Emmanuel’s 26 year career in the semiconductor space began with Texas Instruments, where he was a General Manager for a dedicated product line. As a VP for NXP he designed & executed the wireless connectivity product line, inevitably leading to his Executive Directorate position with Silicon Labs where he leverages his knowledge to developing their medical devices market offering.
Section 1: Microchip Design & Value Creation
1.1. What is your current design philosophy behind your microchips?
First is we’re always trying to predict where the market is going to go in the next few years, because of the design cycle of IC (integrated circuit) design. IC design typically has an ~18-month design cycle, from conceptualization to product launch. It then takes another 1-1.5 years for end-customers to move into volume production. Altogether the process is about two years if you are aggressive, but most likely it will take three years from a concept before you start shipping the device in volume. It is very important to try and understand where the market is going, because you need to have a three-year visibility to make sure you hit the market with the right device.
The second thing is that the semiconductor is a world of volume-based business. You need to make sure that you’re going to produce a minimal volume for the new chips. Margins are getting compressed, specifically because of new players. For instance, players out of China are dropping the prices pretty quickly on the market. Keeping the margin at a reasonable level is very important for companies because this is really what provides the ability to invest in the long term.
One of the big things that is happening in medical device markets, and COVID has accelerated this, is the need to have remote patient monitoring. From a market perspective, if you talk to medical companies right now, a lot of them are going to need to provide solutions for remote patient monitoring. In other markets that’s pretty much what people call IoT, where you want connected devices. On the medical side the end-product is going to connect to your phone most likely and this drives the need for security.
When you think about design philosophy, right now, what we’re trying to achieve here is design secure and connected devices that will provide solutions in medical, specifically for remote patient monitoring.
1.2. How do you design microchips to meet the demands for the automotive market?
All semiconductor companies have a new product introduction (NPI) flow, which is a very detailed development flow that goes from pre-concept to execution. There are different phases in the NPI. The first phase basically validates and solidifies the business case. Then the NPI completes a bottom-up analysis of resources to understand what it would take to design the IC, and then you move the IC into execution.
This process takes a minimum of six months to complete all together. At the end of the process, what you have is a crystal-clear business plan, a list of key customers that you want to target, and you have all the financial metrics for the new program (such as gross margin, volume, revenue, NPV, IR etc.). What you have as well is a bottom up analysis of the resources. So, all the teams have been basically involved and have signed up for the tasks to be executed, and you can move into execution. The execution itself takes about another year and that’s what you have at about 18 months from concept to launch. After launch, the customer is going to take their own time to design their own products.
1.3. What new technologies have you built into your microchips?
From a digital perspective, in terms of geometries, we are in 40 millimeters. At Silicon Labs, we use TSMC. What is interesting is if you look at the digital side, many companies are using foundries. The reason for that is because they cannot own their own foundry. It’s too expensive to build a factory, billions of dollars. They’re working with huge foundry guys, like TSMC, Global Foundries, Samsung, and others.
We use a 40-millimeter geometry. We use embedded flash as well. One of the big aspects with IoT is the importance of software, and software that was tiny in the past, has become a huge element of the entire solution. You’re going to talk about our protocol stacks like Wi-Fi, like Bluetooth. You’re going to talk about cloud connectivity to AWS and others, all this drives a significant increase of the software size.
The next addition will be machine learning and artificial intelligence. We’re very much in the early days, and I think most customers are still trying to figure out what they can do with AI/ML. It is going to take a little bit of time before they’re crystal clear on what they want to do with AI/ML, and how they can source their own AI/ML software.
1.3.1. How do these improvements serve the end user?
If you think about medical, one of the big aspects is patient monitoring and treatment moving from hospital to homes and this has been reinforced by the pandemic. One of the categories that we’re looking at specifically are 65+. There’s been more progress on technology adoption in the past 12 months due to COVID than we had seen in the previous 15 years.
One interesting metric is the rate of adoption, for people using telehealth. It was roughly 5% prior to the pandemic, and is now well over 50%.
Another illustration: if you previously had a heart condition you would call your doctor, set up an appointment, go to a hospital, they will do an ECG or an EKG, tell you what’s wrong. You would then go back home, and take your treatment. Well, here is the new way of doing this. You get a prescription for buying a patch from your pharmacy, the doctor puts it on your chest, and it monitors your heart activity continuously. You connect this patch to your phone. Your phone sends the data to your doctor. Next time you see your doctor, he has 1-3 months of monitoring, which is way better than anything else you had access to before. And that’s the change for the end user.
I think it’s going to be very revolutionary in terms of how we monitor patients and in terms of treatment, I think that the accuracy of the treatments is going to go up significantly due to these continuous monitoring and not this spot monitoring that we had before.
1.4. What is the future of microchip design in the next 5-10 years?
One of the things that is going to be very important for the next 5 to 10 years, specifically in medical, but not only in medical, is security. If you look at the RSA conference, the number of remote attacks on systems and specifically on healthcare, but not only healthcare, has gone up tremendously. Hackers are now using ransomware to extract money from medical and industrial companies. You might have heard that some hospitals were basically getting attacked by hackers while their teams were trying to treat COVID-19 patients.
One outcome is the need to embed on-chip security in the devices. If you take the example of Silicon Labs, we have recently announced a technology called Secure Vault, which is basically fully embedded into the ICs and provides a pretty long list of hardware and software features (both hardware and software) to protect the ICs against remote and physical attacks. What is going to become the paradigm of all microchip designs, is how you make sure that your ICs are protected.
It impacts not only the ICs, it impacts the way you write software, it impacts the way you produce your ICs. Right now, we are in a semiconductor crunch. People are talking about the lack of semiconductor chips in the world, what people don’t realize is that some of the production sites might be vulnerable to remote attacks from hackers who could potentially bring down some of these production sites.
The whole industry is going to have to think very deeply about how to embed more security into the production, into the chips themselves, into the software.
Section 2: Microchip Manufacturing & Production Cost Cutting
2.1. How as the Covid-19 pandemic impact your manufacturing process?
Fortunately, not a lot. We have implemented shifts within manufacturing to make sure that there was no contamination between the different teams. In manufacturing there is the assembly and test site, which is different, but on the wafer side, we use clean rooms, and these rooms are very well-protected against the outside world, because it is continuously filtered. People are also wearing masks 24/7 in those rooms. So COVID-19 impact on the manufacturing process has been minimal.
With COVID I think there has been an impact on the demand. What we’re seeing right now is the combination of two things. If you look at automotive, companies have reduced their inventory in early and mid-2019. They reduced their inventory to a level that was likely too low and now, they’re trying to catch up, while at the same time the market is going very strong.
Because people are staying home, because of some changes like remote monitoring in medical, people are ordering more semiconductor-based products. So, what we’re seeing for instance in Silicon Labs, a very big part of our business is connected to the smart home ecosystem. We have observed people buying many new smart end products for their home. I think it’s directly connected to the fact that they are now working from home. Things like intelligent thermostats, doorbells, video cameras. We are seeing a huge demand for such end-products, and this has impacted the demand significantly. But I can’t see where COVID-19 has impacted the manufacturing process.
2.2. What are the implications for creating an E2E process?
When a new development is started, we’re always trying to predict future demand. If you consider the NPI flow, when you start to enter the planning phase, where all the teams sit down together and review and agree on what needs to be done, in those teams, you have the design team, you have the product engineering team. So, the manufacturing is considered.
Companies like Apple use the latest and greatest state of the art manufacturing technologies. They’re aggressively moving to geometries that very few other companies can touch. You have seen these announcements about five millimeter and beyond. This is state of the art of what can be done in semiconductor now.
Some semi-conductor companies like Qualcomm are also at the tip of the spear in those technologies. These companies have to face the immense design complexity in those technologies, and the unknown of these technologies. They have to plan for it, and they work very closely with leading foundries to be successful.
When you think about the majority of the other players, and sometimes they can be huge players (like even TI or NXP, which are multi-billion companies), they’re not using such technologies. In other words, they don’t have the same constraints in terms of manufacturing. It does not mean that manufacturing is not important, but the technology is well-known.
It still does not mean that you do not involve manufacturing. In fact you might be looking at other parameters that are not the focus of the companies chasing the smallest geometries. One good example is the area of low power consumption. Low power has got a lot of specifics at the manufacturing process level. When you think about Apple, their end products are low power, but they are smartphones that are recharged every day. The process for manufacturing an IC that goes into a phone (that’s going to be recharged every day) is very different from the one needed to designing a chip that is going to go into a gas meter that will be living on the same battery for 20 years.
2.3. How do you scale your production process?
We’re working on the long-range plan. So basically, what we do is provide a long-range plan, which tries to give a five-year visibility in where we think we’re going to be going. We do it deeply. We do it by device. And this basically includes the technology.
We have an annual plan as well. And then of course we have a quarterly plan. From the long-range plan, you build up the need by technology. Once this is summed up, when everything is collected and put up, we basically communicate the resulting volumes to our foundry partners. And we have two things that we specifically look at, because the second one might get ignored. Many people when they think about semiconductor, think about wafer manufacturing, which is very true. That is one part of the problem. The other side is related to packaging, and is equally important.
So, we provide our long-range plan to our assembly and test sites so that they can basically plan the capacity from a packaging perspective.
2.4. How can producers cut their production costs?
2.4.1. Increase wafer size
First about 50% of the cost is located on the wafer side, wafer manufacturing. The big thing that has happened in the past to reduce production costs has been to decrease geometry and increase wafer size. We have been moving from basically 8-inch wafers to 12-inch wafers. There is a substantial cost reduction coming from the increase of the wafer size. This has been the biggest impact and is true for analog and digital.
2.4.2. Transistor geometry
The second aspect of it has been the move to finer geometries. This has been mostly on the digital side. When you move from a 28 or a 22 millimeter, down to a 16-millimeter FinFet, you are talking about reduction of the transistor size. It’s very costly, and it takes a long time to mature. But the reduction of the transistor directly translates into a smaller area size for the IC, which ultimately leads to a lower total cost.
So, wafer size increases one, transistor geometry two. And both items are expensive because increasing wafer size will imply a new manufacturing and basically a new factory. Transistor geometry often means a new factory as well. On the packaging side, I am not sure we have seen a huge revolution because packaging is basically very mechanical, and its cost is directly linked to the cost of the goods. At least this is true for the legacy packages used in the industry.
The forefront right now is the move to finer geometry to <5 millimeters. I think I’ve seen a few announcements where people have gone down to three-millimeter technologies. This is very much cutting edge. On the manufacturing side, there are some very specific technologies been using manufacturing, at least at the beginning, manufacturing in those technologies will not be cheap. It’s going to be expensive initially before it gets cheap. And that’s always been the case in the semiconductor market. That’s why you see less and less people willing to build up their own fabs, because they know that it takes a huge amount of time to get the money back from cutting-edge fabs. That’s why fewer and fewer people are willing to invest money into building their own fabs, except large players like TSMC, Samsung and Intel.