Costello's DAC Keynote

On the Wednesday of DAC, Joe Costello was back giving a keynote. The last time was in 2017 when he was CEO of Enlighted (since acquired by Siemens). Joe said it was the first time he'd given any sort of speech since the start of Covid. But before he started, he dedicated his keynote to Jim Hogan (with tears in his eyes). For more about Jim (including an appearance by Joe in the video) see my post Jim Hogan's Celebration of Life. Or, as Joe put it:

Jim was transformational in this industry and it wouldn’t be what it is today without Jim. Even if you don’t know him, I guarantee you are affected by him.

Joe's presentation was titled When Winds of Change Blow, Some People Build Walls and Others Build Windmills, which is apparently a Chinese proverb.

Change

Joe started off talking about change in general, and would eventually move on to change in the EDA industry. "How do people deal with change?" he asked.

Bottom line, humans don’t like change. Most people build walls because they don’t like change. With the current stock price, the CEO of Cadence, obviously doesn't want change. But people with a crappy life don’t like change either, and they struggle to change.

The most stressful events in people’s lives such as changing jobs, changing relationships, a death in the family. Everything that is the most stressful has to do with change. People don’t deal well with change.

We all live our lives having built a model of the world in our heads. I simulate in my head. That model is called the “ego” (roughly). Change means having to change your model about how the world is and how it works. That’s why people resist change crazily. People will stick with the model through thick and thin over and over again. But to execute change you have to change people’s models. Otherwise, people won’t change their behavior. They’ll just build a bigger wall.

As people get older, in general, they get more conservative. They resist change. They like it the way it is and don’t want to change that much. Truth is that the older you get, the less you have to lose. But people believe they have something to protect: money, family, reputation, career, whatever.

Joe pointed out that this was the tip of the iceberg. Everything he has said about people resisting change is also true of groups, of  companies, of countries, and it is easier to change one person than a group. Groups have their own ego and their own models which are really hard to change. So to effect change you have to figure out how to change people's mental models.

The big changes are macro trends happening at large scale, industry-wide or maybe country-wide, perhaps worldwide. The biggest gains that anybody makes all have to do with macro trends. You can't fight against macro trends but you can be successful by going along with them.

Joe told a story about trying to fight a riptide when surfing as a kid in Santa Barbara. If you've never been caught in one (I have) then it is hard to comprehend just how powerful a riptide is and how scary it is that as hard as you try to swim towards the shore you are going further out to sea. The solution is to swim sideways since if the riptide is going out to sea at some point, the water has to come from somewhere, and so there are other places where the current is toward the shore.

The Birth of EDA

Joe went history and talked about EDA in the 1980s:

I started in EDA in 1984. Mainframes. Manual layout. Very limited software. DRC, some simulation. Internal CAD teams developed all the software. What broke the market open? First, workstations showed up. Unix, C. Every designer could have a computer on their desk. The ASIC design methodology. Suddenly designers could build chips on their own. It was the first golden age of IC design and EDA. All the tools were in the hands of designers, and so there were new freedom and possibilities in design. The energy was palpable back in late 1980s and early 90s. That energy and excitement drives things. Small companies and system companies, not just semiconductor companies, could afford to do design. There were tons of ASIC startups and huge growth in ICs and EDA, I became CEO of Cadence and rode that wave. “Spot the trend and ride the trend”

Joe said that one of the most difficult conversations he ever had in business was at Cadence when some of the team supported going the software-only route versus a turnkey system model (hardware + software) which was the way it was done. Joe said it was obvious to him since there is no value-add to a software company developing hardware. But people fought over it since that's how it had always been done. I remember similar discussions at VLSI Technology about whether we could sell software for a higher price than the hardware that it ran on, since that had never been done before. Now that hardware is cheap and software is still expensive, it is almost comical that these were serious discussions.

The ASIC design wave was driving the whole thing. It wasn't that I had some magic genius moment, it was a group of people recognizing the trends and riding them. If you weren’t riding these trends you weren’t going to win.

Joe left Cadence and EDA and he thinks that there were dark days in the early 2000s. Internal CAD sort of came back with really complex flows...except that they weren't building it all themselves. Server farms had tens of thousands of machines. There was not much room for small companies since the big three had full flows and it was really hard to get your startup tool inserted in there.

And the economics was horrible because these guys had all-you-can-eat deals. So there was very little investment available for chip or EDA companies. The best you could hope for is a sale to one of the big 3.

Joe didn't point it out, but the sale prices of EDA companies to the big three were very low compared to the 1990s. I joined Cadence since I worked at Ambit, and we were acquired for over $300M in 1999, one of the last 9-digit acquisitions.

Today

Joe moved on to the 2020s and what the macro trends are today. There is a new hardware platform in the form of the cloud.

Someone said to me that we don’t really do cloud in EDA. I dug in and it was true. The rest of the world has moved on and we are still building walls in EDA. Cloud is inevitable, it’s happening. It’s economics. Here it’s the economics of scale. Datacenters are not the business of semiconductor companies. Semiconductor is a tough enough business and you have to focus on it.

Joe moved onto a message I've heard him preach before on Cooley's DAC Troublemakers' Panel. See my post Making Trouble in Las Vegas.

There's a new business model Saas [software as a service[. It's just as obvious as software-only back in the early days. And when I say SaaS, I mean pay by the minute. Licenses made a lot of sense in the era of the workstation, but a license model makes no sense in the cloud when I'm already buying my compute by the minute.

Joe calls the license model in the cloud era the "stopped watch business model". Most of the time you have too many licenses, and then when you really need them you don't have enough. You just have the right number of licenses twice, once on the way up and once on the way down.

Processors

Joe moved onto the new design waver that is driving growth: the custom processor market. It's not just a great marketing idea, general-purpose architectures are not going to work anymore. For decades everyone could ride Moore's Law and just buy the next processor from Intel. But the future is going to be specialized processors and open architectures like RISC-V. There are economies of scale with scores of companies focusing their energies.

In EDA we need to look at what others have done, such as the Linux Foundation, Cloud Native Compute Foundation, and so on. And I don't just mean something like the Connections Program [a Cadence program to make tools available to other EDA companies] Open program are happening all around. Kubernetes is the fastest-growing platform in history. Semiconductor has a good start with RISC-V but we have to do it in EDA. There are all sorts of open-source models. EDA needs to make it simple and easy to put together full flows. Don't put artificial barriers to build things. And it needs to be in the cloud too. There will be a renaissance in EDA. There already has been in the stock prices but I'm talking about EDA becoming a place where everyone is going to want to come, where the problems we are trying to solve are cool. More innovation. More young companies. More designers.

It's like ASIC in the old days. Except one engineer sitting in his living room in Mumbai with a cloud platform has more compute that the best engineer at Intel or NVIDIA. That’s revolutionary and changes the world. Those guys are going to be building ton of design-specific architectures.

Geopolitics

Joe pointed out that there is one big cloud: the new geopolitics and the "us against the world" attitude. He pointed out that he can go negative on any politician, and has nothing good to say about any of them.

I wouldn't have any of these people run any company that I'm involved with.

How did we get here?

I thought the world was getting so intertwined it would be tough to have wars. "My kids are going to be safe", I thought. I never would have predicted that in a few years, enough idiots running enough countries could screw things up as fast as they did.

We pushed China into a corner and they reacted. We pushed the dragon into a corner. I’d do the same thing if I was running China. The US did what none of the great minds or leaders could do and unite the Chinese semiconductor and EDA industry. Now incredibly unified in China right now. They are now investing more in semiconductors than the rest of the world put together by a factor of five. There are 5-10,000 ASIC startups. It reminds me of the golden age here back in the 1990s. I think China will become the biggest market for EDA in about 5 years. EDA valuations in China are in orbit. Two EDA startups in China that have existed for just 1-2 years, but with valuations over $1B. Over $300M invested. One going public with $20M in revenue $1.8B valuation.

“Trade wards are good and easy to win”, Trump

“Trade wars are hard to win and impossible if you don’t have a strategy” Joe Costello, December 8th 2021.

What do we do now?

We can’t give up on the fastest-growing marketplace. It won’t change quickly. We’ve go to be involved with politics since they are involved with us. These macro trends don’t ask permission. We have to play the long game. Rebuilding EDA won’t happen overnight.

Here's a shocking and scary thing. I was talking to the best export lawyer in US, she’s brilliant. She said if she was building a technology company today she would be loth to have a single employee in the USA.

Joe recommended what he considers the best book on how to deal with the winds of change. I recommend it too. See my post Clayton Christensen and the Innovator's Dilemma. Not only is the book brilliant for its ideas, it is also the most readable business book you will ever come across.

Summary

Joe wrapped up with three bullets.

• The winds of change are blowing. Don’t resist the change, don’t swim against the riptide.
• Harness the incredible power of the macro trends. I didn’t do much in EDA, I spotted the trends way bigger than Cadence. Harness that.
• And you will find yourself in the center of the renaissance of chip design and EDA.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.

.