Good morning, everyone. My name is Steve Cantor, Vice President of Corporate Relations for Entegris. I'd like to welcome you all today, both in the room and on the webcast, to the Entegris 2017 Analyst Meeting. We have a very full agenda, which we believe will illustrate very well why we're so excited for the future and why the company has a great place in the industry and also, a great trajectory ahead of it. Remind you that we will be making forward-looking statements today. There are risks and uncertainties related to those statements, we encourage you to read our filings with the SEC very carefully. Back to the agenda, which is in front of you.
This shows you a list of our speakers today, beginning with Bertrand Loy, our President and CEO. Following Bertrand, we have a guest speaker, Tim Hendry. He's the former head of Intel's material supply chain. We think you'll find Tim's perspective extremely interesting since it reflects both the perspective of many of our customers as they contend with both technical and supply chain challenges with respect to their electronic materials inputs, which are becoming increasingly critical to their success. Following Tim, you'll hear from Todd Edlund, our Chief Operating Officer. After Todd, we're going to take a short 10-minute break. For those of you in the room, there are boxed lunches outside, we'll ask you and we'll remind you to try to grab one quickly and get back, and we'll start promptly with presentations by the heads of our three divisions.
Following those presentations, Gregory Graves will wrap up the formal part of the meeting with a discussion on our financial strategies and our target model. There are a few housekeeping items that I want to cover before we begin. We're going to try to end the meeting promptly at 2:00 P.M. Eastern. We ask that you hold your questions until all the presentations are complete. You'll notice in the room here in front of you is a survey, and we ask you to take a moment following the meeting to fill that out, because that's really helpful feedback for us to continue to improve these meetings. The meeting is being webcast live and will be available on demand through our website. We posted an electronic copy of the presentation material on our website.
If you need a copy, you can see me and I'll get it to you. With that, I'll turn it over to Bertrand.
Thank you, Steve. Good morning, everyone, and thank you for joining us in New York or on the webcast for our annual Entegris Analyst Day. As Steve mentioned, we have a pretty full agenda, but here on this slide, you can actually see the major themes that we want to be impressing upon you today. First, Entegris has been executing very well, and we have delivered very strong financial performance over the last few years, but we can and will continue to improve from here on. The semiconductor industry, which is our largest market, continues to be a great place to be, especially as you are a materials company. I would argue indeed that the next level of device performance will be mostly coming from advances in new materials requiring higher levels of purity.
Because of its broad portfolio of capabilities, we believe that Entegris will prove to be the ideal partner to enable the next wave of innovation that our industry requires. Lastly, as the largest independently traded specialty material company with a very strong focus on electronics material, we offer a very strong and capable platform for future M&A. This is our business model at glance. We have a very clear mission. We focus, we innovate, and ultimately, we deliver. Our mission is to create unique value for our customers by developing very critical solutions for their manufacturing processes. We focus on some of the most demanding manufacturing environments, and more specifically on semiconductor processes where we can capture the most value for our solutions.
We have, as a result, been able to develop a very rich pipeline of opportunities, which we believe will allow us to continue to outpace the markets that we serve. The result of all of this is a very unique business model that delivers stable, recurring revenues, strong cash flows, and exciting earnings leverage. In 2016, we celebrated our 50th anniversary as a company. More importantly, we celebrated many records. Our top line reached approximately $1.2 billion, as we grew 8.7% versus previous year, which actually was about 400 basis points in excess of the underlying industry growth rate on a constant currency basis.
We maintained a strong focus on improving our bottom line and delivered EBITDA margins in excess of 22%, which is in line with the objectives that we stated at the time of the acquisition of ATMI, and certainly a very nice step up from the EBITDA levels that we are running at prior to the ATMI transaction. We continued to generate very healthy cash flow that we used to pay down our debt, and we brought our net leverage ratio down from about 2.0 at the time of the announcement of the ATMI transaction to about 0.7 at the end of December of last year. Central to our success is our customer engagement model that you can see on this slide.
This is really essentially our internal compass as we strive to reach very strong internal organizational alignment from the very early stage of the ideation phase, all the way to the time when it's time for us to release our products to high-volume manufacturing. This model is constantly forcing us to assess our capabilities and make investment decisions so that we can remain the reliable supplier that we aspire to being. Reliable supplier that is able to meet the ever-increasing expectations of very demanding customers. Third, we'll actually provide some examples of the types of investment decisions that we've been focusing on recently in terms of technology to continue to enable the industry technology roadmap, investments in our global infrastructure to be an effective extension of the engineering arms of our customers in most of our major markets.
Also investments in our own manufacturing capabilities as we strive to reduce variation in our manufacturing processes and increase our levels of quality. For the past many years, the semiconductor demand has been driven by a few discrete drivers, PC, the notebooks, moving on to tablets and smartphones, but all of that is changing very rapidly in front of our eyes. Going forward, semiconductor demand will be driven by many interconnected drivers. For the new digital world to fully deliver on all of the nice promises that it covers, we will require a total revamping of the IT infrastructure so that we can really truly deliver the desired end-user experience. Think about the faster network that will be required. Think about the more powerful computing all around us that will be required.
Of course, think about the new storage solutions that we'll need as we generate this massive amount of data. Think about the autonomous vehicles that will be equipped with over 200 sensors and possibly a server in the trunk. Think generating about four terabytes of data a day. Think about the new stadiums equipped with the new volumetric video capture technology, capturing and streaming up to two terabytes of data a minute, so that you can enjoy a new viewer experience in your living room. All of that, obviously, is already showing some signs in the overall trends for semiconductors. You can see that on this slide here. First, if you look at the period from 2010 to 2016, you can see two departures from the previous decade. First, you can see less volatility in the demand for semiconductor.
Second, you can see a much closer correlation between semiconductor demand and global GDP. I would expect actually both the reduction in volatility and the stronger correlation with GDP to continue in the years to come. Lastly, I would also expect the growth rate for semiconductor demand to exceed the growth rate of global GDP as a result of all of the trends I was describing in the previous slide. This virtuous cycle of growth in the semiconductor market will have very positive impacts across the ecosystem. Entegris will benefit in many different ways from this. Remember that 80% of what we do is consumable in nature. Think about the materials, think about the filters that are used every day in the production cycles of the fab. The primary driver for those products is the level of activity in the fabs.
The best proxy for that, as you can see on the upper right pane, would be wafer starts. As you can see, we expect very steady increase in wafer starts for the next two to three years. The secondary driver for those consumable products that I would warn a flag, is the number of process steps required to process a wafer. As you can see, as the industry transitions to more demanding geometries, we expect a steady increase in the number of process steps. Lastly, 20% roughly of our products are still CapEx driven, and while we expect some level of volatility on that part of our business, we continue to be very optimistic around the health of wafer fab equipment spending in the foreseeable future, as you can see on the bottom part of the slide here.
For the past 50 years, the semiconductor industry has been very diligently following the cadence of Moore's Law. Physical scaling was king, and as we collectively, as an ecosystem, work very hard to reduce the size of transistors, a lot of good things would happen to device performance. We would see greater computing speed, we would see better energy efficiency, and we would see the cost per gate coming down node after node. Let me tell you that this has become a lot harder to achieve recently with physical scaling alone. I would argue that the next generation chip performance will be mostly driven by material scaling as opposed to physical scaling. I would define material scaling as the adoption of new materials, more performant materials, requiring greater levels of purity.
Many presenters after me will explain to you why we believe that Entegris is the ideal partner to enable, but also to benefit from those new requirements. Our portfolio is actually very broad. We have separation technologies, advanced packaging solutions, process chemistries, advanced materials. The addressable markets for all of those products is actually very large. We have consciously decided to focus our development and commercial efforts for the past many years on some of the most demanding process environments, where we can actually capture most value for what we do. As a result, our served market are somewhat smaller, and we estimate our served market to be about $3.8 billion. Many of the divisional leaders that will succeed me in their presentation will actually provide a little bit more detail around the size of their respective served markets.
While we certainly manage those three divisions independently, what makes Entegris unique in the eyes of our customers is the ability to combine and leverage all of those complementary capabilities to create very unique solutions to help them solve very complex yield and integration challenges. Tim Hendry will actually talk to you about some of the process challenges encountered in the fab environment. Tim will also talk about why uninterrupted material supply chain is the lifeblood of the semiconductor industry. Todd Edlund and the general managers for the various divisions will then explain how we are putting all of those pieces together to create very unique solutions, why those solutions matter for the ecosystem.
Why, as a result, we expect to be in a position to grow as we expand our SAM, expand our share in our core semi markets, but also as we leverage our knowledge and know-how and capabilities into select adjacent markets. After all of those contextual considerations, I'm pleased to unveil our growth formula, and here it is. Growth formula calling for a 4%-6% CAGR over the next three to five years. As you can see, this 4%-6% growth objective is the result of stacking up a number of assumptions, starting with the fundamental layer, calling for the GDP growth of about 2%-3%. The middle layer really captures the additional growth we expect in our semi markets as the world invests in the new infrastructure required to make this new digital order a reality.
The upper layer is really what is within our control, and Todd and the division leaders will characterize some of those opportunities available to us as new materials requiring greater levels of purities are being adopted by the industry. Overall, again, the growth objective here is to continue to outpace the underlying industry by about 100-200 basis points for the years to come. In summary, we expect a favorable business environment for the foreseeable future. We expect material scaling to emerge as the primary driver for the next generation chip performance. We will capitalize on all of those trends to continue to outpace our markets by about 100-200 basis points. We'll maintain a very strong operational focus with the stated objective to increase our bottom line at twice the rate of our top-line growth.
We'll continue to be very thoughtful in our capital allocation decisions, the sum of all of those initiatives will translate into a very meaningful expansion of our EPS. Gregory Graves, our CFO, will provide more details around every one of those components in his presentation. Thank you.
Okay. Thank you, Bertrand. Before I turn it over to Tim Hendry, what key takeaways in Bertrand's presentation? One is we're executing really well. 2016, we set records in revenue, EBITDA, EPS. Solid execution. Earnings per share growth the last three years, 17% CAGR. We're in a good industry. The semiconductor industry has changed over the last several years. The number of applications have diversified. We're no longer dependent on just the PC and handsets. As the device makers have consolidated, the industry's become more rational. All that means more stability, but also greater growth as the pervasiveness of devices becomes more important. Last thing, we're in a really good spot within that semiconductor ecosystem. About 80% of what we do is unit driven, only 20% tied to the capital cycle. What we're not is a capital equipment maker.
What we are is a materials player. With that, I'm going to turn it over to Tim Hendry, our guest speaker today. Tim has spent about 28 years with Intel. That time was split roughly evenly between running the fab in Intel's large fab in Albuquerque, which was their first 300-millimeter fab. The second portion of his career, he ran what they call fab material operations, which is essentially think of that as supply chain for the consumables part of the business. With that, I'll ask Tim to come up and give you his perspective on the materials world.
Good morning. First I'd like to thank Bertrand for the invitation to speak today. It's always an interesting time to talk around from the analyst perspective. I've spoken at different conferences before. I've never spoken at Analyst Meeting, really enjoy hopefully the interaction today as well as the Q&As. Today, I would say the introduction of new semiconductor technologies or the success of those is largely based on two measures. One, did the transistor performance hit the measured targets that you had forecasted? Two, is the technology yielding to the standards that you expected? Today, I'll touch on both of those metrics, both on the performance side and how materials has really been a key role in improving the performance of semiconductor technology. Also, I'll talk about the role of yield in terms of the success of technologies.
I think you can probably see from history, when people talk about the success of technologies, they talk about, did you execute on time? Is the technology delivering the performance that you predicted? Is it basically hitting the yield targets that you forecasted? Any one of those, if they're not achieved, will ultimately cause either a delay in the technology or impact your company's reputation. Given a material scientist, I must show at least a periodic chart, right? This is the only technical periodic chart here. If you look at the trend here is more materials are being introduced into semiconductor processing. Two main reasons for that. One, material suppliers are becoming much more adept at handling and delivering more complex materials. In the past, the sensitivity of these materials were such that very, very difficult to actually introduce into manufacturing.
With the greater control that material suppliers have today, we're actually able to expand beyond the traditional types of materials. Two, the equipment manufacturers are delivering new equipment that's able to take advantage of those new materials. Clearly an expansion of the number of materials that are actually used and also being explored. If you look at the right-hand side on this chart here, materials is playing an increasing role in actually delivering the device performance. I've broken it out here into both scaling, design, and materials. It's a little bit difficult to discern exactly between the three of those components. I think suffice to say, materials has delivered a much bigger role in how the performance of those devices have evolved over the last several generations, and I really don't see that trend changing.
The role of materials in delivering performance is going to increase over time, and it has increased, especially as you moved away from the 28-nanometer node with the introduction of high-k dielectric. To go down a little deeper in terms of how these materials have actually contributed to device performance, on the left-hand side is a cartoon, with a cross-section of a 3D of a logic device. On the right-hand side is actually a cartoon of a FinFET transistor. One particular design, FinFETs are the predominant transistor design that's in production today. The efficiency of power is going to be a key factor going forward. If you look at most of the markets, whether it's IoT or data center, power efficiency is key. New materials can have a big benefit in terms of helping to achieve those power efficiency targets. Secondly is still the speed of transistor.
Speed still matters. Certain markets, especially in a data center, the performance of those data centers is both measured on the ability from power efficiency as well as speed. New channel materials clearly will play a role in helping to improve the speed of those transistors. Lastly, how you connect with those transistors. You're going to require new metals because of the contact resistance. You don't want to develop a very fast transistor and then not be able to connect to that transistor or have the connection of that transistor impede the overall performance. You need to develop a total solution when looking at the overall transistor. The other thing to note here is real transistors aren't quite this pretty. This is where you'd like to be.
Very tight conformal materials with very well-controlled, measured depositions, as well as you're going to move to more selective depositions, which will allow you to become more creative in your transistor design. These are two factors here that I think materials will play a key role. If you look at the back end, which is actually the wiring of those transistors and connecting up within the circuit, you can't ignore the back end and spend all your time on the front-end transistor design. If you think about it, as dimensions are getting smaller, the wire signal delays is going to become an increasing factor in the overall circuit performance. As the dimensions get smaller, traditionally, we've used a copper type of back-end wiring to connect up all the transistors.
That is going to be insufficient, and you're going to have to likely move into a new metal, cobalt being one of the leading industry candidates. When you integrate a new metal, you have many different types of other materials, such as liners, and as part of this, I'm showing a via here. You have to change the liner and barrier in order to accommodate that new metal. It's an entirely new system of integration that must be put into play. In addition, the dielectric that separates the wires also will need to move to a more low-k, lower overall dielectric constant. Otherwise, the capacitance will have a bigger role. You have to work on both.
When you start introducing these low-k materials, they also become more fragile, and the integrity of the device is very important from a mechanical perspective because you could have reliability problems if you don't have a well-designed dielectric that has both low-k properties as well as good mechanical properties. The back end, in addition to the front end, will both evolve from the materials perspective to help move forward the performance of the device. Both of them move in concert. Some technologies, you'll spend more energy on the transistor design, and you may leave the back end as it was. In other cases, you can't. You'll have to do both in order to achieve your targeted performance metrics.
On the left-hand side of this cartoon is kind of an evolution of the basic transistor going from a planar device into a FinFET device, as shown here on the bottom, a gate-all-around device, which is one potential transistor design for the future. One thing to note in all of these is that you're going to require more materials innovation, and they need to be deposited at an atomic scale. ALD, or atomic layer deposition, is becoming a more predominant vehicle for depositing films. You're also going to move into ALE, atomic layer etch. You're removing and depositing at a molecule level these films, and they must be deposited very precise in order to maintain the performance and expected design. If you look at the right-hand side, it's basically a cartoon of a stacked 3D NAND device.
As you look into different dielectrics, which separate the wiring in all those stacks, that'll potentially drive new metal precursors for 3D NAND. 3D NAND is a significantly growing market. As you can see, the memory transitions are beginning to move into 3D NAND. A lot of increased capacity put in place around the world associated with 3D NAND. The punchline here is atomic layer scale is where everything is moving towards, and the deposition and control of that is extremely important to maintain the device integrity and the complexity of the transistor design. With this comes more process complexity. Bertrand began to show the number of increased steps associated with these advanced technologies. You can see here as you move up into predicted in terms of seven and five nanometer, a kind of an exponential growth around process steps.
One is driven by the complexity of the transistor design, two, you're moving into a lot of multilayer patterning in order to achieve that because of the delay in EUV has driven a lot of this increased number of steps. The red line is associated. This is kind of a theoretical line. If I didn't do anything to improve my step yield beyond 28 nanometer, I began to introduce those new technologies with all those complexities, I would see an overall yield degradation. This is unacceptable. If you're going to move forward with Moore's Law, you cannot have yield degradation associated with the introduction of the new technologies. You must improve your yield performance with each individual step in order for the cumulative yield of the overall device to meet your targets. This is essential for Moore's Law to continue.
The targets for cost per transistor are predicated on high yields. Yielding a particular device is extremely important. If I showed this kind of graph, I was still working at Intel, and the red line is my prediction going forward, I wouldn't be at Intel. This is unacceptable performance. You must be able to demonstrate high yields for new technologies, and that's a critical component of Moore's Law. On the right-hand side, I'll talk a little bit about EUV just briefly. I'm not trying to make a prediction here about when EUV is going to be introduced. There's plenty of speakers out there in the world that'll talk about EUV introduction timing.
This is a theoretical graph that says, if I can introduce a seven nanometer, I can begin to bend the complexity curve because I can reduce the number of multilayer patterning steps by one EUV operation. It has great economic benefit, likely will have yield benefit, and it will begin to bend this curve with respect to number of steps required for each process node. I think steps will still increase in number, but they'll begin to bend that curve a little bit associated with it. EUV is an important technology. I think it's going to require to enable future technologies, not just for its resolution capability, but also for the ability to reduce the complexity of what it takes to build a device. Talk a little bit about purity.
On the left-hand side is kind of an analogy about just how pure materials need to be today in overall manufacturing. A little bit of analogy. One drop of water is basically the impurity that's in all of the water that passes over Niagara Falls in one day. That gives you an idea of what level of purity we're trying to target for overall materials, and it's the predictability of those materials that's very important. I'll talk a little later on in a few slides about it's not just reduction of purity, but it's control. Once a given purity of a material is defined, it's very important that it stays in that window. If you try to improve on impurities later in the development cycle, that impurity actually may be playing a role that you didn't know in terms of how the overall material performance.
It's not just about going lower, it's about control. Control is extremely important in predictability. The other factor that plays a role in terms of yields is overall basically defect or killer defect size. On the right-hand side is basically a picture of what's the wiring overall pitch at 20 nanometers. You can see there that the defect size that would actually bridge two metal lines together, causing a short circuit. If you look at a seven nanometer typical metal pitch, you can see the defect size that would cause a circuit to fail is much smaller. Fundamentally, you have to reduce the defect size as you move forward. Otherwise, you're going to end up with that red line, which if you predict that red line, you're not going to move forward in a technology until you can demonstrate equivalent yield performance or very little yield degradation.
Very important factors. In this slide here, I've drawn a simplistic supply chain for a material as it's produced all the way to application actually on the wafer. The raw material suppliers that are basically moving into the first segment of supplying electronic materials, they have very little motivation to change their purity and the quality of their materials. Typically, electronic materials is a very small segment of what they deliver into a very large chemical market. They're not going to invest in a new plant to produce simply better purity for the electronic materials market. What do you do? If you're not able to actually get your sub-supplier to move forward in their quality, you have to basically build in additional purification steps prior to the introduction of into the electronic materials. There's actually an intermediate step there that says, let's control that purity.
When I worked for Intel, that's what we did. We basically, in order to control the raw material supply, we worked with our material suppliers to introduce new purification and filtration operations going into the material manufacturing operation. It's very difficult to extract an impurity and a particle way down into this supply chain. Very expensive and very difficult, sometimes with unforeseen and undesirable results. We always tried to work very far upstream to make the material as pure as possible going into the manufacturing operation. This is critical, and I think more and more you're going to see intermediate operations in there to control that. Controlling the larger chemical market is like tilting at windmills. It just isn't going to happen, right? You're not a big enough player to get them to do, you must control your own destiny.
As you begin to manufacture a material, all these elements here are very important, both the manufacturing operation, and today there are many world-class suppliers, such as Entegris, of manufacturing of materials. That part, I think, the market is getting pretty well established and mature. That doesn't mean I didn't worry about the manufacturing of materials, but I worked with trusted partners to basically develop those materials, and they generally did a very good job. Usually our issues didn't result from that actual manufacturing of the material. It was either variability in the upstream or downstream. Also, once you've spent all this time manufacturing this very pure, stable material, transportation is key. I ran a global supply chain. We had factories in Israel, and Ireland, and the U.S.
I had to be able to basically transport materials across the world with fairly long cycle times, because not everything can be shipped on a plane. Many materials must be ocean shipped. You had to rely on the stability. When it arrived on that dock, it had to be the same quality as it was when you left. Transportation is a key element in overall control. The other aspect that you cannot ignore is actually the manufacturing of a plant today. When you build a semiconductor plant, you must control the valving, the tubing. When materials flow through a tube, a tube is not stagnant. It will leach out materials from those tubing. You must be very careful about the types of materials they're using for valves, as well as piping.
That's another source of contamination that can basically ruin the great material and work that was done previously. Finally, at the tool level, to have predictability, you must control the filtration at the tool, and this is your last line of defense and your best opportunity to also make some improvement in that material as it enters into the final application on the wafer. Really, the total supply chain really needs to be managed and developed in order to actually deliver quality of material onto the wafer. What did I look for when I was actually purchasing and developing materials for Intel? It really keeps you up at night when you're running a global supply chain and you have a multitude of materials and multitude of ways that things could go wrong. You want to work with somebody that you can trust.
The other thing is you want to work with somebody when things do go wrong that has the capacity to solve the problems, because you're working the leading edge, things happen. It's not necessarily anyone's fault. It may have developed a new interaction that you didn't see before. I really look for companies that have strong engineering capabilities, and that's not just to develop products, but also to sustain the product once you've gone into manufacturing. The other element is speed of execution. Speed is very important. If you look at the cycle time of a new technology, it's getting very long due to all those extra process steps. You have to be able to develop and execute very rapidly, both to iterate as well as to solve problems. In order to solve problems, analytical capabilities is very important.
If you don't know how to diagnose a problem, if you can't tell what's changed, you have a very difficult time determining what the solution path is. Analytical capabilities is the key, and it's kind of fundamental to actually, if you can't measure it, you can't improve it. If you can't measure it, you can't fix it. Analytical capabilities is also one of the things I worked for. Once you've either chosen a particular product, material, or developed it, or you're in production, sometimes the amount of capacity that you need is not known. You need a partner that can move and has the capacity to actually expand very rapidly and also be very agile. If the prediction of consumption is not necessarily known until you get into manufacturing.
Once you're in place, your filter frequency, for example, may be much sooner than what you'd expected, so you need to change more often. You need to come back and say, "Hey, you need to double your capacity." You need to work with somebody that has the capacity to be able to respond to that demand change rapidly. Finally, someone who can manage the supply chain. I had no problem coming in to a supplier and helping them manage their supply chain. I had the expertise. We had the people to do it at Intel. We were no problem coming into your backyard and saying, "Okay, we're going to help you." I much preferred to have a supplier who had the expertise to do it on their own. I had other things to do.
I look for companies that have the expertise to manage the supply chain all the way up to the raw materials. In summary, materials are going to continue to play an increasing role in overall device performance gains as new transistor architectures are developed. I think this trend will continue. Process complexity is going to continue to be a headwind for new technology node introductions, and you must develop solutions that help reduce that complexity. As a company that can bring in a solution that reduces the complexity without sacrificing the performance is going to win market share. You have to look at it from a holistic system when you look at the supply chain in order to basically deliver materials that are going to be acceptable from a quality level for these new advanced process nodes.
With that, thank you very much for your time today.
Thank you, Tim. Before we shift over to Todd and talk about how all this comes together, I want to just give you a little bit of history. About two years ago, Bertrand comes back from the Intel Supplier Day, and he said Tim Hendry gave a presentation that could have been our IR deck. He talked about the importance of materials. He talked about the purity requirements of materials. Tim was a guy that we really respected. He'd been fast up to a VP at Intel. He'd run, like I said, the whole procurement organization. When we had the opportunity to have him here today, we were really excited about that. What did he tell us? Bertrand talked about how materials are becoming more important. Tim equated that to device performance. He equated that to the importance of materials in driving yield.
He showed a graphic of the relative importance of materials compared to the laws of physics are diminishing, we're now depending on design and materials. He showed a 3D NAND chip and the stacking, and he showed, frankly, where our deposition materials fit into that. He talked about lithography and the importance of purity, how quickly the number of process steps are moving up as we go to 10 nanometer. He also showed that as we go to EUV, those process steps don't fall off a cliff. As Entegris, we're still in a very good place as you move to EUV. The last thing that he hit on, though, was the purity requirements.
If you think about what we do, we're not only making the chemistries and the gases, but we're ensuring that they're pure across that supply chain, whether it's the beginning of the supply chain, where we're filtering product within the manufacturing environment, the containers we make, the shipping products we make, the tubing that we make that moves it around the fab. The whole materials ecosystem, we fit very neatly into that, and that ecosystem is absolutely critical in today's advanced devices. With that, I'm going to turn it over to Todd, who will give you an intro into our three divisions.
I was sitting in the back of the room listening to all the honking out in the streets, I think I realized one of the benefits of self-driving cars, you're not going to need a horn. These streets are going to sound very different if those realities come to us. I want to talk about, you heard a couple of things today that are encouraging to me as the COO for Entegris. One is Bertrand talking about the strength of the industry, what's driving it, why we believe it's going to be a solid place for us to continue to grow and outperform. Then Tim to talk about the importance of materials.
There was a great article in The Wall Street Journal this morning, I'd encourage you to take a look at it, about really design becoming more important and all those elements of design, what Tim was showing you, those architectures, that's what they're talking about, is really being able to enable those architectures. That's how they're driving to improve device performance, and those devices are becoming more ubiquitous across our lives, really. The materials are really how you enable that, how you are able to achieve those designs. I'm going to talk a little bit about how that fits into our portfolio. These are the three divisions of the company, you're going to hear from the general managers of each of these divisions a little bit later today.
Just to give you a broad feel for it, if you look at our Specialty Chemicals and Engineered Materials or SCEM division on the left there, these are really those materials, the periodic table, if you will, being applied to the semiconductor industry. These are the things that are enabling these new architectures and these new devices to actually happen. They're integrated into the actual device themselves. They ship with the wafer in some cases, or they're integral in terms of how they do cleaning to those things that are being deposited on the wafer that Tim talked about. That's 100% unit-driven business for us, we are very focused on the specialty part of that. Again, highly engineered solutions, again, even working with those designers of the chips that I read about in that article this morning. Microcontamination Control.
Microcontamination Control, just really think about that last line of defense. This is where we're actually purifying or filtering gases and liquids and also watching the airborne molecular contamination around particular processes or in the fab. Really trying to focus on contamination control, which gets us to that yield that Tim was asking for. Very important part of that solution. Advanced Materials Handling also, protecting substrates, either wafers or masks or reticles or individual devices at the end of the process, and then fluids as they go through the process. This is really about, again, yield, protecting from physical damage, electrostatic damage, contamination. Those are the three parts of the company you're going to hear more about today.
Let me talk to you a little bit more about how they fit together and why we think it's very complementary to have the three of them. When I look at this really tells a story, I think, about why does it matter to have these three divisions? Really, at the end of the day, a lot of it is about being able to move faster. I think that these three divisions help us to do that. Let me explain a little bit about that. Starting with Advanced Materials Handling, this is really an important entree to us in the fab. One of the first things they start to do when they're going to design a fab is look at the layout. What are they trying to accomplish with the device? They'll look at the layout of the fab.
How are they going to move product around? Are they going to use automation? How are they going to protect devices, and how are they going to protect wafers and reticles throughout the fab? We get involved in that discussion very early because we have this division. Similarly, as they're designing the sub-fab, how are they going to move chemistries and gases around? Same thing. We get involved in that, working with the engineers and architects of the fab to be able to understand how are they going to lay out this fab? How are they going to provide purity in the fab? If you look at specialty chemicals, this is really, again, that enabling capability of the company to help them actually design the device. We're getting early access to what is this device going to look like? What kind of architecture? Is it a 3D NAND?
Are they using FinFET? Are they using variations on that? We can find out actually working with the device manufacturer and the chip designer, what that's going to look like, because these materials have to be an integral part of that solution. When you look at microcontamination control, as I mentioned, this is really that last line of defense for defect reduction and yield improvement. These products have to be very much matched to the application they're going in. It's not simply a matter of having a filter like a coffee filter, where you're sieving out the individual grains of coffee. It's much more, how do you match yourself to the chemistry? That gets to where the interplay is. How does this help us move faster? We do filters, and we do the specialty chemicals.
One thing we've learned over the years, starting many years ago, is that the interplay between those two is critical. How you do filtration today is really not, again, just about sieving, but it's about how you design the chemistry and the surface of that filter membrane to work right with the chemical it's going to see. If you don't know what that chemical is, and the chemical companies and even the device manufacturers guard those secrets pretty carefully, it takes longer to get to the right solution for filtration and purification. Because we understand both of those effects, we can work together to get there faster. These two divisions can help each other move faster to get to the right chemistry with the right manufacturing solution and the right filtration solution that works with these advanced chemistries going forward.
If you look at the interplay here similarly between Microcontamination Control and Advanced Materials Handling, inside of a FOUP or a wafer carrier, which is really a small environment for the wafers, we actually do things to control the environment. We do purge, or we actually do gettering technology to pull molecules out that might outgas off of the wafer. Understanding how to make that happen is obviously something Microcontamination Control knows how to do well. Applying that to wafer handling or reticle containment is something that AMH knows how to do well. Together, the two can move faster. Very similarly for chemicals. If we're doing fluidic solutions in the fab, moving fluids around the sub-fab, we have not only the valves, fittings, and tubing and sensing products, control products, we also have filtration. None of our competitors really have that in a robust way.
It's a unique opportunity for us to move faster and get to more complete solution, that broadly capable materials provider that Tim asked for. Then AMH and SCEM together. Here again, chemical packaging and sensing solutions match to these emerging chemistries. AMH also makes the containers that a lot of these chemicals are put into. Tim talked about the challenge of taking a pure chemical that the chemical manufacturer has created, getting it to the fab in that pristine state. We understand a lot about how to do that in our chemicals or our containers business, and we can apply that very quickly to advanced chemistries that are coming out of the SCEM division.
There's a lot of interplay between these divisions and helping them solve problems faster, and that's probably the key thing I'd like you to understand is that really moving faster is the solution. This makes us more competitive, and I'll talk about that makes us quite unique to our customers. Oops, sorry. This is kind of interesting. You saw this ecosystem that Tim described across the top here, and this is just some representative parts of it. How do these divisions play across these different customer types? It's actually one of the things that's really unique about Entegris is that most of our divisions cross really the whole ecosystem from the production of the basic raw material, be it a wafer or chemical, all the way till it touches the wafer and even becomes part of the finished device, as is in case of SCEM.
I won't cover all this, but the thing you can take away and look at in the handout is these are some of the products that we have that help these different stages of customers as it goes through all the way to the finished device. Wafer growers, chemical manufacturers, these are our customers. The actual gases and chemicals that are coming into the fab, the sub-fab, distribution, incoming wafers, these are our customers. This is where we work to provide solutions. Process tool, where all this becomes a part of the wafer and part of the finished device. Very big customer set for us, and really all of our divisions work across those because that's where our products get actually applied or provide that last line of defense to what actually goes into a device. Beyond.
A lot of the products, obviously some of the chemistries here end up on the wafer, on the device, and we handle those finished devices as well. What's unique here for the customers is that they've got a company in Entegris that understands all of these steps and all of the aspects along the way and can actually, and does, partner with all these types of customers. We're actually in the center of this ecosystem, working with all of the different people that enable this process to happen for semiconductors today. What makes us a unique competitor as well? If you look at this competitive environment, and across the top here, we've got some of the main processes that we're involved in at Entegris in the semiconductor world.
We can see Entegris here, and the dots basically mean that it's a significant participant in those processes. Now, here's a list of some of our main competitors and what makes them unique. There's kind of two types of competitors here. There's these large industrial companies like Danaher, which owns Pall Corporation today, Donaldson, DowDuPont, Parker, Saint-Gobain. These are large industrial companies, and a small part of what they do has got a semiconductor application. If I'm a semiconductor customer, I might be a little nervous how committed are they really going to be to investing in the ways you need to invest, as Tim described, to be relevant and capable materials provider to the industry. We think that we clearly are invested in that. That's clearly our focus and continues to be what we continue to focus on for our company.
The other is smaller companies. If you think about are very narrowly focused, so like the Shin-Etsu company, they compete with us in wafer handling, but really nothing else. If you look at even Versum, they obviously have materials that they compete with us in, but they have none of the other parts of that whole supply chain to be able to be unique in helping the customers, like Intel and others, get to solutions. That's really what we think is making us different. We are focused on semiconductor, and that's clear to our customers, and we have very broad understanding of all the processes and where contamination and yield loss can come from. We think we make us a very attractive partner for partnering on these advanced fabs. What makes it a unique business model?
If you think about it from an investor perspective, how should you think about Entegris? As Bertrand mentioned, very highly recurring or unit-driven business across the top here. A lot of what we do is really consumed or used as they're making products. Wafer starts or millions of square inches of silicon, important driver for us, a predominant driver for us. We also get a very nice boost from these investment-driven or capital-driven revenues. They talked about, gives us that early access into the fab, so we understand how the designs are going, how they're going to lay it out. We're a partner from really day one as they're starting to think about what the next fab or the next line or the next expansion look like. We get a boost from that, and then we get the recurring revenue coming after.
It's actually a pretty unique model, even on the investment-driven side. We're also very highly diversified. Top 10 customers are less than half of our sales. We really serve everybody in the industry, and that's kind of what we're showing here on the right. All the segments of the ecosystem of semiconductor, they're our customers. We know them all. We're involved in joint developments with many of them. In multi-party developments, in a lot of cases, because you can't solve these problems, as I mentioned earlier, just by yourself. You need to do it through the design, the integration, and how it's going to be implemented in the fab. We, having all of these types of customers as our partners, makes us unique. You're going to hear more about these three divisions, again, from the general managers coming up.
They each have a little bit of a different role in the company. Microcontamination control, as you've seen, above-market growth. We made some recent investments that I'll talk about there that are paying off. Strong margins. Specialty Chemicals, Engineered Materials, a more stable business with good market share, but some very interesting emerging growth vectors, as I'll talk about in a moment. For instance, deposition materials, you heard Tim talk a lot about that. Growing very quickly for us, we're kind of focusing investment for that division on materials like that, coatings, some of the other material solutions. Very focused investments on those rapid growth areas inside of a large, fairly stable business. Advanced Material Handling. Pretty stable markets.
It's more industry investment driven, as Bill will talk about, than the other two divisions, but it gives us a very important strategic element of that move quickly, be able to move faster than our competitors and provide that leading entry to the fabs. You'll hear some more in a moment about them. We put the platform together, now we need to make the right choices to invest in that platform smartly, get the returns, get the outperformance that we've been experiencing. Several years ago, really 2012 and even slightly before that, Bertrand and I actually made the decision that we saw some of these trends coming. We saw contamination control becoming more important. People like Tim were telling us, "We have yield challenges for these advanced architectures and advanced nodes. We need help." We knew we needed to invest to make that happen.
We actually invested in certain parts of the company, especially Microcontamination Control, but others as well, to increase our R&D so we could be that relevant partner, so we could have those solutions. And really, we've gotten it now to a point where we're feeling that we're in the right zone. This 9%-10% is what we intend to sustain going forward. That sustainment's very important for us to keeping a healthy portfolio. But it's paying off. And what we're showing in this upper right chart is really the growth in the percent of our revenue that comes from products introduced in the last three years. So it's a very particular measure, but it shows how that's grown from about 20% to near 30% on a four-quarter rolling average basis. This is really important for me.
This shows that the choices we've made are paying off, and we're renewing our business. The product lines are regenerating. There are new demands that customers are asking for, and we're able to deliver those. And we really reached what I think is a healthy zone here, this 25%-30%. If we can maintain that, we're doing a great job of being dynamic and staying up with the industry. Beyond that, there's a very healthy pipeline coming up behind us. So this investment level seems to be working very well for us in keeping the right amount of active projects, the right focus on technology and breakthrough projects, platforms, so larger product lines that we're investing in, and it's providing that outperformance. That's really the important thing for the future. We maintain this investment, we'll continue to be able to have that outperformance.
We've also made investments in capacity and technology. Bertrand talked about the keys to really, to success for our business, have the right technology, have the right global presence, have the right operations excellence. And we've really made the right investments to do that. I'm not going to go through all these with you, but as you look at a couple of things, if you look at the i2M Center, memory manufacturer, filter cleaning, and manufacturing expansion, these were decisions that we made a few years ago that have come online in the last year and are really part of the reason that you saw the big increase in our Microcontamination business this year. We needed capacity. The demand, especially for some of the bulk chemical manufacturers, has been growing. It's been driven down into their manufacturing processes to have this purity.
We've made the investments, that's taken off. Technical capabilities, if you look across that, you'll see a lot of investment in Taiwan and Korea, research capabilities near our customers, advanced capabilities in our research centers around the world for the things that we see growing most quickly, deposition materials, coatings, microcontamination are really the focus for that. Localization. A large part, the vast majority of our revenues come from Asia, as you would expect. So you see a lot of investments in Taiwan and South Korea that are really now starting to populate in a more significant way, and that will continue. You'll continue to see, Greg will talk a little bit about some of the coming investments, and they're going to continue to be focused largely on Asia. Lastly, but very importantly, is investment in systems.
Things like statistical process control for our manufacturing has driven a real improvement in quality, which I'll show you, NPD portfolio management. We're spending a significant amount of money in R&D. We want to make sure we had a lot of rigor around how we make those investment choices, and we've put a very capable system in place to make sure that that's happening. Supplier management, the things that Tim talked about, is really important for him our supply chain. Just quickly, this is the results. If you look at the investment in quality, we've driven our defective parts per million down orders of magnitude in the last few years, and then therefore driven up our sigma level, which is really representative of improving, and reducing customer complaints. We're above 5 sigma at this point. What does that mean?
Really what it means is our customers can relax and trust that we're going to have great quality, and we can focus on the future. We can focus on talking about what's next, what products do we want to be working on together, what solutions do we need. We have the confidence in you as a supplier because you've driven those investments to actually result in our quality. Lastly, of course, that impacts our factories. If we're having better yield, we'll get better yield by having fewer quality issues, fewer excursions. That's good for the health of the financials of the company as well. I think, again, the right investment at the right time and producing the results. We've shown you this before. This was from last year's analyst event, which we talked about some of the drivers, some of the individual opportunities that we're pursuing.
Again, just a representative set of things that are happening within the divisions that are exciting, that are driving some of the growth and that we've been investing in. I'm not going to go through these all again in detail, but suffice to say that we're very, very happy with the growth in all of these. We had very strong results in all of them in 2016, and we're on track for what we've spoken to you about before, $70 million of incremental revenue just from these items by 2018. Continued growth beyond. I think our view today is that most of these really actually we see even broader opportunities than we saw a year ago in terms of scope of what we can do here. Lastly, the growth opportunities that you're going to hear about these from each of the divisions.
Again, I'm not going to cover all of this information here because you'll hear about it from the general managers. These are the things that you're hearing about when you hear about semiconductor and what's happening, what's driving needs in the industry, this node shrink or architecture design, these advanced architectures, 3D NAND and memory, the purity needs, the yield needs that are challenging ramping of new fabs, and then Internet of Things in China. China, our fastest-growing region. Last year, 22% growth in China, and close behind is Taiwan, 17%. Very, very strong growth. A lot of that, of course, is supporting fabs that are working on some of those IoT solutions. Lots of activities, and this is, again, just really representative of some of the things that these divisions are looking at to respond to the things you've been hearing about.
We see these exciting opportunities driving the need for value creation. We have the three divisions. It makes us really a unique partner to our customers. I hope I described to you some of the interplay there, how they connect and help us move faster, which is really the goal of that. These special solutions and speed of development and ramp are very, very important to the industry and for us to be responsive to the kinds of demands that customers have, and as Tim described. We've really integrated the right capabilities and made the right investments in those capabilities, as I described with R&D and some of our capital investments, to make sure we're going to be the successful, broadly capable materials provider to the industry that you heard Tim ask for.
At the end of the day, what you have with Entegris is a unit-driven, very diversified business platform, and as evidenced by recent results, outperforming a growing industry. I'll have Greg sum up.
All right. Thanks a lot, Todd. I think Todd did a nice job of sort of summarizing sort of what he's got here. First, you think about the divisions, very integrated, all really kind of centered around contamination control. Contamination control is really critical to the process. They share a sales organization. They share a CTO organization. It's not three separate businesses. We happen to manage them in three segments, but not three separate organizations. The only other point I would make is really the investments that we've made over the last two or three years have paid off. The folks that have been around a while have heard us talk a lot about the i2M Center. We've talked a lot about our increases in ER&D. The growth that we had in 2016, organic growth of 8.6%, that wasn't an accident.
That happened because we chose wisely and made good investments over the prior two years. With that, we're going to take a 10-minute break. There are bag lunches outside. We are going to start back up in exactly 10 minutes. For trying to sort of like Let's see how this goes, Greg. We'll see if you get everybody back in here. I'm going to start talking in 10 minutes. I hope you guys are back. Thank you.
Tell her that I love her next time you see her. Tell her that I care next time you see her. Tell her that I love her next time I see you. Boy, you better beware. She's just trying to warn you that you're bound to get hurt. I couldn't be her last love, so how could you be her first? I want to tell you, buddy, things are bound to get rough. I know that she's only flirting, but I think that I've had enough. Next time you see her, tell her that I love her. Next time you see her, tell her that I care. Next time you see her, tell her that I love her. Next time I see you, boy, you better beware. If you see her again, I will surely kill you. If you see her again, I will surely kill you.
Next time you see her, tell her that I love her. Next time you see her, tell her that I care. Next time you see her, tell her that I love her. Next time I see you, boy, you better beware. If you see her again, I will surely kill you. If you see her again, I will surely kill you. Next time you see her, tell her that I love her. Next time you see her, tell her that I care. Next time you see her, tell her that I love her. Next time I see you, boy, you better beware. There's no cause to think that I won't stay. Haven't I been with you all the way? There's no time like now to make a way. After all, we are more than friends. This could be the time for you and me.
We could go wherever we should be. Don't put words between us we shouldn't say. Don't be acting halfway when you know we're on our way. There are times when I don't see the light, and I don't know if what I do is right. When I'm wrong, it's never meant for you. Don't confuse my love with what I do. This could be the time for you and me. We could go wherever we should be. Don't put words between us we shouldn't say. Don't be acting halfway when you know we're on our way.
Every morning when I wake, a feeling soon begins to overtake me. Ringing in my ears each time I feel my brain and body surround me.
There is life, there is passion, fever and fury. There is love and there is hate. There is hoping and your worry.
Oh, I am afraid to look at my heart. I might fall apart if I can't go without you. If it should become too cold. I know I can endure the frostbite. Throw a blanket then I'll wrap around it. Keep myself so close to life's light.
There is love.
We're going to start our three divisional presentations with our Microcontamination Control Division. For those of you who have been around the company a while, you'll know that this is a very attractive business for us. For those of you who've been around Wall Street for a while, the Street has always really liked filtration companies, and they've always given them really high multiples, which we aspire to. I think Clint's story will really lay out for you what's so attractive about a filtration company. With that, I'll turn it over to Clint.
Okay. Good afternoon. Welcome back. My name is Clint Harris, and I am responsible for the Microcontamination Control Division. I hope you'll find the next section to be pretty exciting. Really, the takeaway is our Microcontamination Control Division is a growing, profitable division with significant competitive advantage over others in the markets we serve. If I look at the business as a whole, and was to sum up what we do, fundamentally, we're in the business of ensuring purity of advanced liquids, gases, and air. It's interesting, I usually like trivia and history, and as I came into this room today, and my apologies to those on the webcast, but we have this beautiful panoramic view of the New York Public Library.
It's actually quite interesting because if you go back in time into the 1900s, right across the street from us here in Midtown Manhattan was a very large above-ground reservoir. It was known as the Croton Reservoir. It serviced the water needs of all of New York. It had 50-foot high granite walls and a promenade that you could walk around a four-acre reservoir. It's really interesting because, at the time, the filters that the city used consisted of large barrels of sand. If you look over the last 150 years or so, obviously, we've come a long way with filtration technology, but it's a really interesting backdrop. Of course, since that time, the reservoir was taken down and replaced by the New York Public Library. If you walk around afterwards, you can see little foundations of that reservoir just outside our venue today.
Within the Microcontamination Control Division, there's really three main areas that describe our business. The first of those is our liquid filtration, second is gas, and the third is environmental or air filtration. Those not only describe the business, but they'll also form the structure of my talk today. I'll walk through each of those in piece. If you look at our business, it's predominantly a consumables business. What's exciting about that from a financial standpoint is, of course, it's a reoccurring revenue. It's also a very sticky business. Once we gain a customer, they tend to stick with us for the long haul. We do have a segment of our business, about 21%, that is tied to capital expansions of fabs as well. There are many filtration companies in various industries, but we are unique in our depth of understanding and focus into microelectronics.
We understand the challenges and risks, and that gives us a significant amount of technical as well as applications experience, which allows us to lead our competitors. The market for filtration as a whole worldwide is a huge market. It's a multi-billion dollar market. We're focused on a very specific segment of that market, which is high-end, technology-driven, and it's about $1.5 billion around microelectronics and semiconductors. If you look at that $1.5 million market, it's a rapidly growing market, and certainly that was reviewed in some of the earlier presentations by Bertrand and the rest of our presenters. I would like to point out two factors that do stand out in driving the business. The first in the top right shows the growth of the number of manufacturing steps required for each wafer processed in the fab, so you can see that growth over time.
Below that, you can see that as the market has progressed, the purity requirements are becoming more and more stringent. When you take those two factors and you combine it with all of the new materials as well as processes, as well as the need to more frequently replace the filters within a semiconductor fab, you can see that Entegris is poised to achieve significant growth in the years ahead. The first of the major components of the Microcontamination Control division is also our largest. It's our liquid filtration segment. This business services several different areas within semiconductor fabs, specifically wet etch and clean, photolithography, CMP.
Within those segments, we see quite a bit of growth associated with new wet etch steps and increasing wet etch process steps within the fab, as well as with an increased number of photolithography processes as chips get more complex and line widths get smaller. We've been in this business for several decades, and over that time, we've really built up a leadership position. We have fundamentally, really only one major competitor in the segments where we serve. Our proprietary technology, as well as applications experience, is keeping others from really encroaching on the spaces where we're strongest. The second major area is gas filtration. Our gas filtration business is heavily focused on OEM process tools, where we supply small filters which are used for dry gas process steps, such as the etch step and deposition steps within the semiconductor field.
This business has been growing rapidly as we've seen an increase in the number of tools used for new processes such as EUV, as well as to support double patterning as line widths get smaller. Additionally, we do see some new segments which are emerging. For example, 3D NAND, as Tim Hendry mentioned earlier. We see that adoption growing, and as that market takes off, there's more OEM process tools which leverage our gas filters. The last major section is our airborne molecular contamination filtration group. These are solutions which are used for cleaning the air in process tools, as well as in the fab in general. I've chosen to show a picture of the Shanghai skyline on a sunny day. I did this for two reasons.
One is to illustrate the fact that molecular contamination at a very small level can have a significant impact on the environment, subsequently on semiconductor wafers. Secondarily, we do see significant growth within the China market. With that growth is quite a bit of opportunity for AMC filters. This market is not just centered on China. I had the opportunity to listen to a speech given at a symposium by Micron recently. Micron spoke of a situation where they saw yield degradation. As they were trying to diagnose what was causing that, they found a direct correlation to the yield degradation to actually forest fires that were taking place in the Western U.S. Once they diagnosed that, they implemented AMC filters. They were able to see their yield not only go back to the baseline but actually improve beyond that.
With these types of challenges, we see additional need for AMC filters throughout semi fabs. I'm going to talk about just a couple of specific examples and segments, growth vectors, which we see. One of these is an increasing trend for filtration further and further up the supply chain. This is not surprising given the value of specialty chemicals in our industry. As an example, you hear a lot about EUV technology. A gallon of EUV resist can cost upwards of $10,000. When you think about that value, you can understand why fabs, as well as their suppliers, would want to make sure that they protect that and ensure the utmost purity of those materials. For us, within micro contamination control, this is pretty exciting because we're expanding the customers we serve.
As you'll hear in the next talk about one of the other divisions, our AMH division, we're finding applications where not only are we filtering these further and further up the stream, but we're also leveraging our carriers and tubing to ensure that contamination is not introduced into these materials. I wanted to refer back to a chart that Todd Edlund presented earlier, where he showed the increase in our R&D investments over the last few years. This is very important because as we've gone from technology node to technology node, and been focusing on R&D, our customers are starting to recognize more and more value from the technological solutions we're providing. With that value, of course, they're willing to pay for that value. That's helped us as we've looked at growing our top-line revenue. This is quite a challenge.
I also borrowed a chart or a slide from Tim Hendry because as he mentioned earlier, if you think about the fact that we're trying to control the equivalent of one drop of water passing over Niagara Falls in a day from a filter standpoint, not only do we have to identify and pull out that one drop of water, but we have to let all of the other good molecules or good drops of water, which are at similar sizes, to pass through. Filtration, it's not as easy as just blocking everything, but you have to selectively pick out the bad molecules while allowing the good ones to go through. Lastly, let me talk a little bit about the business as a whole. Financially, we had about $360 million of revenue last year. Our margins were in the low 30s.
If you go back in time in 2015, 2014, our business was relatively flat. That was partially due to the fact that we were constrained from a capacity standpoint. As Todd mentioned, we invested heavily in our i2M facility as well as in other capacities to give us the capability to grow beyond where we were. Additionally, besides just the manufacturing investments, we did have significant R&D investments, which led to new products released over the last 24 months. Those two factors combined drove significant growth in 2016, so we had about 15% growth. With that growth, we also saw a good drop through to our operating margin. Currently, we're in the low 30s.
As you look forward, we continue to do our R&D investments, we continue to foster our relationships with customers, we continue to work up through the supply chain, we see the opportunity for above-market growth, as well as driving our operating margins into kind of the mid-30 range. With that, I guess I want to bring it back to the view outside the window, I talked a little bit about the reservoir that used to be right outside where we are today. Today, New York Public Library, I was trying to think through an analogy with the New York Public Library, and there's about three and a half million books in the New York Public Library.
If you were to look at all the characters, all the letters in each of the books and periodicals that's in the New York Public Library, the technical challenge that we face within the filtration group is the equivalent of taking one letter, just one letter, the letter A, that's in one book, and trying to filter that out of all of the other letters within the library. If you think about that technological challenge, you look at the investment that Entegris has made over the years, you start to realize that that's quite a challenge that only Entegris is capable of providing to the microelectronics industry, that's going to drive our growth going forward.
Okay. Thank you, Clint. Next time we give a presentation, we won't have Niagara Falls, we'll have a picture of the New York City Library, we'll be looking for that A. Really three things to remember about Clint's business. First of all, this stuff is really complex. This isn't the filter under your kitchen sink, which is a membrane in a cartridge. It's a membrane. It's surface modification. It's purification. Not just anybody can do it. We have very strong competitive moats around this business. It's a recurring revenue stream. Once we're qualified into a process, unless we do something very wrong, we're going to keep that business. More is better in finance, right? Especially if it's on the top line. If you think about this business, the demands for purification, you go back to a slide that Tim Hendry showed.
Something that's a problem at 45 nanometers is a killer defect at seven nanometers. The whole contamination control becoming more important. On top of that, the number of process steps are going to drive greater filtration. The trends around this business and what's happening in the industry are very strong, we've got a very good tailwind in this business. With that, we'll turn it over to Bill Shaner, who's going to talk about our materials handling business.
Thanks, Greg. Good afternoon. My name is Bill Shaner. I am responsible for the Advanced Materials Handling Division. Thank you for coming today. Back in 1966, the original patent for a wafer-like substrate about 25 millimeters in diameter, about the size of a quarter, was applied for by an engineer in Chaska, Minnesota. That was actually the impetus for the creation of the original company in what is now called the Advanced Materials Handling Division. I don't know about you, by looking at this crowd, facing the crowd, I was not on the planet yet, so it has been quite a long time. What is fairly unique about Advanced Materials Handling is the fact that we have been the market leader for over 50 years. We will walk you through the model today and show you what the future lies for AMH.
You can think about the business in three different kinds of segments. One is wafer handling. These are products that are used to protect and transport silicon wafers, either from the wafer grower or actually in transfer in fabrication within the fab environment. The second one is chemical containers. Tim Hendry talked a little bit earlier about the size of the supply chain and how the purity of the chemistry from when it is actually synthesized and needs to be dispensed onto a wafer needs to be the same at all times. That is kind of like a farm-to-table restaurant. You want that head of lettuce that you got in the morning, picked off the farm, you want on your plate for a salad in the afternoon.
Well, if you think about this supply chain, there are times when that head of lettuce is 16,000 miles away from where it is actually going to be consumed. If you want to put a head of lettuce on a truck in San Francisco and deliver it here to New York in about one week, that is a new definition of a wilted salad. Really, the thing that AMH does is we guarantee that that head of lettuce is going to be the same when it gets pulled off the farm and when it gets put on top of your plate. Finally, in fluidics, these are all the piping, tubing, valves, sensing, photolithography pump technology that is used to actually dispense the chemistry from the drums that are shipped into the fab all the way up into the wafer.
On the fluidic side of things, we work pretty closely with the MC Division that Clint just talked about purifying and filtering the chemistries as they get dispensed on the wafer. We also work closely with the SCEM Division, which will be up next to talk about the chemistries and the process chemistries. We have an early visibility into the develop of those chemistries. We optimize the containers to be compatible with those materials. That gives us a unique advantage as a company. This business has a combination of unit driven, about 45%, and the CapEx driven is 55%. Most of our products are very physical, so if you walk into a fab and you have a bunny suit on in the clean room, you are going to see Entegris. You are going to see the boxes on the equipment front-end modules.
You're going to see them in the overhead transports. You're going to see the chemistry coming out of the drums. You're going to see all the valves, tubing, pipe, and fittings coming in. Those are all polymer-based. Our major manufacturing technologies are all polymer conversion-based capabilities. Moving to the next slide, there's really three dynamics for the business. I want to start off talking, though, a little bit about what does it look like inside the fab. The photo on the right is a FOUP, a Front Opening Unified Pod that holds the silicon wafers as they're fabricated in a factory. If you look at that, it looks like that thing's about 1 mile long. If you walk around in factories, you see these things everywhere.
There's over 5,000 interface points inside of the fab. The Entegris FOUP is used every step of the fabrication process for the silicon chip from when it goes from the wafer supplier to when it's cut up and bonded and put into the mobile device that you have on your phone, lives in one of these FOUPs. That gives us really unprecedented access to fab design and development. These are the two areas that we wanted to focus on. Over here in the capital investment supply chain, when a company decides to build a fab, one of the first things that they do is they bring Entegris in with the fab integrators because they need to know what type of FOUP are they going to use, what kind of valves, tubing, piping, fittings do you need?
What's the type of logic that you're going to use for processing and bringing your chemistries into the fab? We work very closely with the OEMs and chemical manufacturers up front. That's the capital-driven part of our business, is 55%. We're in early. We get specified in with the architectural and engineering companies, and that's what gives us the early entrée into the business. As the fab's built and you get into the life cycle of the fab, there are definitely opportunities for us to get new bites at the apple, so to speak, with new fab conversions, new technologies that are released. Really, the recurring sales component of this is really around the silicon supply and the chemical manufacturers. This is where we have our products that are already specified in, and we have the volume going into the fab.
From a market force perspective, there's two components here. They're a little bit competing. First of all, we talked quite a bit today about the advanced requirements of critical materials handling. Talked about 3D NAND, EUV lithography, 1X process nodes. There's a lot of work that gets done in R&D to make sure that we're developing products that are compatible and then really enabling those technologies. What happens when you win? When you win, you still have to be able to support that ongoing mature supply chain. You have to have the agility, you have to have the capacity to serve that business. You have to have the supply chain excellence to maintain that market position. Over 50% of our business is actually in this recurring sales component, making sure that we're supplying that leading technology node after we get the initial win.
From a market growth driver perspective, AMH is about a $1.1 billion sort of addressable market. It's really broken up into about a third of each of our major areas. What's a little bit unique here, in the fluidics business, this is all the piping, tubing, valves, fittings. This is mostly CapEx driven. When I talked about the engineering architectural companies being early in the fab build, this is where a lot of that revenue is derived. On the wafer solutions, it's a little bit of a combination. It's a combination of FOUPs, which are used inside the fab, and that's mostly CapEx driven. There's our ongoing polysilicon shipping box business, which is the recurring revenue, and that's all under wafer solutions. Finally, in containers, that's mostly going to be volume driven in the mature life cycle of a fab.
You get specified in for the incoming chemistry, and then you get the recurring sales as the chemistry is consumed within the fab. We have, in addition, wet tool shipments and an increasing number of process steps and complexity. Not only are we being driven by the containers for the chemistry usage. As we get to smaller and smaller technology nodes, we see more chemical usage, and that increases the sales of the container business. This is the first business segment, the wafer handling solutions. These are polymer boxes that are used for shipping and transporting polysilicon throughout the fab or in between facilities or from the silicon supplier to the actual fab. It's driven by a couple of things. On the chart on the right, shows the expanding global fab capacity. This is millions of inches of square silicon.
When a new fab is built, that's a major revenue opportunity for us for the initial production of the FOUPs. Once that fab starts to ramp and silicon's being used inside the factory, a lot of the revenue switches over to the consumable wafer shipping opportunity. A couple of things that are important to note here, for a in-fab product like the FOUP, we rely pretty heavily on the micro-contamination technology for filtration inside of the box. These wafers see all sorts of chemistries, and when they're put inside of the box, you don't want to contaminate the box. There's different types of methodologies that we use for either inert gas purge, gettering, and other types of cleaning and purity technologies to make sure that you always have a clean wafer inside of the FOUP.
We also rely heavily on our SCEM division on what types of chemistries are actually being deposited onto the wafer. The surface of a wafer is highly reactive. We need to know what the surface chemistry is, and then what the outgassing, metal contamination things will be on the FOUP, so that we can tune the capabilities of that product to make sure that it's always clean. On the container side, this is really driven by the actual increase in shipments of gallons of chemistry. Lithography materials continue to get more and more pervasive, same thing with CMP slurries and wet etch and clean technology. Clint said earlier that some of these materials are up to $10,000 a gallon. Those get shipped in our products. These are the containers that protect and transport those materials. There are a few major dynamics here that make us unique.
One is the safety. We have unique connectology or connection devices to make sure that we're combining the right chemistries to the right lines inside the fab. If you think about it, in one of these big facilities, there's multiple layers, and the very bottom of the fab, this is where the incoming chemistries come in. You have to know how to connect these correctly so that they go up to the right part of the floor and have the right chemistry. You can't cross-contaminate these chemistries, and safety is a really important factor inside the fab. The second thing around safety is that these are all transported materials, either air shipping or ocean shipping. A lot of them are high vapor pressure type of materials and pretty aggressive chemistry. We have products that guarantee the safety and the purity of that material while in transit.
Finally, the Fluidics section. Maybe I can talk a little bit about the graphic up here in the corner. This is a photo dispense pump, and this takes photochemistry and actually deposits onto the wafer in the most important step in the process, which is the photolithography track. We have a liquid filter, actually, that's embedded into this product. Again, this is a combination of having the materials technology, the dispense capability, but also the filtration to wrap that up into one big solution. The Fluidics business, as I said earlier, is all the piping, tubing, valves, fittings, manifold. It's a lot driven by new fab production. We'll talk a little bit about the growth rate of the business in a while. We had a big CapEx year.
That drove a lot of activity on the engineering and fab build side. That's a big driver for this business. Something that's fairly unique about this business is really the relationship that we have within the supply chain. There was some discussion earlier with Tim that talked about new architecture and how that drives microcontamination, how it drives new materials. What's unique is that the IDM actually goes off to the OEMs and the chemical manufacturers and says, "This is the kind of architecture I want to have in my facility." Entegris has a front-row and early visibility to this because of our relationship on the chemical side and with the OEM supplier. We get a seat at the table, and this makes us fairly unique. We also work with consortia like imec and CEA-Leti.
When you think about the periodic table that was shown earlier, now we have 45 different elements that are used in these different chemistries. They need to know early on what the microcontamination effect is going to be of that inside the fab. They want to know, are the drums going to be right? What happens with the tubing, the valves, the fitting? I can't use metal piping anymore because I have an element that's really critical to metal contamination. They bring Entegris in to understand the new chemistries when they bring up the architecture that we can help specify that and enable the overall supply chain. We've been talking about semi all day long. Here's an opportunity to talk about life science. We have some life science sales already.
One of the things that we're working on in AMH is how do we leverage the capabilities that we have and the investments that we've made into other environments or other industries? This is an example of a gamma-stable fluoropolymer bag that's used for single-use technology in life science. Life science is on a similar technology curve, where the sensitivity towards metal and organic contamination is getting more and more significant. There's a big increase in single-use type technology, a lot of the challenges are very similar to semiconductors. We're using that to launch this into the new business. We released this Aramus bag in the middle of last year. We work really closely with SCEM on understanding the surface technologies and the interaction of the bag material along with the chemistries.
We apply really a lot of the supply chain process control that we have for the high supply chain control and quality levels to really be compliant with the life science initiative. Finally, here's the overall financial performance of the division. We did about $385 million last year. That was a growth of 11%. Generally, we grow pretty much in line with the industry. However, as Todd actually commented earlier, we're putting our R&D in the right places. We actually outgrew the market because not only do we have that base business and we're growing with the industry, but as new fabs are built, new technologies are deployed, new technology nodes, we get to see a bump there. That's what really drove this component in 2016. The big margin drivers for us are leveraging our global infrastructure.
Like I said, we mostly do polymer conversion and manufacturing, we have a fairly large footprint. We really focus on plant efficiency, manufacturing initiative, really driving more volume through the installed base.
Thank you very much, Bill. Bill's business he started out talking about started 50 years ago. Bertrand likes to refer to it as the brand. If you go to somebody in the industry you're trying to explain what Entegris does, they don't really see any of the products, you say, "Oh, we make FOUPs." People are like, "Oh, yeah, okay, I get it." That's Entegris. Business, heavily dependent and really good at material science, particularly polymer science. Tim talked about that supply chain, you can't have a tubing or a wafer shipper that leaches or outgasses onto a wafer. That's what this business is really all about. It's about keeping those materials pure through the supply chain and in that fab environment. Stable business. We've got relatively high shares, most mature business, slowest growth.
The last thing I will say is this is a business that we are very focused on improving the profitability. Bill hears this about once a quarter or once a month that we need to do better here, it is a business that is capable of generating better returns than it is generating today. With that, I am going to turn it over to Stuart, who will talk about our chemicals, gas, and materials businesses.
All right. Good afternoon, everybody. Last division of the three to present. I want to talk about a couple of things. One is, earlier today, Bertrand gave you a broad overview of the company and also some of the macroscopic drivers for the industry. Some of the important things, whether it is artificial intelligence, autonomous driving, the cloud, some of those big macroscopic drivers. Tim talked about the technology which the industry is required in order to kind of meet that next generation need. Talked about 3D structures. He talked about the need for materials and how the design of that circuit is important for integrating those in. Right? What I am going to do is talk about how those forces combine with Entegris' unique capability and materials to propel our growth for the division. This Advanced Materials Division is focused on a segment of the greater materials market within semiconductor.
We broadly have organized these into four separate market segments. The first segment I will talk about today is our Specialty Gases Segment, which is focused on the delivery of high purity, safe gases, primarily for doping applications. The second will be our Advanced Deposition Group, which is focused on deposition materials, which effectively is what makes up the final circuit. The next one will be our Surface Prep and Integration Group. Sometimes semiconductor circuits there is an analogy to baking a cake, a multi-layer cake. Between every layer, you need to be able to prepare that surface. Our Surface Prep and Integration Group is really focused on that, and I will talk about that in some detail a little later. Finally, I will finish up with our Specialty Materials Group.
Whereas the three business segments prior are really focused on materials which are going into the circuit, our Specialty Materials Group is providing materials which are used in the processing of the material, but not the material itself. This business is almost 100% unit-driven. You can see that because it is a materials business. We do sell some ancillary equipment in order to enable delivery of material. More importantly, we work with the different divisions, Microcontamination Control, as well as our Advanced Materials Handling Group, to optimize those delivery solutions so that we are not just a chemical supplier, we are actually delivering materials in high purity to the wafer, to the device in its final form. I will broadly go over some of the macroscopic parts of the business with regards to materials.
I'll talk a little bit about how each one of those subgroups applies materials into the industry, then I'll finish up with a few important examples. We'll start with this cartoon. This cartoon is a picture of a logic device, you can think a microprocessor, transistor, and you can see that it's composed of many different layers. The bottom layer is typically called the front-end processing, where the transistor itself is fabricated, and that has a number of materials which we provide. Some of these materials are very toxic. We talked about the one drop of water going over a waterfall, and being able to purify that. Well, you can think about one drop of water coming over Niagara Falls in a period of an hour that could kill you. arsine is one of those gases that if it escapes, it will kill you.
We provide advanced safety solutions for delivery of arsine and phosphine. Other important materials that are inside the circuit would be the dielectrics in the junction. We provide those as well. There's a series of metal layers, which we provide the deposition materials for, whether they're tungsten, copper, et cetera. Between each one of these layers, like we talked about before, you have the need for both for cleaning and preparation. Whether it's conditioning that surface, which is done with some of the technology used for our CMP pad conditioners, or it's actually cleaning that surface with our specialty chemicals, those are done at each one of these steps. The takeaway from this slide is that there's a large number of opportunities where Entegris provides materials into the final devices.
This is a logic example, and later on I'll talk about a memory example as well. When we think about the materials market, it's a very big market. It's about a $24 billion market for the semiconductor market. Within that segment, there's a what we call advanced materials segment, which is depicted here, about $1.2 billion of our served available market. What's broken out here would be roughly how that segments with regards to the four business segments I talked about earlier. Gases, advanced deposition materials, surface prep, chemistries, then finally, our specialty materials area. The key takeaways from this is that advanced materials are growing for the exact reasons that Tim spoke about earlier. These 3D structures are requiring new materials, which are more difficult to both manufacture and purify than commodity materials, which may have been used years ago in simple planar geometries.
Those new materials are also very sensitive to the chemistries that are used to clean them with. New formulated chemistries are needed to be developed in order to enable those as well. You'll see that the three main drivers with regards to growth are around some of those common themes we talked about earlier today, which are node shrink, which is requiring different materials in order to enable the device performance, particularly speed. You also have the 3D structures, which makes it more complex both to deposit the materials as well as to clean them. Finally, we'll talk about growth, and we'll talk specifically about how 3D NAND is changing, 3D vertical NAND is changing the industry. We'll talk about the four different business segments. The first one is specialty gases.
This business segment is focused on delivering high purity, safe gas delivery solutions to the industry. A lot of these gases are not only toxic but also hazardous. We talked about Arsine, Phosphine. There are other dopant materials which are also very difficult to deliver. Safety in the chemical industry is number one. If you ever go into a chemical factory and they do not talk to you about safety first, please run. Safety is very important for the security of the employees, for the visitors, and in the end, for business. What we provide to the industry really is the leading safety solutions with regards to subatmospheric delivery of gases. Just to put it simply, if the gas is stored subatmospheric and held, then it will not escape.
If it's stored at high pressure, if there's ever a valve failure, you can have an escape of that gas, which could result in damage to the facility or loss of life. Entegris provides the best-in-class, and that's why we're kind of number one position in the subatmospheric delivery. The advanced deposition material group is focused on the materials which eventually effectively make up this integrated circuit. Those materials vary from conductors to insulators, and a few examples here are given, both for solids, liquids, as well as gases. These new materials, some of them are listed here, like aluminum chloride, cobalt, ruthenium, and others, are enabling the circuits of the future and enabling the industry to be able to go down to 10 nanometers, seven nanometers, and below. Many of these are very difficult.
They're actually solids at room temperature, and so they have to be heated up in order to be delivered. That's very difficult to do uniformly, and Entegris has some unique solutions, and not just the material itself, but the delivery mechanisms. We'll talk about a few examples a little later. For surface prep and cleaning, chemistries are enabling the structure by being able to remove contamination. Actually, some parts of the semiconductor processing can be quite dirty. Bill Shaner talked earlier about how contaminants can get inside of these microenvironments, and so we need to create microenvironments that can be purged to get those contaminants out. You also have contaminants on top of the wafer surface because of the preceding steps, and those need to be clean without damaging the structure.
The new materials tend to be more sensitive to structural damage, so you need what we call formulated cleans, or cleans which are designed to remove the contaminants while not damaging the underlying circuit structure and performance. We've talked about the areas we serve. We call post CMP or chemical mechanical planarization, and also after the etch process itself. These formulated cleans are actually growing much faster than the cleans area as a whole, we call commodity cleans area. That's because more and more of the more exotic newer materials are being integrated into device structure. Last one I'll talk about, the specialty materials is the group that provides materials which are not actually put into the device structure itself, but are used to make components in the chambers or coatings of the chamber components, and also used for other industries.
These materials are typically graphite, sometimes silicon carbide, and a number of wide variety of coatings. The example shown here is one of a cellphone. Some of the new cellphones are actually made of glass. They're actually curved glass. That glass is curved by a molding process. Glass, you need to go above 1,000 degrees Celsius in order to mold it. You need graphite to mold it. Our graphite has a very uniform structure, which allows you to mold that curve on the glass without scratching it. That's why I say it enables high-resolution glass forming like you see on some of your higher-end mobile handsets. Our materials aren't limited to that. A number of our materials are being adopted into chamber components. The contamination can come through chemicals that come into a chamber.
They can also result from the walls of a processing vessel causing particulate or metal contaminations which wind up onto the wafer surface. One way to abate that is to put special coatings onto the chamber, which ensures that the chamber itself does not contaminate the wafer surface. That's another area that's growing fast. As we talked about the etch market is growing. That area is growing faster than the industry itself. In both of those segments, we're a market leader. I'd like to talk about a couple of opportunities that we're working on now, which are industry trends. The first is in the area of 3D NAND. 3D NAND or vertical NAND, Tim talked about earlier as being an important trend in the industry.
It's tied to cloud computing, larger and larger SSDs that we put into our computers at home with the insatiable number of photos that we take. What's happened here is that you can think of this as a car park versus a multi-story garage. In a plainer parking lot, you can only fit so many cars. If you want to put more cars in, you build a parking garage. That's what this is. This is the equivalent of a parking garage where you have 64 alternate steps. Entegris is providing the deposition materials which go into these steps, both for the dielectric layer, insulating layer, as well as the metal layer, as we call wordline, and a lot of the other smaller layers, which would be high-k dielectrics in the structure that forms this single memory cell.
The key takeaway from this is there's more materials which are being put on per square inch. Those deposition materials, in many cases, require new types of precursors. Entegris is really leading in that, not only in delivering the materials, but protecting the wafers by chamber coatings around the chamber elements. Second one I'll talk about is the fluorine-free tungsten. One of the applications of the industry is using tungsten in the metallization step. Tungsten usually comes with fluorine, WF6. Fluorine actually can damage the structure, particularly dielectrics. In cases where you need to avoid that, you need the ability to have a fluorine-free molecule. Entegris has developed a molecule which is fluorine-free, actually a number of them. What's unique to Entegris is that not only can we synthesize these molecules, but we can also purify them.
We have special delivery mechanisms for delivering solids, which is actually very difficult. These materials are very corrosive in themselves, and they will corrode the vessel that they're in. We're able to utilize the Entegris capabilities in coatings to make sure that material is delivered in a pure form all the way to the wafer surface. To make sure that the stability is there, we also have some analytical capabilities to enable that as well. It's really Entegris capabilities in delivery coatings as well as chemistries that enables us to take advantage of this trend, and this is both present in 3D NAND applications as well as in advanced logic where tungsten integration is critical. Finally, just to summarize on the business, this business is about $28 million in this past year in revenue.
It's been growing slightly above the CAGR for the industry as a whole in terms of MSI, and it's really poised to grow significantly at a significantly higher rate. We talk about in the 200, 300 basis points based upon the growth of those key drivers in 3D NAND and advanced logic. Margins will improve both due to product mix as well as being able to leverage our supply chain as more volume for those advanced precursors take hold. Thank you.
All right. Okay, thank you, Stuart. Key highlights on that business, essentially a unit driven business, 100% unit driven, high recurring revenue stream, again, sticky products. It's a niche-y business, though, when you peel it back. Stuart talked about his SDS system, his gas business, and how it's around safety for some very toxic gases, that we're not competing broadly with an Air Products or a Praxair, but we're doing very well on a relatively small niche. Same thing, he talked about formulated cleanings. We're not participating in the spot market for cleaning chemistries. We're focused on the advanced side of that business. The last thing that he spent time on is this is our business that's probably the most leveraged to 3D NAND. We get a lot of questions from investors about, what do you see in 3D NAND?
Which of your businesses are exposed to 3D NAND? This would be that business. With that, I'm going to switch over and talk a little bit about our finances, and really, Bertrand gave some history. I'm going to focus, excuse me, primarily going forward. Before I do that, I want to just acknowledge two people. You've heard from everybody from Entegris that is here today, but Jolene Bergner, who helped you with the registration, countless hours putting this all together, and Bruce Beckman, who's in the back, who's our VP of Corp FP&A, who's also played an important role in putting this together. Starting out with our objectives, just to crystallize what Bertrand talked about, our goal is to grow 100 to 200 basis points more than the market. The market for us, 80% semiconductor unit production, 20% capital.
Continue to expand our EBITDA margins from the 22% we had last year moving forward. I'll talk more about that. Historically, we've been a strong cash flow generator, continue to drive higher cash flow levels, and then grow that bottom line or that EPS line at twice the rate that we grow the top line. From an investment perspective, you can expect to see us to continue to invest in R&D at the levels we're investing today. We're going to continue to invest in our internal capabilities. I look at what we've done with the investments we've made in some of our processes, like statistical process control, like some of our process around R&D selection, and it's paid off. You'll see us continue to do that. Obviously, acquisitions are something that we've got on our radar screen.
The last piece here from an objective standpoint, we will continue. We've used the target model since 2009, and we'll continue to use the target model as we operate the business, and I'll lay that out for you today. When we think about the three divisions, I just want to give you from the CFO's chair how I think about this portfolio. It's a relatively balanced portfolio. You think about the Microcontamination Control business, I would clearly put it in the star category, high growth, high margins. We've invested heavily in the business. We have high expectations for continued strong growth and continued margin expansion. The AMH business, Bill's business, that's our legacy business, relatively limited investment going forward, and we will continue to invest in FOUP technology, which has been the cornerstone of that business.
We're not expecting a lot from that business from a growth perspective. Roughly market, maybe 100 basis points more than that. View this as a strong, it's historically been, and we would expect it to continue to be a strong cash generator. Stuart's business, the SCEM business, kind of falls somewhere in the middle. We are investing pretty heavily in that business. A lot of what we invest there is for growth. We're investing in our cylinder and canister fleet. We're investing in deposition materials, and we'd expect that business to grow a couple hundred basis points in excess of the market. Talking about growth and our relative growth, this is my favorite slide in the deck. On a relative basis, the slide on the left is how did we do in 2016 versus our market and our peers.
Our market, again, I talked about that 80% units, 20% CapEx. If you put that together, that's about a 4% blended market. We were 8.7%. We did have a little bit of a tailwind from currency, but even without that, we outgrew the market by about 400 basis points versus our company A and B, other specialty chemical companies that have a strong component of electronic materials. Chart on the right is just how have we done since the ATMI acquisition, so a two-year CAGR. Market's grown about 3%, we've grown about 4.5%. Really right in line with that 100-200 basis point commitment to outgrow the market. Our profit growth path, where do we go from here? The column on the left is where did we end 2016?
I showed a chart like this in the middle of 2014, right after we did the ATMI transaction, and I did a walk to 2016. It looks a lot like what we reported. But in 2016, $1.175 billion in revenue, 22.4% EBITDA margin, $0.94 in earnings per share. Bertrand talked about a 5% top-line growth rate based on an industry that would grow kind of 3-ish, 3% to 3.5%, and us outgrowing that industry by 100 to 200 basis points. 5% on a CAGR basis, that would take the $1.175 billion to $1.36 billion in 2019. That 5% growth, our internal planning is based on a 3% industry number. We've put external forecaster's industry numbers in the appendix of the deck just for perspective. That 3% is a relatively conservative number.
We've all been in the industry a long time, we're going to play on the conservative side. From a margin improvement perspective, take that EBITDA margin from about 22%, a little over 22%, up to the 25%-26% range. What does that do? It takes us from $0.94 in EPS to $1.25-$1.32. That is assuming that all the cash we generate, we keep it on the balance sheet. It doesn't take into account capital allocation, and I'll talk about that in a couple of slides. Target model, we're going to an annual target model in terms of how we guide the street. You see the 2013, you see where we were, about 19% operating margin. 2016, we were at 22% operating margins or EBITDA margins.
As we move out from here at $1.2 billion, we'd expect that operating margin to be 24%-ish. Improvement from where we were last year at almost 22%. At $1.3 billion, 25%, and at $1.4 billion, kind of 26%. This is really consistent with what we've talked about before. We talked about kind of a 40% flow through. If you were to take the midpoint of all of this, it reflects about a 40% flow through. That takes us on an EPS basis from $0.94 in 2016. At $1.2 billion, we think $1 plus or more. At $1.3 billion, call it $1.15 plus. The street today, if you look at estimates, they're kind of somewhere right in the middle, and the estimates, the consensus is in that $1.07-$1.08 range.
Really lining up with what we view as kind of our longer-term target model. We talked about cash flow. Historically, we've generated very strong cash flow. Last year, the operating cash flow was over $200 million, and the free cash flow was approximately $130 million. As we look forward, we think the business over the next three years can generate operating cash flow of about $700 million, or free cash flow of about $400 million over the next several years. That gives us a fair amount of capital to work with as we think about our capital allocation strategy. Where are we going to invest? We're going to really invest, I talked about continuing to invest in the P&L, ER&D, and continuing our trends there. From a capital perspective, for 2017, we'd expect to invest $80 million-$90 million. Same kind of number in 2018.
As we come through this heavy investment cycle, you should see that come down a little bit as we get out toward 2019. No surprise, the big investment initiatives, filtration. Within that business, it is about membrane capability, it is around cleaning capability, and just in general, continuing to drive the capabilities of the MC business forward. Stuart's business, the chemical and gas business, the largest area for investment there is going to be within deposition. As I talked about, we will also continue to invest in the fleet of cylinders and canisters to support the overall growth of that business. The other area we are investing in that business is the coatings business, which Stuart talked about. In the AMH business, we talked about the next generation FOUP and also updating some of our molding capabilities. We continue to have a very strong balance sheet. We have $406 million in cash.
$140 million of that cash is in the U.S. We have brought our debt down to a little more than $560 million from over $800 million. Excuse me. We brought our debt down to $590 million from about $860 million at the time of the ATMI transaction. We have brought it down relatively significantly. Just last week, we repriced our debt, so the notes stay the same at 6%, but our term loan went from LIBOR plus 275 to LIBOR plus 225. That will have a modest impact on our interest cost going forward. We do intend right now, our capital allocation strategy calls for us to continue to pay down about $100 million in debt this year. $25 million in each of the quarters this year. Let's talk about our capital allocation strategy.
If you think about our options from a capital allocation perspective, on one end, we can stay really flexible, pile cash up on the balance sheet. I think ultimately our investors are not going to let us do that, but that would be an option. At the far end, in terms of driving EPS growth, we think that M&A is likely to drive the greatest proportion of EPS. That said, we have also got good options in the middle. It is not critical that we do M&A to continue to drive our EPS. What this chart shows, on the left side, we accumulate cash on the balance sheet, our capital allocation is going to have no impact on our earnings per share.
If we just continue to pay down the debt that we have got outstanding with the free cash flow that we generate, you will have about $0.10 impact over the next three years. If we were to buy back shares, assuming we buy back shares at a forward price in the mid to high 20s, it would have about a $0.14 impact on our earnings per share over the next three years. M&A, about $0.16. Let me just talk a little bit about some of the assumptions here. On the left-hand side of the page, that $1.25 to $1.32 is what I showed you a couple slides ago on the operating EPS walk forward. The debt, it is just paying down the debt that we have at the interest rates that we currently have.
I talked about the buyback assumptions and the M&A assumptions, we assume that we could acquire at about 10x EBITDA post synergies. The capital that we're employing here is essentially that free cash flow times the portion that is domestic with $200 million of repatriation as well. This doesn't reflect specific targets. At one level, it's just math, you can do the math based on the footnotes. That would take us, Bertrand showed early in the deck, 2019, kind of a $1.40 number. We show $1.35-$1.48 here based on the operating EPS walk I showed and our capital allocation choices. The other thing that I would point out here, though, is this doesn't assume any additional leverage.
Obviously, if we had an attractive acquisition come along, we'd be able to take our leverage back to kind of that 3.5 level, and we'd potentially do better than this. What should you take away from the meeting today? I think we believe we're well-poised for continued growth on the top line. We're in a multi-year cycle within the industry. The industry is a good industry. It's less volatile than it's been historically. It's driven by more than just sort of PCs and handsets, the number of applications continues to expand. Materials are absolutely critical, there's really nobody better when you think across the material supply chain than we are. We're pervasive across the fab environment. We've got a broad range of products.
We think, in the end, we're well-positioned to continue to grow the business on the top line, grow EPS at twice where we grow the top line, M&A could put potential icing on the cake. From an M&A perspective, we look at what we did through the ATMI process, where we've essentially, I talked about 17% CAGR in EPS since 2013, which was the year prior to ATMI. We think we're good at it, we're good at integrating, we'll continue to look at that as a strategic option. With that, we've talked a lot. We haven't taken a lot of questions. I'm going to have the rest of the team come up, we'll take questions from the group. Bertrand is going to moderate that, we're willing to take questions for any one of us as well as for Tim.
Thank you.
I'd like to ask if you do have a question to use the microphone.
Thanks, guys. Amanda Scarnati from Citi. The first question is for Greg on kind of the operating margins. If you could just kind of walk through that a little bit. You look at the margins of the three different business units, they seem to be growing by an estimate of 2%-4% target-wise, whereas the overall corporate averages or margins are growing by 1%-3% in the targets that you put out. Just kind of talk about what's happening in the margins there and what's happening on the corporate side that's kind of pulling the margins down below the targets in the three groups.
The targets in the three groups. The whole model rolls together. We've got relatively aggressive targets in the three groups. We'd expect the corporate spending to be relatively consistent. I'm showing operating margin expansion in the businesses, and we're showing EBITDA expansion when we talk about 300 basis points at the corporate level. The delta is really around
The EBITDA takes into account relatively significant CapEx investments that we talked about. That's why you get 300 basis points of operating margin and a couple hundred basis points of EBITDA margin.
Okay. Bertrand, just to talk about the China strategic partnership that was announced a week and a half ago. Can you just talk about the benefits of that and how that's going to help to grow Entegris over the next year or two?
Sure. I'll take the first stab at this question and probably turn to Stuart, since this partnership was really established to enable some of his business units. As we mentioned, China has been a very important market for us. We have been growing very steadily in China, 2015, but also in 2016. We grew about 22% in China. Our business has been very strong, both with the foreign investors in China, the large, traditional international semiconductor players, but also increasingly with the emergence of new indigenous players in China. We really wanted to be in a position to better support what we expect to continue to be a significant wave of investment and greater activity at fab level. Some of the chemistries that we are developing are not necessarily traveling very well.
In addition to that, some of those process gases and other materials that we're making can be subject sometimes to very continuous regulations. In order to still convene all of that and to put us in an agile position to support those Chinese players, we decided to structure a partnership with Spectrum Materials. Stuart, maybe you want to add to that.
Yeah. I think Bertrand mentioned that the Chinese market is growing very rapidly. Expansions are planned in this year, next year, and the following years. Being able to support our customers in the region is important. I think Tim Hendry actually mentioned the fact that many of these chemicals cannot be shipped by air. They have to be shipped by boat. If you think about the amount of time it takes to ship a chemical from North America to Asia by boat, you wind up with a very long supply chain. Being able to shorten our supply chain reduced our cost and actually increases the surety of the chemical arriving on time for our customers in China. It's a strategic move for us to better serve our customers in China.
Does Spectrum have any control over IP, or is Entegris managing the IP relationship as well?
The relationship with Spectrum is they're what we call a toll filler. We own the canisters, we own the technology, we actually even own the equipment that's used to fill the cylinders or canisters. They're operating the filling process with their personnel on their side.
They're essentially acting as a contract manufacturer for us, and we are maintaining the direct relation with the end customer.
Yeah.
Okay.
Thank you. Patrick Ho from Stifel Nicolaus. Bertrand, maybe a big picture question for you, looking at the business and how it's evolved to this point. Given a lot of your relationships with your customers are becoming deeper and deeper, particularly in the development stage, what's the challenge for you in driving more value proposition and particularly where I'm focused on, and this was on the gross margin side. You've increased the operating margin since the ATMI deal, how do you extract more value, particularly as you work closer to some of your biggest customers?
Well, it is in the constant balance that we're trying to strike. At one level, we do believe that all of the new solutions that we're developing in collaboration with our customers are increasing in value, enabling them to advance the performance of their chips, improving their yields. We're trying node after node to capture more of that value. Having said that, we also recognize the fact that without this very intimate relation with those customers, we will not be able to develop the next generation technology. There's this happy symbiosis, if you want, in the ecosystem, when we're going to try to capture as much value as we can.
We also need to be cognizant of the fact that we need to maintain and preserve the strength of our relation with our customers so that we get invited to participate on their next process technology challenge. I think we've been able to do that well. I think the proof of that is that the fairly steady margin profile that we've been able to guarantee across our business lines and the steady expansion that we've been able to deliver at the bottom line, in spite of the increase in R&D spending that we've done, in spite of the many investments that Todd was describing in terms of new tech centers, new metrology capabilities, and new talent that we've been adding to operate those tech centers. Again, I think that we're seeing, maybe it's another way to say that we're seeing the benefits of scale.
That was, if you remember, that was actually one of the reasons why we acquired ATMI. We felt that the size of Entegris back in 2012, 2013 was not as sufficient to allow us to be able to, again, live up to the expectations of our customers. I think now we have the appropriate scale, I think we have demonstrated that we can put that to good use.
Great. Maybe as a follow-up for Greg, in terms of the total corporate operating model that you highlighted today. We saw the different business groups give their target models and especially some of the three-year targets that you mentioned. One of the things I noticed was operational efficiencies in all of them. Can you give specifically, I guess, what could drive additional operational efficiencies? Because you have had good operating margins growth. What else more can you extract, say, over the next three years from each of these business groups?
If I go business unit by business unit, I would say in Stuart's business, in the chemical and gas business specifically, that's our business where our assets are the most underutilized. Volume alone is going to play a meaningful role in improving the profitability of that business. I think you move to the AMH business, we've owned all of those businesses and operated those businesses for a long time. It's the first time they've sort of, in the configuration they're under today, have been under one senior leader who has actually been at the company a long time. Our view is that there's just opportunities for additional efficiency there. We view it as the cash cow, additional efficiency, more probably on the operating expense line.
Clint's business. That's our highest growing business. We're going to continue to invest heavily, but it's a function of volume leverage as much as anything.
Hi, Ed with Company. Greg, Patrick, question, actually, I'll ask a follow-up. Entegris, combined with ATMI, has quite a number of facilities, right? Around the world, frankly, right? You guys talk about, for example, your current facility investing because you want to localize that. Any thoughts about site consolidation? Does it make sense to do it? Or you think you can grow to your footprint and therefore no need for that? Any thoughts on that? I have a follow-up one to.
What I would say is two things. One is we're constantly looking at our cost structure. We're constantly looking at the footprint. If you think about it across businesses, though, most of our facilities are running at very high levels of capacity utilization. Most of the floor space is used up. I would say I talked about it when I answered Patrick's, with the exception of the specialty chem and gas business. There, I think over time, we'll consume the footprint. That doesn't mean we won't consume it, meaning we'll grow into it, that doesn't mean we won't continue to sort of look at opportunities to be more efficient on the footprint front.
Okay, great. That's helpful. I guess the two last questions on R&D. First, I guess just staying with you, Greg. R&D, as you've shown, has been staying at nine, 10% of your sales. Is that kind of the optimal level? Do you see an increase in that? You talk about obviously leverage on your model as you grow your top line. Should we expect R&D to stay that level, but leverage all mostly come from sales and marketing? How do you kind of think about that? In terms of where the R&D is spent, if you don't mind me asking a follow-up question. I think you talk about a quarter of it on platform, a quarter on kind of disruptive, and half of it being almost on ongoing business, right? Do you see that mix of spending changing?
Do we expect more maybe platform type or disruptive technology down the road?
Yes, I can take maybe the first question, Todd, you want to take the second one. We will constantly look at the appropriate amount of R&D. We know that there's a cost to be the technology leader in the various product platforms that we have chosen to have in our portfolio. We want to be spending the adequate amount of R&D to be viewed as the technology leaders by some of the most demanding customers. Right now, based on everything we're seeing, we believe that nine to 10% is probably the right number. If we need to spend more, we will spend more. We will do that by making choices, as we have always made.
When we increased the level of R&D spending, we did, to your point, look for efficiencies elsewhere in SG&A, we were able to increase R&D without really compromising on the overall commitment we made in expanding our bottom line. You should expect us to continue to do that on an ongoing basis. I think that's really the quality of the execution that we keep referring to. In terms of the makeup of the portfolio, I turn to Todd, you should expect us to also constantly look at the optimum shape.
Yeah, if you looked at that pie chart a few years ago, you would've seen a higher percentage from derivatives, more than half. We've actually evolved it quite a bit towards the more advanced solutions because that's what the industry's been demanding. I think the balance actually is right where we want it right now. To be a responsible supplier to this, to all the players of the industry, we're always going to have derivative work to do to support the fabs day-to-day, as well as the kind of backward-looking part of the business, the core of the business. We have the right balance in that pie chart that I showed you. I'm very comfortable with that and the number of projects, to continue delivering the results that we're getting from that decision we made on R&D a few years ago.
I feel like it's the right balance. I don't think it'll change much more.
Management describe how they're incentivized in relation to total compensation changes in stock price.
The question was about management incentive compensation.
We have a base salary. We have a variable component on top of that ranges from anywhere between 50%-100% for myself. You have a stock-based component. Let me describe a little bit more those last two components. The variable plan is primarily driven by our bottom line performance, and that accounts for about 75%.
Bottom line meaning?
Means EBITDA, and that's 75% of the plan. The remaining 25% would be a series of corporate objectives, some of which have to do with top-line growth, some of which have to do with new product introduction, and some of which have to do with operational performance, such as quality levels, on-time delivery, inventory turn. Finally, the stock component, the equity component, is a mixture of RSU, performance shares, and stock options. Performance shares, it's really TSR. We're tracking the semiconductor, the stocks index.
Do you want to do stock buybacks? Because I recall stock buybacks.
The question was, some of the returns that we hope to achieve will be through stock buybacks. The question was really challenging us on the fact that there haven't been significant stock buybacks in the past. You're correct. There was a significant stock buyback.
10 years ago.
10 years ago, roughly. Since then, we had some small opportunistic stock buybacks. You're right, none of any magnitude. Recently, we have committed to have an ongoing stock buyback program to eliminate the dilution coming from stock-based compensation. You should expect that to continue going forward. I think what Greg was suggesting is there may be a point in time where it would be the right decision for us to commit to a larger stock buyback. I think Greg prefaced that comment by saying we continue to believe that M&A is the best way for us to create shareholder value. We have demonstrated that with the ATMI acquisition. We believe that we are a very effective integrator and a very effective acquirer.
If we are not able to find the right target at the right price, down the road, we may have to actually revisit the prioritization of our capital allocation options. That's really what's behind the comment that Greg made.
How long have you been acquisition targets? Will they be lateral or horizontal?
The question was, what type of M&A targets will we be looking for? Will they be lateral or horizontal? I want to be sure I answer your question. What do you
Do you add on to existing
I think you should expect actually a little bit of both. In some cases, we believe that we could benefit from adding to the existing technology platforms. You should also expect probably most of the M&A focus to be about adding to the portfolio and find technology platforms that will be very synergistic with the existing technology portfolio.
Cabot understand why that
Well.
change and what breaks or have the material change for it. Do you have a standard for it? There was more identification.
The question is about Cabot's acquisition of NexPlanar.
NexPlanar. We were actually a shareholder, as you probably know, of NexPlanar. We didn't feel that NexPlanar would necessarily be thriving as part of our portfolio, as a result, we decided to not express any interest for those assets. I want to believe that this acquisition is probably going to be benefiting from the more intense focus on the CMP process that Cabot is providing for those assets.
Take the next question from Christian Schwab.
Thanks. I want to follow up on the M&A question first. I have one quick follow-up. On the M&A front, you said that you would look for possibly new areas. I just wanted to clarify. Let's say microcontamination, your fastest growth, highest profitability, lowest market share. Maybe logically you would look there. Would that be to expand market share in served markets, or would that be to go after underserved markets? If you could give us some clarity there.
Some of you know that I started my career in Mykrolis. I probably have a great affinity to separation technology. Yes, I would love to find ways to add to our filtration and purification business, whether it is to support existing applications in our core semi markets, also potentially look at broadening our market focus. I would say we would look at both, it's going to be a question of actionability and affordability. Putting that aside, I would also say that, I think you've heard that from all of the presenters, we tend to like, as a management team, recurring revenues. You should expect us to probably also be primarily interested in unit-driven types of businesses. That's the other area of focus for our M&A strategy.
My last question is for Greg. Could you quantify 2016, what your exposure is to 3D NAND, and elaborate on what the potential of that could be as we transition the rest of the way for starts?
I really can't do that, I mean, specifically, but Bertrand or Todd can.
It is not something we can easily do because what we sell to the companies who are making 3D NAND use our products, similar products in other processes. We don't always get that division. Clearly, though, as Tim described, that form factor and what they're trying to accomplish there requires a lot of the kind of things that we do, be it for layering or for deposition materials or clean. Definitely driving an important part of our growth, especially as we talked about the growth in deposition materials last year. The rest of it starts to get a little more blended. FOUPs and filters and so forth kind of go across all processes. We like the growth in that because complexity in the chip, be it logic or 3D NAND, drives needs for our products.
Chris, a specific process is always hard for us to answer. If you go to the pie chart that Todd showed, only about 55%-60% of our revenue is direct to a fab. The balance of it is to a materials supplier or to an equipment maker or a chemical supplier. It's tough to say exactly.
Chris Kapsch with Aegis Capital. I wanted to drill down on the operating margin target and the progression. You emphasized both, I guess, Bertrand, in the opening remarks, and then Greg, in your summary and the Q&A, that a lot of that delta is driven by leverage associated with just greater sales volume. Clearly, as the industry is progressing along Moore's Law and the increased complexity at the advanced nodes, you're presumably getting better sales mix contributing too. I'm just wondering, how important is leverage and how important is mix in that progression, and how dependent is your model on the progression to the advanced technology nodes actually staying on track?
What I would say is it cuts both ways. I mean, leverage is very important to it in volume and the new product growth. If you look at, Todd showed 30% of our business is new products. Our margins in that business, as you might expect, that new products are higher than in our legacy businesses. At the same time, some of the products that have been around a long time that are supporting the Internet of Things, we continue to do well in them, but we do experience some price degradation. We plan for 1%-2% of annual ASP erosion.
Okay, if I could just follow up with Tim. Tim, thank you for your comments, there was an interesting chart that you emphasized talking about the yield degradation that comes with these advanced nodes. You almost suggested that maybe the node transition being delayed is inevitable unless something happens to improve the yields dramatically. I think at 10 nanometers, it looked like a 60% yield, and going well below that for seven nanometers. What's your crystal ball say about how the industry plays out in terms of progressing to these advanced nodes? Is it just that there's going to be fewer winners that are capable of actually making that transition based on the technology and the complexity required? Thanks.
Yeah. First of all, that was a hypothetical of degradation if you didn't make any improvement based upon 28 nanometers. Clearly, the industry's making improvements in terms of overall yield. I don't expect that degradation to occur. The complexity of number of steps and number of problems that we have to solve as you move forward is really probably the issue I think that will cause some delay in the nodes. You just have a longer cycle time. You have to figure out how to do development across that, and your number of information turns simply is going to go down. There's things you can do to increase your learning rate by segmenting the process. That's one factor, is just purely the length of the process itself has a force of how many information turns you can get within a given development cycle.
That's one force. The other one, you have to solve these yield issues before you go into a ramp. Otherwise, it's economically very impactful. The rate with which you can solve these problems is key, and that's why I emphasize speed of execution and new development of materials. The companies that can actually solve problems faster are going to win business because there's just more problems to solve in a given period of time. You do the math. That's going to be another factor that I think will put pressure on node transition. The other is, I think companies are going to want to extract more value out of the investment that they've made for a given node. The given node investments are going up. You're going to want to extract more value from that given investment before you move on to the next node.
I think there's several factors that will put pressure or headwinds, as I call them, on node transitions. There's still a very strong economics for making those transitions. There'll be immense amount of pressure to try to figure how to expedite those transitions and solve problems faster, work with partners more closely to develop more integrated solutions. I think Moore's Law is not dead. It's just going to take a little more time, I think, to solve some of the problems. You certainly want to be ready to launch with a well-characterized process that you can commit to customers in high volume. That's the thing, you just got to be able to commit. I think that in order to have that predictive schedule, you might want to exercise a slightly longer node transition to just have that predictability that customers can count on.
Changing a schedule is very impactful to your customers that have planned platforms around a given product and a node design. You want to be predictable, and you don't want to disappoint your customers. That may drive some elements of pushing out the node, just to be careful. That's why I know that's not an Intel view. That's just my personal view and my experience. I've done about eight different technology transfers, and you really need to make sure that you have a very good predictive and solid technology base before you commit to the outside world. I don't know if I answered your question.
Hi, this is Toshiya Hari from Goldman Sachs. Bertrand, thanks a lot for the day. Very helpful. You guys are positioned pretty well across the spectrum when it comes to your portfolio. I think your SAM today is about, as you pointed out, $3.8 billion, in a TAM that is maybe north of $50 billion. You do address a fairly small percentage of the overall market. I totally appreciate your focus on technology and your focus on margins. On a three to five-year view, how big could that SAM number be as a percentage of the overall TAM?
Thank you for the question, Toshiya. I agree with you. I believe that we have a very diversified product portfolio that should be leverageable to other markets and applications. We have characterized a few of the initiatives that exist and that we are sponsoring organically. It's probably fair to say that if we want that number to grow significantly bigger, we probably would need to supplement the organic efforts with some well-targeted M&A. Simply because when you work with very demanding semiconductor customers, it's extremely difficult for those research teams and new business development teams to really find a mind share to appropriately focus on the other markets.
We have actually carved out some dedicated resources and organizations. I think at some point in time, once we know that some of those other non-semi opportunities are real and are really exciting and could be actually a lot larger than what they are today, we probably would have to make some meaningful investment. We'll talk about that when we are ready to talk about that.
My follow-up is for Tim. Thanks for spending time with us today. You've probably spent a significant amount of time with Bertrand and the broader team. I was hoping we could get your unbiased view of Entegris. Maybe the two or three things that differentiate the company from the rest of the industry, and perhaps the one or two development areas for the company. Thank you.
One is the speed of problem-solving resolution. I think being able to integrate packaging technology, the filtration technology, and have a knowledge of the chemistries allows one to solve problems and present solutions more rapidly. The number of problems you have to solve is growing, if you can partner with somebody who can be more efficient and solve those problems quicker and requires less of your own internal engineering resources, that's a plus. That means I needed to devote, when I was with Intel, less engineers to solve a problem with Entegris because they had the ability to integrate across that. That's kind of very important. The other one, I think, is the agility to expand capacity and broaden it across a global base. Forecasting demand in the filtration business is not easy.
I never knew how long a filter was really going to last until I got into production, then I could set my expectations around that with the R&D team. That sometimes requires the ability to ramp very quickly on unforecasted volume. I loved it that I could ask them, and they would deliver, right? They had the foreseen investments to be able to do that. Other companies maybe didn't invest as much in that ability in the future, therefore they just, basically, couldn't meet my demand. To tell one of my factory partners that I couldn't deliver, that just was unacceptable. They go, "Tim, go figure it out," right? Working with Bertrand's team, they've rarely disappointed me when I ask for a lot more. That's very important.
I think the ability to integrate and expedite solutions and having also the agility to meet unforecasted demand, those are some key attributes. That's what I think, that's about supply chain, right? As well as technical expertise.
Yeah. You mentioned you guys are running the ATMI acquisition. Just curious where you guys are comfortable with the state to do that?
Let me.
I want to-
Let me take that one. First of all, I think, Greg was really using a number of really plain vanilla assumptions to build the model.
Right. Yeah.
Don't take that number as a hardwired number. I think it's really, again, stressing the type of assumptions that we use to build the accretion model. In terms of the level of leverage that we'll be comfortable with, at the time of the ATMI acquisition, when we announced the deal, the gross leverage was about a little over three.
A little over three.
At the time, we said that we would have been comfortable pushing it to probably about 3.5. That's, I would say, at high level, the framework that we still have. I think, of course, all of that would depend on the types of target, the nature of their business. We probably would have a different answer if this is a business that is more CapEx driven as opposed to a business that is more recurring in nature. A number of factors will probably impact the answer to your question.
If you were to find that.
Potentially.
I mean, ultimately, we've got to generate favorable accretion, and we've got to generate a favorable ROIC over time.
Mehul Trivedi, Cooke & Bieler. My question is related to CapEx. A couple of years back, you invested in the i2M facility. ATMI had expanded into Asia, so you would've kind of expected CapEx to kind of trend down, and it did a little bit, but it seems like it's going back up again. Can you speak to, is that because you've already filled the volume that you added as part of the facility? Or is it because you feel more confident around some of the growth initiatives and the traction that they've gotten, so you've got to invest in the sort of CapEx for potential sort of growth out there?
I would say our biggest CapEx initiative, there's really literally nothing. There's a very modest footprint expansion in that CapEx dollar. Most of it is capability driven, so it's greater capabilities. It's a new membrane system, meaning it's a membrane that we don't make today in MC. It's expanding our capabilities in deposition materials. It's expanding our coding capabilities, all places where we've got good growth ahead. The footprint really has not changed since we did ATMI. I would say, I think your question's fair. What I would say is, sitting in my chair, investing in some of these capabilities, frankly, has cost more than I would've anticipated two or three years ago.
Yeah. Hi, this is David Silver from Morningstar. I had kind of a big picture question about M&A, but maybe from a different perspective. When I look at the newspapers, Toshiba may be selling its business in the semiconductor equipment area. There's been a number of M&A transactions, and I think Tim has implied that working with fewer suppliers is better than maybe working with more numerous suppliers, even if the overall capabilities are the same. Is it your opinion that the bigger picture consolidation in the industry is going to drive similar consolidation momentum at the materials and supplier level that you operate in?
I can take that. I would say that, yes, certainly we've seen a fair amount of consolidation among the semiconductor manufacturers. Frankly, I think that this has had some really positive impact on the industry. I think that those consolidated platforms have been behaving a lot more maturely in terms of capital allocation, in terms of really trying to find the right balance between supply and demand. I think that's one of the results, probably why we haven't seen in a long time some of those very nasty cycles in the industry. I think that consolidation has had a positive impact on the semiconductor manufacturers and, again, positive impact on the overall industry. I think it's going to drive us certainly to do the same as an ecosystem.
I would tell you again from our practical experience, when Mykrolis came with Entegris, it was because of the realization that as two, three to $400 million companies, we could not afford building the right technical centers in all of the major geographies. Number 2, we were not even very attractive as sub-scale platforms to attract top talent away from some of our customers and establish, again, the application knowledge and the process knowledge that we needed to win in the marketplace. We have actually went one step further with the acquisition of ATMI, I will tell you that as a $1.2 billion company, it's a lot easier to justify making those investments without, again, compromising on the financial results of the company.
I would say that yes, we need to continue this virtuos cycle, you should expect us, again, assuming we can find the right targets, to play a role in that industry consolidation.
I think we have time for one or two more questions.
You received a bunch of questions on the M&A, just as we think about M&A as kind of the best way to complement the organic initiatives more dramatically expand the SAM, we kind of focus on the platform and scale, which I think was highlighted more this time than in previous years as a competitive advantage. If I take a step back and go to the capital allocation chart in Greg's presentation, when I look at the left end, how do you kind of balance flexibility and patience? Clearly, in the short term, it's not as enticing as the $0.16 carrot on M&A. If I think back to ATMI, it was kind of overcapitalized, then really ready for something more transformative. Just how do you kind of balance those two in the short term?
Well, it's a great question, you're right that it's going to be up to us, frankly, collectively, to decide what is the right amount of patience that we need to have. I'm glad that you collectively, as investors, did allow us to keep some level of cash on the balance sheet back in 2012 and 2013. Without that, we would not have been in a position to act on ATMI. Period. I think you would all agree with this, that we have generated a lot of value for customers, for investors as part of that transaction.
I think the reason we are insisting so much on M&A is that we want you to know that we are focused on it, we're going to have to find the right time, the right target, at the right value, we don't know, we won't know until such time as all of those stars will align. I hope that you would allow us to, again, exercise judgment. Now, again, what we wanted to say, that's really what was the message from Greg, is that we're also recognizing the fact that if we are empty-handed in a few years from now, we probably will have to do something with the cash that will be on the balance sheet, we have a commitment that we'll be responsible and thoughtful as we have always been in our capital allocation decisions.
We'll do that with a view to create value and return cash to shareholders one way or another.
Any last questions?
This is Susanto from Gabelli. You highlighted IoT and connected devices as one major growth driver, and I think there's a market view that a major portion of that is tied into mature node, mature process technology or older nodes. How should we view that area? Should we view it more as a unit driven or there is more to that where you can see some complexity and high-value solutions driving your new growth opportunities?
Yeah. I will answer it in two parts. Yes, you're right that IoT will require a lot of sensors. I think Tim made the point that a lot of the sensors will have to be a lot more reliable than they are today, a lot more energy efficient than they are today. I would expect actually to have some advances to the current process technologies used to make those sensors. Again, remember that when you hear Todd mention that 30% of our revenue come from new products, it means that 70% of our revenues come from older generation products. A lot of those products go into those trailing edge fabs. We are also very focused on a very large number of customers running older fabs.
The one thing I, again, want to stress, and I think you've heard that across the board from all presenters, is that those sensors are one, in my opinion, one real but yet small components of what will be required to enable this new digital order. I continue to believe that faster computing power, more energy-efficient servers will be required to crunch the massive amount of data that will be collected. A lot better storage solutions will be required. The amount of data that we're generating every minute is just growing exponentially, and I don't see any end to that. A massive amount of investment will be required to 5G network, and I'm sure the next generation after that will be right around the corner. Don't think of IoT as just benefiting older generation IC devices.
It will have a very holistic impact across many, many, many different types of semiconductor architectures.
I think that's it. Bertrand, do you have any final comments?
No. Again, thank you very much. It's been great to see you actually stick around for this very deep and busy day. We really appreciate your attention and hopefully see you very soon. Thank you.