconductors because we have good friends in the areas that I have mentioned, like Dado [Banatao, at Tallwood Venture Capital], Mark Stevens at Sequoia, Irwin Federman at USVP, Rob Chaplinsky at MDV, Andy Rappaport at August Capital, Vijay Parikh at Global Catalyst, Steve Dominek at Sevin Rosen, Tony Sun at Venrock.
Who are your most frequent collaborators in deals?
The people I already mentioned.
Who are your biggest competitors for deals?
That is what this industry is like. Look, I have two deals I’m doing with Mark Stevens at Sequoia. I have three deals in competition with Mark Stevens at Sequoia. It’s the same with Venrock and MDV. Unfortunately that is part of the industry. Sometimes you end up on the opposite side of the fence from people that you are working closely with in other areas.
How much money have you made at Walden through semiconductors?
A Oh, I’ve never added it up. We’ve never calculated that. It has to be in the hundreds of millions though, conservatively.
So overall, 5x to 10x on capital invested to date?
We have a lot of deals that have returned 10x, lots that have returned 3x to 5x, and a few at 2x. We had four semiconductor IPOs in the past year alone, and one still in registration: Ikanos Communications.
Where do you see that going in the future?
The future-with God’s blessings, more good returns, continuing to be lucky, and being in the right place at the right time-will be as good. I am working very hard now. My responsibilities are in line with semiconductors and China.
How long before the investments made in 04 start to pay off in IPOs or M&A?
Well let’s walk through it, although it’s typically five years to a return: one to one and a half years to tape out, one year to qualify your products with customers, and one year to ramp sales. Then you have an IPO or exit in years four to five. But the days are gone of the $1 billion sale of a VC startup to Intel or Broadcom. Now its $100 million or $200 million.
What’s the future of semiconductors in China?
China is the future for three reasons. Five years from now China will consume 25% of global consumer electronics sales. Look, they’re already No. 1 in consumption of televisions, No. 1 in cell phones and No. 1 in DVDs. Then you start looking at other areas, like white goods (refrigerators, household appliances) and automobiles. Each of China’s three major consumer electronics companies already consumes $400 million apiece of semiconductor products. Today 90% of those [chips] are from the United States and Europe. In five years it will be 60% of those semiconductors coming from China and 40% from elsewhere. Second. China is producing more engineers than the United States. That’s two to three times the United States. Their talent pool is enormous. Add to that all of the returning engineers from Intel, Cisco, TI, Marvell and it’s very powerful. Third, the government is very supportive of semiconductors and provides incentives and grants for companies working in the field.
Will China ever overtake U.S. in developing new semiconductor technology?
Not soon. They’re still focused on lower-end semiconductors, but over time the gap will narrow. It’s like what happened in Taiwan 20 years ago when the foundry business started. Today they dominate the business. So maybe after 20 to 30 years.
Last year you told VCJ that you were willing to help other VCs enter China. Were there any takers?
And then some. Yes, today all of the top VCs have a China strategy and several subsequently came to us for deals. We took NEA to China, for example. We’ve done deals in China with Bay Partners, Interwest, Lightspeed. It’s a very long list [of VCs] in which we’re helping to do their deal or their first deal in China. And I’m happy to learn by working with these best firms in the business. I try hard to learn from these guys who have real depth of experience in our industry.
Does the current competition between competing stock markets (LSE, HKE, NASDAQ and the NYSE) have any impact on your work?
Yes. And while we’ve had IPOs on the NYSE and on NASDAQ, we’ve also had IPOs in Singapore, Malaysia, Hong Kong and Taiwan.
How long before China becomes a viable and competing stock market for VC-backed IPOs?
We think that the Shenzen and Shanghai exchanges are going to be good exchanges for Chinese domestic investors over the next five years.
Any final thoughts from “Mr. Chips,” the world’s most active semiconductor investor?
I’m lucky. I like what I do. I know that I am in a field of giants, like the Sequoias and Global Catalysts, with people who are really First Class in their fields. So I am humble about being able to work with people like them. I’m in the right place at the right time, as the market is changing and coming to me. I’m hopeful that I can apply my 20 years of experience and that I can take advantage of the wave of semiconductors and China as it arrives, just at the right time. I am in China once a month trying to add to my skill sets. We’ll know in five years if I have made the right bets. But I am also enjoying the process.
Check out the print edition of VCJ for plenty of semiconductor data charts.