Experts At The Table: MEMS Challenges

Semiconductor Engineering sat down to discuss the challenges of MEMS with Rakesh Kumar, senior director of the MEMS program at GlobalFoundries; Tak Tanaka, managing director for Applied Global Services at Applied Materials; Paul Lindner, executive technology director at EV Group; and Alissa M. Fitzgerald, founder and managing member at A.M. Fitzgerald & Associates. What follows are excerpts of that conversation.

SE: At present, the MEMS IDMs generally outsource little or none of their production to the foundries. Will that ever change?

Kumar: Right now, more than 70% of the MEMS business is still in the hands of the IDMs. Once again, you have to look at the overall MEMS business. MEMS means microelectromechanical systems. So MEMS is considered a system, not a component. When that word systems comes in, that means you have to provide a complete solution. The IDMs have been successful because they provide a complete solution, such as the design, process, packaging and test. They also enable the applications from a systems point of view. So the IDMs have all of the internal capabilities to provide a systems-level solution. And in terms of MEMS processes, this is also a core competency of the IDMs. But now, I see where MEMS is moving towards another direction, where the foundries can provide at least a processing solution for customers. And at the same time, the fabless companies are no longer just design houses. Fabless companies, such as Knowles, InvenSense and others, are basically systems houses. They operate like an IDM, but they outsource the fabrication part to the foundries. As a foundry, we want to enable fabless companies. But will IDMs outsource to the foundries? Some of the successful IDMs already have a huge amount of fab capacity in-house. In terms of outsourcing, I don’t think they will duplicate those efforts unless they change their CapEx plans in the future. For the foundries, this is not very attractive to get into that type of business in the first place. Generally, the foundries are looking to enable mostly fabless companies with new technologies. So, by offering more efficient platforms for fabless companies, we can grow faster, as compared to being dependent on some spill over from the IDMs. On the other hand, there are some IDMs that have 6-inch fabs and smaller. Those people are considering investments other than new fabs. Those IDMs may consider foundries, rather than building a new fab. Foundries like GlobalFoundries already have 8-inch fabs for MEMS production.

Lindner: The biggest opportunity for MEMS lies in the foundries. The foundries have the possibility to offer standard processes. But I would agree with Rakesh’s point of view about IDMs. The likelihood that an IDM transfers its proprietary process to a foundry to expand the IDM’s capacity is probably small. The IDMs have the ability to expand their own capacity. At the same time, the foundries, which have higher growth rates, are successful with the fabless companies. My prediction is that outsourcing—that is, the foundry business versus the IDM business–will be more balanced in the future. In addition, there is also the concept of fab lite. There are many customers out there who have pilot line or R&D line capabilities. They may come up with a device that they want to manufacture, but they don’t want to invest in a new fab or line. We support those customers with our equipment in a semi-automated fashion with the same process recipes. So, the ability to transfer those processes to a foundry is possible and easy. This is because it’s the same process module.

Tanaka: Some IDM customers are considering growing their own fab base in MEMS. Many IDM customers won’t outsource. They are already doing a great job. As far as we can see in the next few years, we don’t expect significant changes in that. And there are also a lot of older 150mm and 200mm fabs out there, many of which are not being used for MEMS. Certain customers may see the opportunity to convert existing capacity into MEMS and provide a manufacturing solution.

SE: What about the fabless-foundry model in MEMS?

Fitzgerald: The fabless model for MEMS is in its infancy and is pretty weak. If you look at Yole’s list of the top 30 MEMS rankings, only two are fabless companies. The rest are captive IDMs. There are a few companies doing fab lite. They have their own fab, but they also have a second source to another facility. I would say that’s evidence of the fact that we don’t have a strong fabless model. Now look at the famous Jerry Sanders’ quote: ‘Real men have fabs.’ No one believes that anymore. It’s a new age and no one can afford fabs. Certainly, a few people can afford to build semiconductor fabs. On the other hand, MEMS fabs are much less expensive. But the problem is the investor money is not there. No investor is going to fund and build a new MEMS fab today. Almost by default of the economic and investing climate, the new MEMS companies will be fabless. But they still face some significant barriers, because of the lack of process-specific information in the foundries. This, in turn, is because of the lack of full simulation capabilities. This, by the way, is not the fault of the EDA vendors. The EDA vendors know how to execute these kinds of simulations and put these capabilities in their software. The problem is the data is missing. That is, the data for the processes. I think everyone sees that the fabless model is the way of the future. It has to succeed. But the question is how and who will pay for that? So, that’s why there is a lot of interest in standardization. A lot of foundries are interested in how they can attract more fabless customers. So, for example, in the last year or two, we’ve seen several foundries offer inertial sensor platforms, such ST, Dalsa and X-Fab. I see this as a desire to reach out to fabless customers and offer them something more. But these are very early efforts and there is still a lot of work ahead.

SE: Is China a factor in MEMS?

Tanaka: We may want to watch China. We have a lot of inquiries from customers in China for MEMS. China is not a driver yet, but it is something we are watching. The same thing happened in LEDs. Four years ago, a lot of Chinese companies entered the business. Many of them would say: ‘I have the equipment and technology.’

SE: Where is MEMS going in the future?

Tanaka: We will see the integration of MEMS on other devices. But definitely, we will see more CMOS and MEMS integration. That, of course, will give us more challenges from an equipment point of view. I think we will see a lot more opportunities in the consumer and medical areas for MEMS. For example, there is bioMEMS. There is also MEMS for the energy and environmental markets. In addition, we will see double-digit growth on a CAGR basis. That will continue for the next few years. So that’s why Applied Materials is interested and seeking opportunities in this industry.

Fitzgerald: If you look around, we are using innovative devices in areas like medical. You see pacemakers, insulin pumps and quality-of-life devices, such as artificial joints and knees. Near term, the exciting MEMS applications are medical devices for life saving and quality-of-life. But this is a very fragmented market, because the technology is very specific to the procedure, organ or surgical instrument being used. The challenge for the foundries will be to capture that business without having to have 10,000 different customers who all want something different. I am also bullish on infrared technology. The technology is established. But if it could be less expensive, we would see an explosion in volume for lower cost applications. Right now, these chips are in the hundreds of dollars. They go into very high-end security cameras and scientific instruments. If we could see the costs go down, we could see them in automobiles and personal electronics.
Lindner: If you focus on manufacturing technology, I see more 3D integration for MEMS coming, driven by low-power requirements. I see where the industry will maintain the same form factor, but customers will integrate more functions on the same device. Potentially, we will see MEMS being integrated with more advanced CMOS nodes, which could bring along the migration to 300mm. This is because the very advanced nodes are not available at 200mm. I don’t see that next year. I would guess within five years we will have 300mm MEMS fabs, at least in pilot lines.

Kumar: We will see the integration of MEMS with ASICs, plus microcontrollers and maybe wireless capabilities. It is also possible to see more integration of CMOS and MEMS, but this will be for specific applications. It’s not for all kinds of devices. From a process point of view, you can do the CMOS-MEMS integration at the chip level or the package level. At the fab level, the benefits of the monolithic integration of MEMS with CMOS are increased levels of integration, low power, high speed and the reduction of parasitics. It is particularly suitable for making an array of sensors. It offers the ability to reduce the form factor and cost. The trade-off of CMOS-MEMS integration is the increasing complexity of fabrication. There is a reduced thermal budget for fabrication of MEMS-on-CMOS. For the most part, package-level integration offers greater flexibility. This allows the designer to optimize and integrate MEMS and CMOS separately using the most suitable technologies for each device.

To view part one of this roundtable discussion, click here.
To view part two of this roundtable discussion, click here.