Follow the Money
Betting on Moore's Law
Porsche 911s, BMW 850s, Mercedes 500SELs plenty of hot cars jam the freeways of Silicon Valley, and the good fortunes of many of their drivers can be traced to Moore's Law. According to this law, the amount of information storable on a silicon chip doubles every 18 months while the cost of making those chips drops by 50 percent. These happy economics continue to fuel a long-running technology boom that has made a lot of shareholders extremely rich.
Although there's some debate whether Moore's Law can hold up in the long run, expect the law to remain in force for another decade, at least. The manufacturing dynamics that drive Moore's Law are changing, however, and investors would be wise to change the way they evaluate the semiconductor industry.
A fab decade
For the past 10 years, factors such as growth in wafer size and manufacturing yield have been the key to the semiconductor industry's stupendous growth. But as these yields and wafer gains begin to taper off, manufacturing throughput will become more important (see How the Chips Fall). That puts the spotlight on chip-making equipment manufacturers such as Genus Inc. (GGNS), Plasma & Materials Technology (PMAT), and, especially, Mattson Technology (MTSN).
Here's why: Historically, chipmaking costs have fallen as a result of the constant decline in feature size (the size of individual circuit lines and transistors) and the constant increase in wafer size (the platter of silicon on which chips are made).
Average yields or the percentage of good chips per wafer tripled over the last decade as equipment improved and engineers concocted better recipes. But today's yields are already at around 85 percent, and you just can't do better than 100 percent. So the gain in yield for the future is naturally constrained.
Feature size decreased by a factor of 4 and will continue to shrink at that rate. But wafer size, which grew by a factor of 4 over the last 10 years, will only increase 2.2 times over the same period: when wafers get larger, it's harder to keep yields up. Factory utilization improved 50 percent over the last decade, but you can't work more than three shifts and equipment can't have more than 100 percent uptime. Add these factors up and, since the mid-'80s, the production process has experienced a whopping 7,200 percent boost in productivity.
However, these gains came at the cost of a decline in factory throughput, as manufacturers were forced to move from batch processing of wafers to single-wafer systems to maintain yields. Declining throughput meant net productivity was up "only" 1,800 percent.
The throughput thesis
So, over the next decade, improving productivity will depend on improving throughput. Fortunately, new technologies have been developed that can share a single set of robotics while processing multiple wafers. And those technologies are in the blueprints of more than 90 factories scheduled for construction over the next three years.
Given these manufacturing improvements, I expect throughput to reverse its decline and grow five times over the next decade enough to make up for the tailoff in yield and wafer size.
Net productivity should rise 1,900 percent over the next decade at about the same rate as the last one guaranteeing that the cost of semiconductors will continue to fall 50 percent every 18 months and that the world will continue to wire itself at levels sufficient to keep the Valley's highways jammed with luxury automobiles.
Make it Mattson
What's the best bet for taking advantage of the new dynamics pushing Moore's Law? I am putting my money on Mattson Technology (MTSN), a semiconductor equipment manufacturer based in Fremont, California. The company's Aspen architecture for photoresistant stripping, rapid thermal processing, and chemical vapor deposition can process four to eight wafers simultaneously, giving the process control advantages of a single wafer system while providing the speed of a batch system. In addition, the system sports an inner load lock that allows wafers to be swapped in and out without having to pump down to vacuum pressure, thereby reducing wafer handling time.
Mattson is growing more than 100 percent a year, and I estimate earnings at US$1.10 a share in 1996 and $1.50 in 1997. I would buy the stock for up to $15 a share with a target of $30. Background information is available on the Mattson Web site at www.mattson.com.
TWIT$
I'm covering the Netcom short; with the entry of AT&T as an Internet service provider, the bad news is out on the profitability of ISPs. I'm also selling Sun Microsystems stocks at a huge profit it looks like Microsoft NT and Unix are about to go to war, and I've found pacifism is a good strategy when investing. I used the proceeds from the Sun sale to buy 30,000 shares of Mattson Technology.
The Wired Interactive Technology Fund
|
| Company | Primary | Symbol | Shares | Price Apr 1 | Since Mar 15 | Action
| Netcom Online Communications Service Inc. | Internet | NETC | 5,000 | 25 1/4 | + 3/4 | cover
| Sun Microsystems | Hw/sw | SUNW | 3,000 | 45 3/4 | 3 1/4 | sell
| LSI Logic Corporation | Semiconductors | LSI | 7,800 | 19 1/8 | + 1 1/8 | hold
| Applied Materials Inc. | Semiconductor equip. | AMAT | 4,000 | 35 5/8 | + 7/8 | hold
| The Walt Disney Company | Entertainment | DIS | <1,500 | 64 7/8
| Apple Computer Company | Hw/sw | AAPL | 4,800 | 25 1/2 | 1 3/8 | hold
| Tele-Communications Inc. | Cable television | TCOMA | 4,800 | 18 5/8 | 2 3/8 | hold
| Intel Corporation | Microchips | TCOMA | 3,000 | 57 1/8 | + 1 1/16 | hold
| Adobe Systems Inc. | Software | ADBE | 5,000 | 32<7 8 | + 1/4 | hold
| Mattson Technology | Semiconductor equip. | MTSN | 30,000 | 13 5/8 | buy
| Portfolio Value | $1,739,356.25 | (+73.94% | .94%
How the Chips Fall
| Equipment Productivity
| Last 10 Years
| Next 10 Years
| Yield
| + 3.0x
| + 1.1x
| Feature Size
| + 4.0x
| + 4.0x
| Wafer Size
| + 4.0x
| + 2.2x
| Utilization
| + 1.5x
| + 1.1x
| Total Gain
| + 72.0x
| + 11.0x
| Capital Productivity
| Equipment Cost
| 3.0x
| 3.0x
| Equipment Throughput
| 1.3x
| + 5.0x
| Total Loss/Gain
| 4.0x
| + 1.7x
| Net Productivity
| + 18.0x
| + 19.0x
The TWIT$ fund is a model established by *Wired*, not an official traded portfolio. *Wired* readers who use this information for investment decisions do so at their own risk.
*Michael Murphy is a money magager who publishes the California Technology Stock Letter in Half Moon Bay, California.*
