Cadence’s CEO looks at what comes after mobile and where the real innovation is happening.
Lip-Bu Tan, president and CEO of Cadence, opens up on the next big things, what will drive them, and what will change to make that happen. What follows are excerpts of that conversation.
SE: What are the biggest changes in the semiconductor industry over the past year?
Tan: The whole system approach to designing hardware and software is really happening now. It will continue to expand from here. Other than just looking at the silicon, you have to look at the overall system and the end user. That includes 5G and the whole automotive wave. The mobile market is becoming pretty saturated.
SE: How do you define a system?
Tan: It is the silicon, the software stack, the IP stack taken together, and even the vertical applications. What’s happening in the datacenter is a big shift, too. It used to be that you got the next server chip and you were done. Now it’s much more complex as the workloads are changing. There’s more data and more emphasis on I/O and memory. So now it’s a question of how you maximize your investment.
As a developer it used to be that you primarily had hardware and virtualization to worry about. Now the developer wants to have everything much faster and much cheaper. They’re demanding changes. Those affect a lot of the architecture requirements in terms of storage and the network speeds. Silicon photonics is happening, along with 100 gigabit per second SerDes. Then you have the network switch. It used to be that a switch that was 2 terabits per second was enough. Now the big web services want 4 terabits per second or much higher performance because of workload demands. They need to be much faster and lower latency. That opens up tremendous opportunities for the silicon guys. So rather than focusing on consolidation, there are huge opportunities opening up.
SE: But there aren’t many startups yet, right?
Tan: Sometimes the pendulum swings too much to one side. There is too much consolidation. The end customers feel nervous, and they also have new requirements that appeal to new startups. For example, there is a new company called Innovium in the high-speed switch market. They are developing products with three to four times higher performance than what existing companies are developing.
New startups in the market are more focused, and they see an opportunity. And customers are looking for opportunities. But there aren’t too many VCs willing to fund semiconductor companies, and, for EDA, there are almost none even though there is so much innovation needed. This is why we’ve pushed a culture of innovation at Cadence. There isn’t much out there to buy. You’ll see higher performance and some new IP blocks coming up from us. It’s exciting. There’s a new suite of companies coming up with new requirement levels that we’re focusing on. That includes automotive, cloud and IoT. In industrial IoT, there are some really good ones coming up, too.
SE: Is this spread out globally, or is the opportunity concentrated?
Tan: Right now we’re seeing opportunity in Israel, some in the United States and Europe, and more in China, Taiwan and Korea.
SE: Is China still focused on startups, or is it becoming more acquisition driven?
Tan: China is trying to be self sufficient and doing more buyouts. That’s starting to slow down a little though because the scrutiny is tighter from the Committee on Foreign Investment in the United States. That’s why we’re encouraging startups. There are a lot of good entrepreneurs coming up in areas like artificial intelligence, robots and drones, and there is a lot of good technology there. In drones, Chinese companies are not following other companies anymore. They’re the leaders. There’s a lot of computer vision intelligence in there. These are very advanced systems compared to the other drones.
SE: Where do you see the market for drones?
Tan: There are a lot of good industrial applications. They’re used for surveying. In mines, they can do mapping. With agriculture, you can send a drone out to see what’s going on. There are a lot of opportunities.
SE: What’s the biggest threat you see for this kind of innovation?
Tan: The talent pool. A lot of college kids feel the semiconductor business is a sunset industry. They’d rather go to Facebook or Uber or Airbnb. That’s why we have a very active college recruitment program. About 12% of our recruits every year come out of college. That’s important as you want a good mix of fresh, new thinking and industry experience.
Recent college grads are fearless and willing to try new things. We have been actively recruiting and mentoring millennials.
My phone can’t last beyond 11 o’clock in the morning with all the apps I have on it. That’s an opportunity for the semiconductor industry. If you’re looking at computer vision, the chips are great for scientific research, but automotive applications will drain your power very fast. The good news is this is all opportunity.
SE: What’s the bad news?
Tan: You may not have the talent or enough funding. The labs from companies like IBM and AT&T are cutting back. And you’d think that university funding comes from corporations, but it’s actually foreign companies. Almost no VCs want to invest.
I used to have at least 20 VCs willing to invest with me. Now I get one or two. And usually you want to bring corporate investors in at the later stage of the funding process. Now I have little choice but to include them on the first round. I’ve funded about 16 semiconductor companies. I can’t think of one involving any other venture capitalists. We need more. Hopefully we will have enough success that we can encourage others.
SE: So the business climate has changed enough that it’s hampering innovation. Is that going to change or is that the new order of things?
Tan: VCs are not too interested in hardware and are investing more in things such as Uber.
You need to really specialize. And then, because the industry’s growth is less than 10%, the revenue is not that high, and it takes too long to see a return. Big companies fund their own innovation, of course, and some, including Qualcomm and Intel, have venture arms as well, but it’s not across the board.
You still have to invest, of course, but you also have to strike a balance. From a Cadence point of view, we invest more than 35% of our revenue into R&D. We continue to perform well financially. Our operating margin last year was 27%. We need to continue generating profits and continue investing money into R&D.
SE: Is the technology an issue, as well? As we move to 7nm, it gets a lot more complex.
Tan: Doing this kind of development is very expensive, so that prohibits startups. Mobile is moving there. But it’s too expensive even for automotive. You need the volume to justify it. A lot of startups tend to be mixed signal because they don’t have to push to the most advanced nodes. Some of them are trying new materials, like gallium nitride. Those can be more cost effective without having to move to the advanced node.
SE: But not everything has to move to the most advanced node, particularly with the IoT, right?
Tan: That is a major shift. Data is going to the cloud. And the cloud itself is changing. In the past, every company had its own datacenter. That wasn’t scalable. That’s why we’re going to see more startups. They can just move the data to the cloud, and they’re up and running. That cost, depending on configuration, can be as little as two-thirds of running your own datacenter. That’s significant. There will be adoption that’s quite fast. And it’s more secure than your datacenter.
Cloud vendors see all the problem cases, and they fix them. Even the CIA is willing to use them because they are very safe. That’s a major shift. On top of that, these web services almost have software margins. They can be so profitable that they can massively scale. And they buy the silicon and define the chips themselves. They can drive massive economies of scale. The companies in this space have very special requirements in regard to speed, latency and storage. There is software-defined storage and software-defined networking, so they can fully utilize everything. For them it’s scale, cost and efficiency.
SE: There are a lot of business models in transition these days. How does that affect chip design?
Tan: You really have to design with the silicon in mind. All of these systems companies have special requirements, so there is room to customize the silicon. IP becomes very important, too. They only create a few pieces of IP themselves. The rest they want to buy from the industry, and that has to be silicon proven and high quality.
SE: One big shift that we’ve seen is a recognition that trying to create mixed-signal engineers isn’t working. What happened?
Tan: Analog is a very different experience. Digital is ones and zeroes. We can optimize the experience so analog and digital can work together, and some of the digital flow can come into the analog side. Instead of two separate processes that are not optimized, we facilitate collaboration.
SE: Without trying to make a single systems engineer who understands everything, right?
Tan: Yes, that’s very difficult.
SE: It’s similar with software engineers, isn’t it?
Tan: We’re looking at what the software stacks are for verticals, such as automotive and aviation, that can be more optimized for our hardware. We want to make it easier for application developers.
SE: Where does Cadence play from a security standpoint? What do you bring to the table that’s different?
Tan: Security and safety are getting more important. It’s not just your mobile phone anymore. Now your car can get hacked. A lot of security tends to be on the software side. We’re looking at that with IP. For hardware, security is embedded and it’s expected. We did look at software, but it’s a little bit tricky on the software side because you’re going against some of the big players.
More security issues come up when we put our tools in the cloud. We hired a new CIO, who came from Symantec. We also have people on his team focused on security. So this small team isn’t just IT to support our operations. We’re also asking them to look at cloud security offerings and evaluate the pieces of security that need to be embedded. Collaboration becomes very important, and we need a trusted environment to do that. It requires a lot of trust between the foundry, IP vendor and the customer. With some of them, we have three- or four-way agreements.