“The time has come when we have to understand that we can only compete WHEN we collaborate.” (Sudipto Roy, SMIC, at ECTC 2013, May 2013.)
In part 1 of this article, we covered some of the challenges dealing with the business of complex 3D IC supply chains in a conversation with Alan Levine, Director, Wright Williams & Kelly, Inc. Now, in Part 2, we want to look more at putting into practice supply chain business modeling using issues specific to 3D ICs. Let’s start with the tools WWK brings to do the analysis required for 3D IC supply chain modeling.
3D InCites: Tell us about your modeling tools Alan.
Alan: We employ a variety of operational modeling tools. One in particular called Factory Commander® lends itself well to 3D ICs. The tool was initially developed for flat panel displays, which required 2 or 3 distinct process flows with merge points, sometimes in different factories. We were already thinking ‘multi-factory’ when the product was initially developed nearly 20 years ago. We embedded concepts like transforms, which allows a wafer to become individual die, then for multiple die to be merged into a module. We also embedded a lot of flexibility to look at changes; learning curves, process flow changes, etc. We decided early on we’d do the industrial-strength version that handles real world complexities.
3D InCites: What is the level of complexity you can address with your tools?
Alan: The questions we answer are pretty challenging and have a lot of money riding on getting the answer right. The questions being asked of the model may be complex, but it’s not as complex to use Factory Commander as you might think. It’s really well structured. We built it on a relational database, which simplifies the data management a lot. We’ve gotten a significant chunk of our business from companies who tried to develop Excel spreadsheets to do this job. One of our customers said they had descended into ‘Excel hell’. We got them out of there.
3D InCites: Are 3D ICs even more challenging in that regard?
Alan: 3D ICs present plenty of complexity, but we’ve handled that before. In practice, LEDs and MEMS both have relatively more of their value added in the package. As a result, they need closer alignment between the fab side and the packaging side than commodity IC folks. We’ve learned from working in these types of scenarios. For example, we addressed operations for an LED manufacturer, with one model concurrently predicting material flows at three factories, each operating on a different continent.
3D InCites: Risk seems to be an important part of developing a 3D IC supply chain. Too much risk and it’s too scary; not enough, and all your competitors will be so close on your heels that your business will suffer before it’s even off the ground. Is risk assessment a part of what you cover in your models?
Alan: Yes. Risk is an important part of it and we look at it in a large variety of ways. For example, if the end product does not sell as well as hoped, maybe because it’s too expensive and the supply chain feels it. If the technical approach taken is largely driven by material costs, you can cut those costs relatively quickly by matching the material purchases to the demand.
3D InCites: Material purchases are the kind of thing that can be ramped up or down pretty rapidly. What about on the 3D IC equipment side – dedicated Deep Reactive Ion Etch (DRIE) tools for TSVs, for example, or dedicated CMP tools for removing thick plated copper?
Alan: On the 3D IC equipment side, you have much less flexibility. You buy what you need to address the demand, but if demand is less than expected, you can have very expensive tools sitting idle. Etch tools in particular can be very pricey, especially with their modest throughputs. Looking specifically at 3D ICs, the wafer fab is more capital driven, the module is more material driven. The different operations look at risk through different lenses. Not surprisingly, they arrive at different strategies to mitigate risk.
Of course, the equipment and material suppliers know these risks as well, as they are part of the supply chain. And their agenda is part of the equation. There is an upside case regarding risk as well, adding capacity isn’t always easy or quick. For example, if demand exceeds capacity, equipment takes time to order, get built, get delivered and get qualified before it adds the needed incremental capacity increase. Sometimes you need to do some factory remodeling to add that capacity, which can also introduce a delay. All it takes is one part of the process to hiccup for the supply to be constrained during a period of high demand. The key takeaway is that different parts of the supply chain look at risk very differently. Across the supply chain, those differences can become problems if the agendas differ. We can quantify risk in many areas, such as yield, demand, costs, pricing, adoption rate, learning rate, availability, etc using both single and multi-variable scenario analyses. Quantifying risk helps companies to bridge their different approaches to risk.
3D InCites: Can you cite some examples of how agendas can differ in a 3D IC supply chain?
Alan: For example, one company might ask another company in that supply chain to purchase a critical piece of equipment to mitigate their risk; here, the parties are concerned about large capital outlays and equipment that might have a short life cycle. A second company in the chain might want a guaranteed minimum business level, knowing that they will get paid for delivery. A third company might want exclusivity related to the product, the guarantee that they have the market for at least a period of time. A fourth might want royalty-free licenses to use the technology to pursue other customers, seeing a way to grow the business into multiple supply chains. It’s OK if everyone wants something different. It is not good if you don’t understand the agendas, and the reasons behind them. The modeling efforts help each company address their specific agenda, while showing how that also supports the entire supply chain.
3D InCites: If each situation or set of interlocking agendas in the 3D IC world is unique, can you describe a general case study to look at differences?
Alan: Yes. You have two basic drivers that lead to money. You can change the cost of building your product with things like increased productivity, less scrap and more efficient material use. And you can change the value inherent in your product, which usually means you can sell the product for more money by making it differently. Among the more common ways of adding value are making products with increased functionality or with reduced power consumption.
3D ICs are driven by a change in value. Smart phones, for example, place a value on how much functionality you can get in a very small volume. 3D ICs create value by increasing functionality within the same volume. For a case study, I’ll consider a wafer fab going from running only 3 devices to running 30 different devices using 3D IC technology. Why would a fab consider that? Simple, the newer devices are being better targeted to specific end products and can be sold for more money. However, with the added products, that also means substantial cost additions in several areas. Each device needs new test programs, new qualifications, etc. Litho, with all the new reticles, would see a substantial uptick on costs while concurrently reducing capacity due to all the reticle changes needed at the scanners. In contrast, a shared process module like wafer thinning may have no change in requirements. The increase in designs would create added product value, and must override the process costs in the fab and elsewhere. Meanwhile, the thinning process would still be driven by cost reductions by things like improved use of materials and increases in productivity – with the increase in products from 3 to 30 having minimal influence on that particular step. Identifying the costs needed to attain the increased value helps guide the decisions made at each operation.
3D InCites: How do you start thinking about this 3D IC business process?
Alan: You can start anywhere in the supply chain, but the most valuable results are the ones driven by the end user. The folks that make smartphones see value in greater functional density. I spent a lot of years in litho and in the wafer fab world, where we always thought of density as a 2D issue. Better designs and improvements in processes allowed us to produce more transistors per unit area. The mobile device maker wants more transistors per unit volume. The smartphone maker knows the value of their end product and whether it’s worthwhile to make it better with 3D IC technology. They have to determine if the added value is enough for their supply chain to incur the costs needed to get the job done. In the early adoption, we’re seeing about two-thirds of the silicon disappear with the wafer thinning processes, taking off many hundreds of microns, and then placing thinned devices into modules instead of individual packages. A lot of volume goes away, it is low hanging fruit. But what comes next? Will taking off a few more tens of microns pay for the additional processing complexity? Smartphones (and other mobile device makers) are the ones driving the business, and they need to drive that answer. If the answer is yes, this supply chain analysis will help the smartphone manufacturer to get their supply chain on board and ready to move ahead at a rapid pace. Those players won’t get on board and stay on board if the money part stops making sense for them.
3D InCites: With 3D ICs, and the 3D IC supply chain, where do you fit in?
Alan: We see ourselves as helping in two basic areas. The first is strengthening the 3D IC supply chain. If everyone in the supply chain understands the risks and rewards, you get better commitments to stay in the chain for both the short and long term. Second, we can make the pie as large as it can be, by making sure the total available profit is understood, that everyone knows their part, and that everyone gets to take part. A more reliable supply chain means less risk for each individual member, less risk for the end customer, and, probably, a happier end customer. A smarter 3D IC supply chain means more profit for all. We think that matters.
3D InCites: Thank you Alan!
Alan Levine has spent over 30 years working in high technology manufacturing, with an emphasis on manufacturing productivity. Alan has been with Wright Williams & Kelly, Inc. since 1995 and focuses on helping clients increase the value they receive from their complex operations. Previously, he held positions with Fairchild Semiconductor, KLA Instruments, and Ultratech Stepper. He holds a degree in Chemical Engineering from Cornell. Alan can be reached at 408-323-9780 or at alan.levine@wwk.com.
From Petaluma, CA, thanks for reading. ~ PFW
