Tougher choices in capital equipment
By Ed Sperling, Editor in Chief -- Electronic News, 8/17/2007
Rick Hill, chairman and CEO of Novellus, sat down with Electronic News/Electronic Business to talk about what’s changing in capital equipment, why there won’t be any new competitors and why it may cost more for companies to develop manufacturing processes. What follows are excerpts of that discussion.
edn.com
Q: How’s your business?
Hill: Most people would classify this as a trough.
Q: Does it look like part of a normal cycle?
Hill: It’s a normal cycle, given the buying patterns today. Because of the phasing in of fabs, instead of putting in large fabs all at one time, the quantums of purchases are less. The peaks are lower and the valleys are higher. That continues to be the case.
Q: Is memory dominating purchases of capital equipment?
Hill: Memory is big right now. But memory has anticipated Vista, and Vista has not materialized. That is causing somewhat of an oversupply of DRAM. What you’d like to see is lower power consumption, faster turn-on speed and lighter weight. That’s the fundamental PC market, and those products are just starting to hit the shelves.
Q: Has the market for Novellus’ equipment changed from a geographical standpoint?
Hill: It’s still mainly based in China, Taiwan, Japan, Korea and Singapore.
Q: We have other businesses starting up, though, such as solar cells, which are not conducive to shipping long distances?
Hill: We don’t see any shift from a classic semiconductor standpoint. To me, the solar cell business feels like a fad business.
Q: Isn’t there normally a bubble followed by a real market?
Hill: This feels like 1976.
Q: What happened in 1976?
Hill: That was when we had the first oil embargo and what then was an astronomically high price of $45 a barrel. Many initiatives for solar energy were started, only to fizzle out because of the economics. I don’t think the economics have changed. Efficiencies and material costs haven’t changed.
Q: How about other market changes? Is it all geared toward memory, or is logic becoming a big consumer of equipment, as well?
Hill: Logic is the shift away from IDMs (integrated device manufacturers) to foundries. It’s causing new capacity to come on line for the foundries, which is a net positive for the equipment business. In the long term, though, it might be a net negative. You’d like to have as many buyers as possible filling their capacity to 50 percent. That means they buy twice the amount of equipment the world needs to produce those devices. As you shrink into foundries, they’re able to operate at higher levels of capacity because they agglomerate demand, which means they buy less equipment.
Q: You’re referring to the digital IDMs, right, versus the analog manufacturers?
Hill: That’s right. They (analog chip makers) are not the ones who buy new equipment. They buy used equipment. They use the tool that they need to use for the business that they have. They’re very successful companies, but they haven’t found a need to go to 300mm wafers.
Q: Does it make it harder to sell to a limited number of customers?
Hill: There’s always that pressure.
Q: Then is there a way to differentiate between capital equipment makers to command a better price?
Hill: There has to be a way to differentiate yourself. Clearly, Novellus has both technology and productivity, and that’s a differentiator.
Q: But is there enough room for differentiation so that one capital equipment maker is not competing head-to-head with another?
Hill: [Applied Materials] is going into the solar business and Novellus is going into the industrial business. Those are clear opportunities. Applied has largely turned into a service company, not a technology company, in the semiconductor business. They’ve largely written it off as a growth business. They’re looking for new growth businesses. We don’t think that’s the case. We think there’s still growth, but you have to go find it.
Q: Is the growth opportunity getting more fragmented?
Hill: It’s a little more fragmented. Some segments will grow faster than others.
Q: Such as?
Hill: Clearly, the NAND flash market is going to grow faster. There are some unique demands in that industry you have to address. It is very price sensitive, so you have to have productivity. At the same time, it’s probably the most technically advanced business. If you don’t have the technology and the productivity you can’t take it to the next level. But I think it’s going to be a big opportunity for years to come.
Q: How about DRAM?
Hill: The DRAM business is a cost-driven business. But it’s also shifting to copper. That’s a niche that, given our product positioning, bodes well for us.
Q: Is there a big opportunity at the front end of Moore’s Law?
Hill: When you get down to 22 nanometers, you have 220 Angstroms. When you figure the average atom is 5 Angstroms, that’s roughly 40 atomic layers. You don’t have much material to work with. I think getting to 22 nanometers is a tough path.
Q: Will that still be top-down manufacturing or will it involve self-assembly?
Hill: Self assembly is a nice concept, but making order out of disorder is a tough task. Typically you engineer that. You don’t hope it statistically occurs. There are more self-aligned processes coming out to take out the need for machinery at that level, but this business isn’t going to change markedly from the standpoint of the process technology.
Q: Does Novellus work with the nanotechnology research lab in Albany and IMEC?
Hill: We do. We have no equipment at Albany. We work with IMEC on some copper initiatives. It’s a long way from the university to the real world. They have great ideas, but they don’t have to make money. They write papers.
Q: But it’s still a good idea to do that research, isn’t it?
Hill: No question. And that’s why we fund it.
Q: You mentioned copper in memory. Is that going to be as much of a problem as it was for other chips?
Hill: No, the kinks are out.
Q: You talked in the past about the problems of low-k, 300mm and copper coming all at once. Is that all solved?
Hill: Low-k isn’t. That’s still a fundamental tradeoff between structural integrity versus dielectric content.
Q: What’s the big issue in manufacturing?
Hill: High-k gate is the biggest gain, and hafnium oxide is half the way there. The reality is that gate to drain leakage is a large power consumption factor in semiconductors, and power is probably the number one issue as you shrink. Advances in high k would be well rewarded. As you make the lines smaller, you’ve got exponential growth in resistivity. It’s not because of the material. It gets dominated by sidewall scattering.
Q: The problem is pushing electrons through a narrower and narrower tube, right?
Hill: Yes. They run up against the wall, hit little cracks, and don’t go in a straight line. That’s a big issue.
Q: What are your customers asking for that Novellus has to get involved in?
Hill: The industry has changed a lot to the point where a semiconductor company uses a lot of equipment companies to research areas they’re not sure of. They tell one vendor to research this and another to research something else, then they go off and pick and choose. We’ve got to be aware of all the possibilities, and only sign up for the one we think we can with. That’s the toughest challenge.
Q: Is it getting tougher to place your bet on the right horse?
Hill: It’s very difficult.
Q: It sounds, then, like one of your big issues is how to allocate R&D budgets.
Hill: That’s a big issue. It’s a whole new wrinkle in the equipment business.
Q: So what’s the solution? Do you work with partners or try to do it alone?
Hill: It’s a combination. Internally, you have to have a lot of really smart people who look at all the permutations and have arguments about the pros and cons of each, both from standpoint of the technical likelihood of their success and the economic likelihood of their success.
Q: So now it sounds like your R&D decision is no longer purely technology driven, right?
Hill: It’s very much economics. It’s a big pressure point on equipment companies. If a Samsung or Intel comes to you and says they need a new gate oxide, and you have to design a totally new reactor to be able to put down that film, that could be a $20 million to $50 million investment. And you may not get it back, because the amount of equipment required in the fab when it’s only one layer and very thin might be two units. When you make a $50 million investment and you’re only going to sell two to each fab, that equipment will have to be very expensive to pay for the material it takes to make it and get a return on the $50 million I had to put up front to do the engineering to build it.
Q: How does it affect your business model? It sounds like it’s a top-down decision, rather than an engineering-driven, bottom-up model.
Hill: It is, and it’s something we’ve been criticized for in the past. They say we don’t have R&D going into all these areas. We haven’t because we figured out that economically we can’t afford it. We pick our areas of focus. Copper interconnect is one. It turned out to be right. The number of layers kept increasing and it all went to copper. And it went to damascene, so we focused on low-k dielectrics. We have a regimented process to determine what’s likely. We pulled out early from the high-k gate tantalum oxide and barium strontium titanate. These are materials that you heard about in a Superman comic book. The reality is that putting those chemicals down requires totally new reactors. Then you didn’t even know which one they were going to pick. So we chose not to work on this.
Q: Does the customer contribute to the research?
Hill: I think that will be the next step going forward. If the customer wants it, they’ll have to pay for the research.
Q: Have you been able to get money for that?
Hill: Not up until this point. That only happened in the early days. This is back to the future. Economics will drive it. It’s like the decision to go to 450nm (wafers). No one wants to go there on their own nickel.
Q: Because of all this, does your research cost change as a percentage of revenue?
Hill: No. You’ve got a business model you stick to in order to attract capital. You have a fixed amount for research and development. If you spend beyond your model, you destroy economic value, and then it’s only a matter of time before you can’t spend it anymore.
Q: Let’s swap topics. Has the time that it takes for your customers to order equipment and get that up and running changed?
Hill: It’s shortened dramatically. If we went back 14 years, it was nine months to a year. Today it’s two to six months.
Q: What does that do to your inventory?
Hill: As long as you can increase your cycle time, it doesn’t do anything to your inventory. But historically, what you’ve seen is that inventory has stayed flat. Given the tools that we have, it should have gone down more than it has. You have to take some risk.
Q: What does that do to your cost structure?
Hill: It drives our investment up, which forces us to get a higher return on that investment.
Q: Given all of this, and all of the constraints and shifts in business, will there be any new capital equipment players in the market?
Hill: I think it’s almost impossible for new CapEx vendors to come in, be successful and grow without a major acquisition. You need a market of critical mass. If you’re a small company and you launch a product, if you can’t go global nobody will buy it. But you can’t afford to put the global infrastructure in until after you’ve developed the revenue. It’s a Catch-22. |
