Interview with Ritesh Saraf, Co-Founder & CEO of OmniPhy, about Building Networking & Communication Intellectual Property for IoT
In this interview with Ritesh Saraf, co-founder and the CEO of OmniPhy, we discuss the emerging opportunities of the Internet of Things (IoT) and what OmniPhy is doing to build essential building blocks to enable IoT in consumer, automotive and industrial markets. OmniPhy is a semiconductor Intellectual Property, or IP, supplier, that provides these communications and networking building block to their customers who build chips for this rapidly growing market. OmniPhy delivers specialized interface and communication technology to their customers. Their technology provides greater design margins and fast time-to-market for established and emerging networking and communication standards. Ritesh is a seasoned semiconductor executive with 17 years of leadership experience – leading teams to deliver over 10 high-speed Serializer/Deserializer (SerDes) designs into silicon across multiple foundries, or manufacturing partners, and multiple customers. A critical member of the founding team at Prism Circuits – and the Vice President of Engineering – he oversaw its phenomenal growth and helped scale the company from a small privately held startup to an IP powerhouse with over fifty employees and millions of designs shipping in volume. Prior to joining OmniPhy as it’s Chief Executive Officer, he was the Vice President of SerDes Engineering at MoSys, through its acquisition of Prism Circuits, where he oversaw the organization implementing the entire line-up of SerDes designs. Ritesh holds a Bachelor in Electronics and Communications from the Delhi Institute of Technology and has co-authored several technical publications and US patents. Patrick: Hi, this is Patrick Henry, the CEO of QuestFusion, with the Real Deal…What Matters. We’re here today with Ritesh Saraf. Ritesh is a co-founder and the CEO of OmniPhy, which is a semiconductor intellectual property, or IP, company supplier. OmniPhy delivers specialized interface and communications technology to their customers. Their technology provides greater design margin and faster time-to-market for established and emerging standards including high-speed serializers/deserializers- these are, in our industry, called SerDes- and high performance Ethernet solutions. Ethernet’s the technology that connects all computers together and is now starting to be used in automobiles and other technologies, which we’ll talk a little bit about. OmniPhy delivers this IP in an advanced process technology, which is important to a larger number of companies. In the semiconductor business, you have different processes as you go into higher performance and higher geometries, more circuits on a printed circuit board. OmniPhy is able to work in these very deep sub-micron technologies, which is very important to a number of customers. The company services the industry industrial, automotive, consumer, and networking segments through its broad portfolio of silicon-proven designs. Here’s a little bit about Ritesh. He’s a seasoned semiconductor executive, with 17 years of leadership experience, leading teams to deliver over 10 high speed SerDes designs into silicon, across multiple foundries- these are factories or manufacturing factoring partners that are used in the semiconductor business- and multiple customers. A critical member of the founding team of Prism Circuits and the Vice-President of Engineering, he oversaw its phenomenal growth and helped scale it with over 50 employees and millions of design shipping and volume. Prior to joining OmniPhy as its Chief Executive Officer, he was the Vice-President of SerDes engineering at MoSys, through its acquisition of Prism Circuits, where he oversaw the organization and implementation of the entire lineup of SerDes designs. Ritesh holds a Bachelor in Electronics and Communications from the Delhi Institute of Technology and has co-authored several technical publications and U.S. patents. Welcome, Ritesh. Ritesh: Thank you, Patrick, for having me. Patrick: Ritesh and I got to know each other because, when I was running Entropic, OmniPhy was a key supplier for us. We were actually OmniPhy’s first customer when they left MoSys to start this company. Ritesh: The very first one. Patrick: How did you make the decision to start OmniPhy? Ritesh: It was an interesting story. When we got acquired at MoSys back in 2009, we were prosecuting the IP business. I was running that business as part of MoSys. End of 2011, MoSys decided to exit that business because they wanted to focus on IC business. Patrick: Doing products versus intellectual property for other people’s products? Ritesh: Exactly. MoSys was doing IC as well as IP at that time. I was managing the IP business. When the company decided to exit that business, my cofounders and I felt that there was still a lot of value in being in that business. We had been in that business at that point already for six years. We had been working with Entropic, also, at that time. It seemed like a logical step for us to keep doing that business, because we had known that business and had been successful in it. As MoSys shut down its IP business, we came out of MoSys and started OmniPhy, to run an IP business company. Patrick: I recall this was driven mainly by our Vice-President of Mixed Signal Technology and our head of overall engineering, in terms of the relationship with OmniPhy. I like this about startups. Sometimes there are challenges working with startups because there’s a little bit more risk. However, you guys were really flexible. You had a seasoned team. That was something that was really important. We wanted to design on a process technology. Some of the big guys, like a Synopsis or a Cadence, were unwilling to do those projects for us. It was really nice being able to work with you guys. What was the original business idea and the business model? Has it changed over time? Was there a pivot that happened within your company? Ritesh: Good question. Yes, originally, when we started OmniPhy, our idea was to do an IP business in a similar domain that we had done in the past, which is SerDes IPs. When I say SerDes IPs, these would be protocols like USB, which are consumer protocols, and protocols for enterprise, which now go into high-end computing or high-end networking in the cloud infrastructure. These would be protocols like 10 GKR and more recently 25 GKR towards 100 gig Ethernet. Our thought was that we wanted to be at the leading edge in terms of bandwidth, in terms of data rate, because that was something we had been doing in the past. We wanted to continue that. As we got started in this business, we had to react to the market conditions. We had to react to how the competitiveness of our products was with relation to what was already there in the market. We had to react to what the customers were demanding and where there was a scarcity of solutions. The first set of products, which we worked together with Entropic, was focused on the consumer market. Entropic was developing a set top box solution for the consumer market, so our IPs were tailored towards the consumer market. As we tried to sell those, we experienced a lot of competition. We were trying to compete with the biggest players who had been doing the same things. Three or four years back, mobile and consumer were considered very hot. There was a lot of competition. For a smaller company, it took a lot of investment to create the differentiation that we wanted. Through this process, we had developed certain IPs, like Ethernet again, which we started working with Entropic. We went to the market and tried to sell those out. Surprisingly, we saw a whole lot of traction for those technologies, because those technologies were not available as IP. Ethernet is an interesting story. Ethernet as a protocol is probably 25 years old. Patrick: At least, yes. Ritesh: At least. The first Ethernet files were defined in 1990, with 10M meg transfer rates. Maybe in the year 2000, the gigabits came about and so forth. Surprisingly, all the installations of Ethernet, until a few years ago, had been as ICs, not as IPs. Companies like Broadcom, companies like Marvell, had been selling a lot of Ethernet chips into solutions started as enterprise but later on even as consumer. Today, if you look at Smart TVs, you’ll have an Ethernet port. Your Gateway router at home has an Ethernet port (if you have yet set up your D-Link router, now could be the time!). Your set top box has an Ethernet port. A few years back, all those functionalities were supplied by IC companies. As process nodes started becoming newer and the integration capacity of those processes became higher, our customers started thinking, “Why can’t I integrate this function?” It’s the SOC mindset. The whole system becomes a chip. For the longest time, Ethernet had been outside of this SOC. Patrick: So much cost was driven out of the Ethernet PHY, that it was relatively cheap to buy a discrete solution. There wasn’t a 100% attach rate for that technology on all of these different Internet of Things type of products. Ritesh. Sure. However, a lot of the consumer companies still had to have it, even though it may not always be used. They had to put it in. As this integration increased, the cost of silicon also decreased. At some point, there was a crossover. Patrick: There always is. Ritesh: Even if an Ethernet port was selling for $0.50 or $0.70, if some company was going to sell 5 million of them, 4 million of them, or even 10 million of them, it didn’t justify that level of investment if they could get it for 1/6th or 1/10th of the cost. Patrick: Exactly. Isn’t that crazy? When you get into the millions and tens of millions and hundreds of millions, pennies and half a pennies start to matter on these kinds of solutions. Ritesh: The interesting thing is that initially we saw a lot of challenge in the consumer market with our other IPs. Then the popularity of our Ethernet was the in consumer market initially. It is so cost conscious. If someone’s saving $0.20, $0.30, $0.40 by doing it, they’re willing to invest certain amounts of money to get that saving per piece. Patrick: The Ethernet market, as I understand it, is evolving. There are various different flavors. You have the traditional, what we used to call, 10Base-T and 100Base-T. Now a gigabit Ethernet is what goes into computers and now into TVs and things like that. You were mentioning, off camera, about industrial Ethernet and especially automotive. There are these new flavors of Ethernet. Can you talk a little bit about that? Ritesh: Sure. Ethernet is having a rebirth of sorts. Patrick: Kind of a renaissance. The renaissance of Ethernet. Ritesh: Its variations and its flavors are being adopted in different applications and different spaces. Some people may club it as IOT. Ethernet is a way of connectivity. Wireless is not the solution everywhere. The default solution for connectivity becomes Ethernet. The reason for that is because it’s been there for so long, it’s well understood, there’s a whole ecosystem that supplies the Ethernet installation. The cost of installation and maintenance of that installation is also low. That’s why it becomes so attractive to become the connectivity solution. More specifically to your question, automotive and industrial, let me talk about industrial first. Patrick: Those are typically areas in the industry where wireless doesn’t work very well because you have lot of noisiness, either on a factory floor or in an automobile. Having a wire-based solution like Ethernet is very important. Ritesh: Exactly. You hit the nail on the head. These installations in the factory, production lines, they don’t want to have wireless. They want to have predictable, guaranteed connectivity. Ethernet became the solution there. There’s a plethora of high-level protocols in the industrial Ethernet space. By higher level, I mean at the controller and application layer. They’ve all agreed to use the Ethernet 100Base-T and 10Base-T, 1000Base-T and 100Base-T as the physical layer connectivity. Patrick: The analogy I use is the Ethernet or these communications protocols are like the freeway system. You’re going to drive different kinds of cars on them. All the upper-level protocols are how you deliver information, the kinds of information you deliver, but the pavement and the roads and the plumbing are a standards-based solution, which is Ethernet. Ritesh: Exactly. You said it right. Even in our homes, we may be watching video, we may be playing a game, we may be reading a book online- we’re doing different things, but all that data, all that information is going over the same plumbing. Similarly in industrial, it may be doing different things but it’s going over the same plumbing. In industrial, there are some new features that are required to enhance the Ethernet for the application use. Patrick: Versus traditional computers. Ritesh: Versus tradition computers. Here are a few of them: time synchronization. You think of a robot on a production line. The robot arm has to move very precisely to do an action, maybe to pick a part. There needs to be very precise time synchronization of when that arm needs to go down and pick up a part, in the order of milliseconds, nanoseconds. Now there are some additions to the Ethernet protocol that enable that. That’s an enhancement to that for industrial. The other is latency, which is tied to the same thing. It has to have a deterministic latency, from one end to another. Patrick: Standard Ethernet has historically been pretty bad with predictable latency. Tightening up latency is a big challenge. Ritesh: Those are challenges that need to be solved. Even though you may think that the base technology is the same, there are unique challenges in the industrial space that are required to be solved to make it applicable. We’ve gone through that process. We’ve established how to do these things. We’re seeing good traction in the market. Again, a lot of the competition we really see today is not IP providers. The competition we really see is IC providers. Today, there are companies in Japan like Renaissance, which has been an industrial supplier. There are companies like Texas Instruments that have been a supplier to the industrial world. We have a big advantage compared to them as a small, nimble company, we move fast. We adapt to the requirements. Patrick: You’re doing things that would violate their own DNA as a big company. They’re not going to want to license their technology to their competitors. You’re approaching the market in a completely different way by being someone who can provide this somewhat standard, but a lot of differentiation based on all these new things based on all these new things to a variety of different customers. These other big companies are not going to supply this as IP. They’re going to want to sell chips. Ritesh: Exactly. That is happening. One of the other trends from the industrial space is they had been using commercial, off-the-shelf silicon to build all their systems, because of so much focus on IOT for industrial, especially in Europe where the German government has a put a fund of close to half a billion euros towards industrial automation. Many enjoyed process engineering with the support of industrial automation. They have christened it as industry 4.0, as the fourth industrial revolution. Everything in the industry is going to be cyber-physical. What that means is there are no people who are doing jobs on the shop floor. Everything is being controlled remotely. They want to retool the production line in a matter of minutes, instead of days. A lot of efficiency is being driven. The backbone for all of that network capability is Ethernet, with such network devices used to connect businesses to their customers, some may find that network device monitoring could be a good way enjoy that network performance is constantly working towards the goal of the business. A lot of customers are looking into unique ways to solve those problems. There isn’t commercial silicon available to solve those problems. People are looking at doing their own customized SOC solutions. That’s a really good fit for us. When some industrial company wants to do a custom solution for themselves, we are right there to enable them with the Ethernet piece of it. We’re creating a huge differentiation for them. Patrick: This big interesting Ethernet opportunity, the rebirth of Ethernet, came out of one of your customers forcing you to do something you really didn’t want to do, but you were going lose all the other business if you didn’t end up doing that. Ritesh: It’s funny. I guess the learning out of this has been: never avoid looking at a market opportunity and keep your eyes and ears open to what your customers want from you. Listening to the customer has been a key learning. That education really tells you what will make you successful. We talked about industrial; automotive is the other space where Ethernet is making big waves. Traditionally, automotive has had its diverse set of protocols that are used to communicate between the different engineering control units, or ECUs. Patrick: There are multiple subsystems within an automobile. Ritesh: Yes, and they have to talk to each other. There’s a subsystem for the engine control. There’s a subsystem for the power train. There’s a subsystem for communication or the debug controller, infotainment for the driver. Traditionally, they have had a lot of different protocols like CAN or LIN. CAN is for control area network. Patrick: These were proprietary protocols that were eventually, through alliances or some standards organizations, made into standards or they’ve always been proprietary? Ritesh: Some of them are proprietary, like MOST, which is used mostly for multimedia in the car. The others were more standards. They’ve grown in a unique way and they’ve become standards. All these connectivity options, they were a very low bandwidth. Most of the connections in the cars, a few years back, had a few hundred kilobits of bandwidth. Patrick: Very slow. Ritesh: Very slow bandwidth. Infotainment applications were the first ones to drive bandwidth. Really what has tipped the scales is this automated driving. The goal of the car being fully automated means it has a lot of cameras, it has a lot of sensors, and that means a lot of data has to go back and forth to enable the car to function. That has necessitated an increase in the bandwidth requirement, which these protocols could not support the way they were defined. That’s where Ethernet came in as a solution for the car network. Two goals: one was increase the bandwidth and provide a path for incremental updates. Ethernet as a standard, like we discussed, has been there for 25 years. They have consistently been able to up the bandwidth. Patrick: They probably, automobile electronics manufacturers more specifically, want to preserve a lot of the good features that they’ve developed in these historical protocols that may be set above the PHY layer. Is that part of what’s making automotive Ethernet different from other flavors, incorporation of some of these other bells and whistles or fixtures and faucets that exist currently in cars? Talk to me little bit about that. Ritesh: Automotive has its own specific needs, like you mentioned. One of the things is EMC, emissions. That needs to be incorporated. That by itself creates a new design requirement. That’s one thing. The second thing is weight. Cabling weight is a big thing in automotive. For every couple of pounds you add on the car, your fuel efficiency is going to down 10% to 15%. That became a real constraint on how you decide the cabling. The other thing is sleep modes. You don’t consume power. You consume zero power when it’s not being active. On the other hand, when it needs to be active, it wakes up as quickly as possible. All of these constraints in some ways are unique to the car for automotive electronics. All of these are being incorporated into this automotive Ethernet solution. EMC is one. Cabling weight. For example, there’s a standard Ethernet Cat 5 cable, which everyone’s familiar with. The automotive industry doesn’t use the Cat 5 cable. The automotive Ethernet uses a single pair unshielded. It’s very thin, very lightweight. That was a unique solution. Patrick: Totally different requirement. Ritesh: Totally different requirement. Two things were accomplished by that. One, the weight was reduced. Second, the cost was also reduced. The whole solution is targeted towards lower cost as well. These are some of the unique things that went into this process of defining automotive Ethernet. There are a couple of alliances which are working on it including IEEE, with active participation from all the car makers, whether it’s European car makers like BMW, Audi, or the U.S. or the Japanese car makers. Everyone sees that as the future. The autonomous car is the future. This is a step towards the autonomous car. The first step is to define the network of the car. Patrick: How does it communicate? Ritesh: How does it communicate? This is defining the network. Patrick: How big is the market for your types of products? It seems like with the rebirth of Ethernet it could be pretty massive. Ritesh: Pretty massive, especially automotive. As an example, if you talk about automotive Ethernet, there are projections that in the next four to five years, let’s say 2020, a standard car would have 18 to 30 ports of Ethernet. There would be about 100 million cars manufactured per year. It quickly goes into billions of ports. Again, how fast does the market mature and how fast does it reach it? Patrick: That’s only one segment of what you’re going after. Ritesh: I’d definitely, say from a volume perspective, that’s a huge market. Patrick: Yes, that’s massive. Ritesh: The other market, industrial Ethernet, again, that’s less in terms of volume, but still could be more in terms of number of customers. The third segment in the Ethernet space is the consumer Ethernet. Patrick: You’re seeing opportunities there for you because people are integrating the Ethernet into their system on chip products, where historically the PHY was external. Ritesh: Exactly. We have all the biggest TV manufacturers using our solution. Easy to say, we probably supply over 80% of the TV manufacturers. You talk about the biggest ones in Korea, both of LGs and Samsung; in Japan, you have Panasonic; then some of the companies in China. That’s a big market we’ve supplied our solutions to. Patrick: Terrific. How do you guys make money? Do you charge royalties? Is it an upfront licensing fee? Is it a combination of both? Does it depend on the customers? How does it work? Ritesh: It depends on a number of things. In a lot of cases, it’s a combination of different things. Certain situations or certain scenarios, more specifically in the consumer world, royalty becomes an overbearing thing for the customer. In that space, it’s purely license-based. In the industrial and automotive domains, all of our agreements are royalty-based on top of the license fee. Patrick: Okay. That lowers the license fee? Is there typically a buyout provision where someone can buy out the royalty or does it pretty much continues to come down when the royalty comes down? Ritesh: Most of them, the royalty continues and it scales down with volume. There are certain scenarios, where we’ve even done a buyout. It varies customer to customer and what kind of value we see from them. Patrick: You’ve pretty much explained how your product works. I know that you had really good revenue for quite some time and you’ve been break even since your first year of operation. Very impressive. You guys have bootstrapped this from the beginning. You’ve never taken venture money. I’m so proud of you guys. Ritesh: Thank you. Semiconductors, today, is a very challenging market to raise venture money in. Patrick: Yes. It’s tough to get venture money into semis, for sure. Ritesh: We did put in our own contributions to get the company off the ground. After the first year, we’ve been cash flow positive. Patrick: Great. What’s next for you guys? Are you looking at raising outside money at some point? This sounds like it’s a pretty big opportunity. Is that something you guys are thinking about or are you pretty much going to continue to organically grow it and take advantage of opportunities that way? Ritesh: Our direction right now is to keep doing more of what we’re doing. It’s been a successful method for us. Going forward, we would get into more of these strategic areas, like automotive, keep extending ourselves as the experts, going to next generations there or in industrial. In certain cases, we’re doing some strategic agreements with the customer or customers, where they might do some level of funding to develop certain things. We work closely together, enable them with technology, and they benefit by early access to the market. Overall, we want to focus on technology. We want to be the experts in a particular technology. Going and becoming an IC company, that’s something people used to think was the best thing to do, maybe a decade or 15 years ago. It’s gotten more challenging now. The return on investment that you make is a long time out and it’s also risky. In the model that we have today, we have a quicker return to the investments we make, although the multiplier may not be as high until the royalties kick in. It seems to be a successful model that has worked for us so far. We want to take that model itself to the next level. Again, if there are opportunities for us to become part of a bigger company and leverage our technical expertise into something much bigger, that’s also an area that we’re open to. Patrick: What does the competitive landscape look like for your products right now? Is it primarily other IC companies or are there other IP companies now getting into this space? Ritesh: One of the reasons for the Ethernet product line, specifically the industrial and automotive, and us having such a huge success was that there’s no competition for us in the IP landscape. Our competitors are IC companies. We have a different kind of discussion when we’re competing in those opportunities. Whereas in the IP space, if there are several providers, it becomes a race to the bottom, where a customer would pitch four or five providers against each other and try to get the best cost out of them. That’s something we’ve tried to consciously not get into, a race to the bottom. Have ourselves differentiated, bring some unique solution to the market where we’re not fighting for price, but we’re actually reducing the cost of the customer by giving them solutions which they wouldn’t have had. Patrick: Even though there’s a basis of a standards-based technology, which inherently means things would be commoditized, there’s enough differentiation and that differentiation is evolving over time and some of these specialized applications of the base technology. Is that the right way to say that? Ritesh: Yes, exactly. The key is to keep up the speed or be ahead of the curve. You keep being unique, whether it’s a new standard or whether it’s enhancements or features that create that unique application. Patrick: Is there anything else you’d like to share with our audience of entrepreneurs out there? You obviously have your battle scars from running a couple of companies. You’ve been around the entrepreneurial landscape for some period of time. What are the key words of wisdom you have for our audience listening? Ritesh: There are a couple of things from my personal learning. First is, when you want to become an entrepreneur or people who are already entrepreneurs, you will have ups and downs. It’s like a rollercoaster. The best thing is to ride it through. Don’t try to jump off. You ride it through, something good will come at the end. The other learning I’ve had, which we talked about is, listen to the customers. From knowing what is required in the market, what will succeed, the key thing is to listen to the customer. A lot of times, people who have engineering backgrounds, starting companies, think they know exactly what’s going to work. Then you go and build it and then you go and talk to the customer who says, “I could use something else.” That’s a key learning. For a business to be successful, you need to have your eyes and ears very open. Patrick: It’s an interactive dialogue. Sometimes they know what they want, generally. “I want a cake,” but then you give them some kind a cake and they say, “I want a different kind of frosting.” You have to get that feedback and have an ongoing dialogue with the customers. Their market landscape is changing. It’s a dynamic environment. The quicker you can iterate things. It’s a lean startup concept. I think you guys have done a great job implementing those things, where you test things, you get customer feedback, and you tweak it. You’re able to move quickly, which is impressive. Ritesh: Also, I think what the learning has been is there is path of least resistance for you to move forward. That path of least resistance, we figured out by talking to people, by listening to the feedback. You might think, “I’m so good at this. I’ll do the best design among everyone.” However, when you do that, spend time and energy doing that, go the market, and that advantage you built, whether it’s 10%, 20%, 30%, may not be what the customer cares about. Even though you build an advantage, it may not be so critical for the customer. They may be valuing their advantage on certain other parameters. Without that dialogue, you would be on the wrong side. Patrick: This is awesome. This has been really fantastic. I know you’re a super busy guy. I know you have this huge operation over in India, as well. You’re always off visiting customers. I’m glad you could take time to talk to us today. Ritesh: Glad talking to you. Patrick: Good to see you. This is Patrick Henry, the CEO of QuestFusion, with the Real Deal…What Matters. This is Patrick Henry, CEO of QuestFusion, with The Real Deal…What Matters.