What's in the Cards for Chips
H&Q analyst Rob Chaplinsky tells why the "chipless" microprocessor firms have a bright future and finds something nice to say (sort of) about Intel.
By Eneida Guzman
What's to become of the semiconductor industry? It's stumbling over slowing sales, shrinking margins, intensified competition and fallout from Asia's economic crisis. And share prices are suffocating at several chip makers that have been Silicon Valley leaders for years.
Rob Chaplinsky, semiconductor analyst at high-tech boutique brokerage Hambrecht & Quist, admits that the group's near-term future looks rather bleak. But he sees promise in newer, less-traditional areas for growth in the sector that are just beginning to take shape. Instead of microprocessors, memory chips and graphics chips driving growth, he sees networking and communications applications used in modems, routers, switches and the like as the sector's biggest growth opportunity. He also likes a group of chip stocks that are known as the "semiconductor intellectual property" group, or SIP for short.
In an interview from H&Q's San Francisco headquarters, Chaplinsky told Investor about the latest innovations at the "chipless" semiconductor companies, the difficulties of running a chip company without a fabrication plant to call your own and Intel's business prospects.
The SIP Model
What do you see as the most significant trend in the semiconductor intellectual-property industry right now?
The most interesting trend is that a lot of companies are pursuing the new chipless semiconductor business model because the cost of capital and the requirements to compete in next-generation devices is just escalating. So some companies are pursuing this software-like semiconductor model as opposed to actually distributing and selling chips.
Describe the business model.
The SIP companies basically just design and distribute chip design via software or database. Someone else integrates the design, makes the chip, sells it and distributes it. This party usually pays a licensing and royalty fee to the SIP company. The SIP company is almost like a software company, but instead of selling a CD-ROM, they are selling semiconductor logic.
In this group, I like Rambus (RMBS). It is a high-speed memory interface company for next-generation memory-chip technology. They have unbelievable patents and a very strong royalty-based semiconductor intellectual-property model.
Do you expect this model to be successful?
Yeah, I actually do. I think there are some natural hurdles that will happen because sometimes just connecting all the different parts of a chip together and hoping that it all works is pretty difficult. You have to implement a lot of standards and interfaces to make it work. But it does help companies come to market on next-generation devices faster because they don't have to reinvent every piece of silicon or transistor in there. They can reuse existing, proven blocks of transistors, and that's really what's attractive about it.
The new deep-submicron semiconductor-fabrication process will enable companies to design systems on a single chip. What is the cost savings?
A leading-edge, 5- to 7-million-transistor device takes several years to design, develop, verify and get into production. What the new process does is really shorten the design and layout and verification process by one-third or one-half. It's a significant savings of time and money.
The SIP company is almost like a software company, but instead of selling a CD-ROM, they are selling semiconductor logic.
People are moving from motherboard-level solutions to chip-level solutions, and that's what they're calling a complete system on a chip. In many cases that's where the value of these new intellectual-property kind of semiconductor companies come into play, because designing one big system on a chip is very complex. So if you can reuse proven and existing technology, what they call cores of transistors, you get to market much faster.
Programmable Logic
I understand that you believe the programmable-logic sector is one of the most attractive investment segments of the semiconductor industry. Tell us why.
Programmable logic is where you can actually make adjustments to your chips and tune your chips in the field after manufacturing. The hardest thing about semiconductors is once you actually fabricate, it's basically fixed and you can't change it. However, if you have a programmable-logic device, you actually have some memory and programming capability in the chip to alter its functionality and tune it for a specific application after the actual chip is made. I think this will to continue to be used and become more important as the product cycles continue to accelerate.
Are you concerned about price erosion in this segment?
Yes. It's becoming very competitive. There are a couple of players who have large market share. And everyone is trying to grow market share. You have an environment where these programmable devices are trying to penetrate new markets where typically they are at a price disadvantage because they are offering this programmable premium, if you will. So, when they try to penetrate new markets, they have to be very aggressive in their pricing. That's what happened in '97 to this sector. The companies in this group were over-aggressive in their pricing. They have been trying to manage their pricing models a little more prudently in '98. I think the price erosion we saw in '97 will slow down in '98.
With orders slowing down, economic conditions in Asia stalling market growth and the fact that we are moving towards a seasonally slower period for the group, which programmable-logic names make sense?
I like companies that focus on the networking and telecommunications of semiconductors. The chip companies with exposure there that I'd buy are Altera (ALTR) and Xilinx (XLNX). Those are two good industry leaders in the front of the logic sector.
One of your favorite stocks, DSP Group, is in the telephony space. Its fundamentals are very impressive and the stock is trading at a discount to its growth rate. Tell us about this one.
I have a strong "buy" rating on the stock. DSP Group (DSPG) develops and markets digital-signal speech products targeted at the consumer telephone and computer telephony markets. Over 80% of the company's revenue today is derived from selling speech processors for use in digital telephone-answering devices. It's not very glamorous, but they have a dominant market share there. I especially like the company because of the other 20% of their revenues -- and 40% of their earnings -- are from licensing and royalties they receive from their DSP or digital signal processing business. I think the stock will experience earnings-multiple expansion as royalties ramp to comprise a larger portion of their revenue mix.
Details
--------------------------------------------------------------------------------
Company Report
3-Year Chart
Advisor FYI
Earnings Estimates
Insider Trading
Show Intel
You have lowered your estimates for Intel for calendar-year 1998 and 1999 to $3.09 and $3.62 respectively. They have limited visibility as far as a rebound in orders, and revenues are flat year-to-year. Do you see a turnaround on the horizon?
In the past, Intel (INTC) used design processors for performance and it was a great business model as long as people were buying up in performance and they needed next-generation processing power. But I truly believe the application-software treadmill has really slowed, and Intel's hardware treadmill is exceeding the software treadmill -- meaning applications aren't consuming Intel's MIPS (millions of instructions per second) as fast as they used to. As a result of that, Intel has to now head in a different direction.
Low-cost computer chips must be high on its list of options.
Yes. Clearly, the low-cost computing trend is going to continue to accelerate. So Intel needs to develop products for those segments that are very profitable as well as products for the high segments, and that's a challenge near-term.
Intel sells $25 billion worth of products a year. To sustain its historic growth, you need to start finding $5 billion industries where there are new opportunities for growth. There are not too many $5 billion industries around. As you get bigger, it's very hard to sustain the growth. I think what Craig Barrett will bring to the party as the new CEO is that he will start looking outside the box for new areas of growth.
As far as a turnaround, I think there's still some downward adjustments coming [for Intel].
As far as a turnaround, I think there's still some downward adjustments coming. Some of those downward adjustments are already reflected in the stock price. Investors seem to be waiting for signs that orders will pick up and margins will stop going down and stabilize. As soon as there is evidence of that, a period of growth will follow.
It will be interesting to see if Craig Barrett's leadership can steer the company to the same level of growth that Andy Grove experienced at Intel.
Robert Noyce and Gordon Moore put Intel on the map and Andy Grove put Intel on top of the world. Tough to guess on how Craig Barrett can repeat it. We'll know the answer to that five years from now.
Facing the Coming Challenges
Because of your belief that that the second quarter will be difficult for the semiconductor group, you remain fairly cautious. How much does the Asian economic situation factor into your caution?
Asia is a significant reason to remain cautious on the group's long-term outlook. Initially, people were just attributing the correction to end-market consumption of the region. We're becoming more global, so I think that the impact on our domestic markets and sales is going to be at stake, too.
The growth in the semiconductor capital-equipment area is definitely going to slow down. Asian countries no longer have access to capital like they used to. And we're seeing significant cutbacks in orders, as well as financial estimates of these semiconductor capital-equipment companies all related to Asia. At the same time, with the weak currencies in those countries, the costs for them are significantly lower and as they lower prices on exports that will make U.S. products much less competitive. That will dramatically impact us as well.
Asia is a significant reason to remain cautious on the group's long-term outlook.
In this environment, is it better for chip companies to have fabrication facilities or to be "fabless"?
Three years ago the fabless business model, where you outsourced all your manufacturing, looked pretty attractive; however in the environment we're in, the fabless model for memory companies is pretty difficult because you need much tighter integration between design, manufacturing and logistics. In an environment where you have very large excess supply and inventory levels, you need to balance your needs and be in control of your costs so that you are actually more profitable. It's just basic economies of scale.
The semiconductor industry is experiencing two below-average years of growth. With lower PC sales and falling chip prices, what is going to be driving silicon demand in the next few years?
The PC market is not dead. I'm still a secular bull on the long-term growth of the PC market. Pricing is going to continue to play a role in that industry's growth rate. But PCs will continue to be the dominant consumer of semiconductors well into the next millennium. Steeper growth will come from the communications and networking applications side of the business. Companies that are selling digital-signal products, programmable-logic devices or ASICs -- which stands for application specific integrated circuits -- to the Ciscos (CSCO), Nokias (NOK/A) and Ericssons (ERICY) of the world will be the big growth drivers for semiconductors in the next three to five years.
|
