Buno Pati, Numerical Technologies President, CEO

-- Semiconductor International, 7/1/2001

Buno Pati (Source: Numerical Technologies)

Y.C. "Buno" Pati has served as president, CEO and a director since he co-founded Numerical Technologies in 1995. He has been an assistant professor of electrical engineering and computer science at Harvard University, and has conducted research in computational and system sciences applied to IC manufacturing at Stanford University. Pati has published numerous articles on signal processing, communications, fast lithography simulations and automated phase-shifting mask design. He received his bachelor's and master's degrees, and a Ph.D. in electrical engineering from the University of Maryland.

SI: You've had a record 2000Q4, a 350% increase, and ended the year with $23M plus. What else is going on?

Pati: We're still growing in the revenues and earnings front. We just reported the quarter, and it shows continued growth during a down cycle.

SI: What do you attribute this to?

Pati: I see it as a reflection of the fact that the industry can be divided into two parts — a capacity-driven part and a technology-driven part. Any time there's a slowdown in the capacity-driven part, everyone scrambles and refocuses their attention on technology. This is why everyone is quick to adopt new technologies during a down cycle; it's what will get them out of it and up again.

SI: So, no retrenchment for Numerical?

Pati: No, we haven't reset our guidance; we continue to beat our numbers.

SI: Is there anything you're planning to change in the company?

Pati: No, the company is still relatively new. There are two parts to our business. The business itself is centered around licensing our patented process technology to manufacturers — and we license it for royalties paid either on a per-fab-per-year basis, or a per-unit basis. Since the IP model by definition is one where you get paid as the technology is used, in order to drive the use of it we have a second part of our business, an infrastructure business that supports and facilitates the use of the technology. Thus, the growth and revenues you're seeing are primarily coming from the infrastructure business, which has to be in place to drive the long-term growth through the use of IP.

SI: Can you describe the structure of your business?

Pati: We play in four infrastructure segments—the design sector, manufacturing data preparation, the photo- mask area and the semiconductor equipment space. In each of these, we attach certain components to ensure that the fab technologies that we're implementing are supported. Coming out of design you want to make sure your design complies with these new design technologies. Going through data preparation you want to handle this properly, so that the manufacturing data adequately supports it. The photomask must be made correctly and, lastly, the equipment must support specific things such as inspection and metrology and so forth.

SI: So your IP is your main focus?

Pati: Yes. Our strategy is that our long-term growth is derived from the use of our IP, and to facilitate its use, we supply the infrastructure. We're unique in that the infrastructure component is one we get paid for. Most IP companies have to build similar infrastructures, yet they rarely get paid for — it is something that they either give away or is paid for by other companies.

SI: You're in an excellent position in that you essentially have no competitors.

Pati: True, there are no competitors for patents. Today, ours is the only commercially proven technology that allows 0.10 µ m and below manufacturing.

SI: What do you see as your next growth areas — how are you planning for this?

Pati: Our driver is adoption — adoption of our process technology by the fabs. As I mentioned, we're still in a relatively early phase, and have several of the leaders adopting our technology right now, and the infrastructure is going into place. Over the next two to three years, we' continue to drive adoption as it enters mainstream with 0.13 µ m generation and beyond, and continue to build infrastructure that standardizes the industry on this technology.

SI: So for that period of time your end users should expect more of the same?

Pati: Absolutely! Our growth driver is adoption. (Laughing) Technology's great, but if no one uses it, it doesn't do any good! Our focus is on getting the technology adopted and making it as standardized as possible.

SI: What about three years from now?

Pati: For us that is the sweet spot. Let me tell you why. We started out of Stanford University as a research group. When you're in a university, you have the luxury of being able to look forward and try to figure out what'll happen in 10 years. We spent about three years working on this at Stanford, and decided it was something that we'd like to be able to bet on. In a sense, we were anticipating the subwavelength gap where feature sizes on silicon dive well below the wavelength of optical lithography tools.

SI: So you were predicting the future?

Pati: In a sense we were. Had you done a survey then, the answer you would have received is that nobody needs the technology, nobody expects to need the technology and, in fact, they do not want to need the technology, because they expected equipment solutions to keep driving the shrinks. We bet everything on this gap opening up. In 1999, the gap opened up with the coming of the 0.18 µ m generation.

SI: And now?

Pati: If you look at what's going on, for anyone doing 0.13 µ m with high-performance or low-power devices, they need small transistors, so phase shifting automatically becomes part of the roadmap. Looking forward to the 0.10 µ m generation, its transistors are 50 nm. So what can you do, even if you do have 193, which isn't available today? There just is no alternative to our technology.

SI: What about non-optical technologies?

Pati: The so-called "non-optical" next-generation lithography technologies are all on the horizon and, with few exceptions, are fundamentally good. However, they require you to rip out a 30-year-old infrastructure supporting optical and replace it with something else — an extremely difficult and expensive task to accomplish in this industry. It'll be done when there is no choice — in other words, if optical can be made to produce these smaller feature sizes, then it will be. People will bend backward in order not to change, which, incidentally, is what fought us on the way up — "Don't want it, don't need it, hope we never need it," until they did.

SI: You mentioned providing infrastructure as an important part of Numerical's business. Tell us how you work with your customers, the fabs.

Pati: Our production licensees include Intel, Motorola, TI and UMC. We' "For my 0.13 µ m generation I need feature sizes of 70 nm, I need a certain amount of CD control to get the desired yield," and so on. So we essentially help develop the process. We have considerable lithography expertise and we apply it. Once we have completed that stage we license the technology for production. So there is an R&D engagement and there is a production engagement. When it reaches production stage, we license the IP for production and help them implement a transfer into production, including the flow of data. In the meantime (and here is where the infrastructure comes in), we're building up the infrastructure. For example, KLA, Applied Materials, Zygo, Lasertec — every inspection tool they sell has embedded in it our software, which resides in the machine and handles data and, in particular, subwavelength design data to accommodate the process flow. Our Virtual Stepper sits on these companies' inspection tools and enables these tools to look at what has been flagged as a defect and conclude whether the defect will print on silicon or not. It's that type of activity. We do something similar in the photomask area.

SI: What is the next step, design?

Pati: Yes. However, here our only interest is to ensure that the designs that are coming out are phase-compliant, ready for the phase-shift and subwavelength processes. We have no intention of building design tools any more than we want to build photomasks or equipment. But what we need to add is some intelligence in the way of software that's able to participate in the design activity and ensure that the design is compliant with these new processes. Again, every physical design tool that Cadence sells — whether custom layout, place-and-route, or physical verification — has our embedded software.

Once you have the design flow, you must feed those design flows with the building blocks: the cell libraries. These must also be phase-compliant and characterized for these subwavelength technologies. We acquired a company called Cadabra that provides software tools to automatically generate cell library layouts given process information and functional requirements. At that point, you are looking at a situation where cell libraries can be created in a phase-compliant manner; they can be fed into these design flows that support phase shifting and OPC.

SI: Do you maintain FAEs at the fabs, or is everything handled internally by the fabs?

Pati: We have personnel at the fabs because we're still introducing new technology. In many of these settings we're the ones with the experience in this portion of the process. We have a number of resident engineers who spend considerable time at the fabs and actually run the steppers and the metrology tools, get the measurements and take the customer to the point at which they understand the process.

SI: You mentioned you are always looking for new technology — anything coming up on the horizon?

Pati: Well, nothing that I can mention right now (grinning), but I can talk a little bit around it. Processes take a long time to develop. To be part of a process generation, you must begin two to three years in advance. That is what put us in our current position — we were able to start well in advance and get our technology baselined into these processes for the 0.13 µ m generation. Looking at 0.10 and 0.07 µ m, the requirements are somewhat different. Until eight months ago, everyone thought that they'd have 193 available for 0.13 µ m — it's not happening and there was concern. Fortunately, we had a solution. If you consider 0.13 µ m critical features, you're primarily looking at the transistors, because they go down to 0.07 µ m or so. At 0.10 µ m, the chip's whole density changes, so now you need a more aggressive version of that same technology, because many features that weren't critical before are now, from the wavelength's perspective.

Much of what we're doing now is focused on what we call a full shrink, which is taking an entire design and transferring it to 0.10 or 0.07 µm, using available optical lithography tools. There has been a lot of hoopla about EUV, and it's so far out — it's good technology, but it's at the stage at which X-ray seems to have remained indefinitely — forever five years out.

SI: Looking at a time period of two years from now, what should we be preparing for right now?

Pati: Everybody involved in semiconductor manufacturing better forget the idea of DUV. Some talk about 2003, but I don't believe it'll happen.

SI: What are your thoughts on the consolidation frenzy that is taking place?

Pati: In my opinion, any growth that stems from consolidations is artificial. It also stifles innovation. In times such as this, when we are in a crisis mode, you need a lot of innovation to successfully ride it through. Large-scale consolidation is the wrong thing to do, yet it seems to be the tendency. We' perceived, artificial.

SI: What industry trends do you believe we should be paying more attention to?

Pati: Feature size reduction. This may sound like nothing new, but the industry depends on it to continue — it has fueled our growth for the last 30 years. A trend I would wish someone in the roadmap committee would notice is that Moore's Law executes in two-, not three-year periods.

SI: Does this mean you believe the roadmaps are inaccurate?

Pati: I don't believe in roadmaps because what I see at customers — and we're very close to their internal roadmaps — is completely different than what you see in published roadmaps.

SI: It seems like only yesterday we came out of a bad downturn and now we're in the middle of another. Is there anything we should do, as an industry, to minimize these cycles?

Pati: No. We're inherently subject to these cycles, which are driven by macro They turn inward and once again focus on technology. I believe companies do the right things automatically during these times — the ones that survive.

SI: Is there anything that you'd like to see changed in the industry?

Pati: One of the industry's endemic problems is inertia — a reluctance to change. In good times, inertia keeps the good times going as long as possible. This isn't a situation that lends itself well to dramatic changes. As a company, we've had to overcome these issues. For example, why was it that it wasn't until some eight months ago that everyone realized that 193 wasn't going to be available for 0.13 µ m? The signs were obvious, but there was a disconnect; it is as if people were looking but not seeing. This whole time could have been spent focusing on how to better use 248.

SI: Any acquisitions looming?

Pati: No. At present we're not looking at anything. We're focusing on our business's key drivers. We've pretty much completed the acquisitions we've needed for our purposes.

SI: Describe Numerical five years from now.

Pati: We expect to be the dominant provider of technology for smaller circuit features. We expect phase-shift and other attendant technologies to be the industry standard. I believe we've done a good job of setting ourselves up to be in that position, with the major players adopting the technology — 95% of all inspection tools have our software embedded in them; for data preparation, virtually every chip in the world goes through us. We're in a great position to establish a standard.

SI: Any advice to your peers?

Pati: There's an increasing need for companies to work together. It's no longer possible for everybody to live in his little corner, with blinders on, and say, "I'm going to do my job and hope everyone else does his." As the industry has continued to disaggregate, relationships between the different sectors have broken down. It' You work on a new engine and, when you're done, you realize the transmission isn't there. Right now, more so than in the last decades, partnerships are required to make everything work.

— Alexander E. Braun