The Challenge Of Fitting In

Connections between players in the semiconductor industry are becoming critical for survival. Whether the focus is a connected car, home automation, health care or the energy grid, each company in each of those markets relies on others to build useful products.

There are several forces at work here. One is an emphasis on connecting everything, regardless of whether it is inside a single vertical market or between markets. A second, is the need to bring products to market quickly, inexpensively, and customized for specific use cases. And a third is the ability to reach into new markets where companies have limited expertise. All of these are well beyond the capabilities of even the largest companies, which makes ecosystems essential.

“Connections between us all are what amplify the activities of any one organization,” said Simon Segars, CEO of ARM, during a keynote address at the recent TechCon conference in Silicon Valley. “Our future really is dependent on these connections between us. Not just connections between the objects that we seek to connect all the time, but the connections between the businesses, and it is those connections that make everything really happen.”

For many decades in the past, he noted, the name of the game was optimizing supply chains. For example, the Ford Model T cost $800 when it was first launched on the market, which was too expensive for most people, but through optimizing the way cars were built by optimizing the supply chain, the cost came down over time to the point where at about $250, many more people, including the people that worked in factories could afford to buy these products. By driving efficiency, costs came down. This dynamic can be seen in everything from vacuum cleaners to DVD players.

Segars calculated the cost to build a smartphone 25 years ago would have been approximately $3.5 million — which today can be had essentially for free. “Fortunately though, we’ve collectively driven price out through scale, through active management of supply chains, and through innovation. Innovation has driven cost out and enabled huge volumes of devices to be shipped, which we now take for granted.”

Over the course of time, integration of functionality has changed. “As ecosystems started to emerge, they were really about third- party aftermarkets, planetary ecosystems, if you will, with products at the core, surrounded by peripherals and other things you might add to it.”

A good example of this is video game consoles. Ecosystems are a way to drive technology and innovation, and they represent an industrial shift that will become even more important in the future, he said. “Old ecosystems were built around single products. New ecosystems are built around entire product categories, entire industries, and it is the linkage between the players within these ecosystems that are as important as the products themselves. As devices become increasingly connected, there are a small number of degrees of separation between anything you might own. That connectivity is forming this Internet of Things. While that’s a very broad term, we are headed toward an Internet of connected things. As we get there, the ecosystems, which are ever-expanding, are going to become really, really important.”

Pravin Madhani, General Manager, IC Implementation division at Mentor Graphics, agreed. “Semiconductors are touching lives in every part of the world, and the systems are getting more and more integrated. Remember 10 years back, we used to just talk about system-on-a-chip. Today, it is truly a system-on-a-chip on a smartphone because you have a complete mobile office on your smartphone. Everything from processing to video-on-demand, cloud computing — everything is available on the mobile device. This could not have been possible without, in some sense, the accelerated rate of innovation in every aspect of the design ecosystem.”

The semiconductor ecosystem consists of chip design, methodology, EDA tools, IP, manufacturing, and system-level software. “Now what happens will depend on what application you are going after,” Madhani said. “That will define what ecosystem you need for it. For example, if you are doing something processor driven, you will need everything around a system of processors and what it needs around it. On the other hand, a big advancement is happening on the networking side with including LTE, 4G, 5G — very high bandwidth for users to watch video on demand, which is not as much about processing as it is about the SerDes and the PHYs on that side. Today the biggest chip at 16nm is 17 billion transistors, and in the next five years we will see 100 billion transistors on a chip at 5nm, which is what is needed to go after self-driving cars for very complex functionality and super-fast response time. As a result, every aspect of the design ecosystem will have to advance based on the application you’re going after.”

Third-party IP
When it comes to the IP ecosystem, Frank Schirrmeister, senior group director, product management in the System and Verification Group at Cadence, observed there’s a distinct ecosystem from a development perspective in three areas: software development, verification, implementation. “The ecosystem corners around the IP just like you have an x86 ecosystem for processors. You have an ARM ecosystem, and there are new ecosystems forming right now for the IoT. It’s really about the associated enablers around it. If you look into the ARM ecosystem, for example, you have software development, you have applications, you have development enablement for verification and software development. You have implementation partnerships where it’s essentially how to implement this — which technology, and the examples where people have done it before. These ecosystems will strengthen and develop further around the IP. There are similar things happening for Tensilica cores.”

Partnerships are key here, he stressed. Does this IP work together with the processor correctly in the subsystem? The cornerstone here is the processor.

In addition to hardware-related aspects, couldn’t an ecosystem be centered around an application or platform like software, or around an operating system (Android or iOS) as opposed to the processor?

Schirrmeister said it can. “The cornerstone always enables something in the hierarchy above that. That’s how ecosystems are structured. The Android ecosystem enables applications on top of it. In the ecosystem then everybody who supports Android in a specific incarnation becomes part of that ecosystem implementation, so yes, there is definitely some centering around operating systems.”

“Another ecosystem in this vein would be centered on the mbed initiative from ARM,” he pointed out. “Of course that’s trying to also foster the implementation, but then you have already seen IBM and ARM announce what they are doing. The cornerstone here becomes the operating system for the pieces above it. Who benefits from it? The application guys on top of it can all use the same resources and they don’t need to worry about the implementation because they will simply take care of that.”

Whether or not the application becomes a cornerstone is another consideration, Schirrmeister continued. “It might not seem like it at first glance, but considering RF applications or algorithms — if you look at the system level above us, part of the ecosystem for system-level design and verification are most certainly the tools like LabView and Mathworks. I wouldn’t consider them the cornerstone, however. If you look into the Mathworks, at the end, the processor or the operating system, or the combination of both becomes a cornerstone again because out of the Mathworks tools software can be generated and targeted to a specific application, which is ARM-based, or just DSP-based. I would still claim here the cornerstone is the processor because that is what you target and execute on.”

The impact of consolidation
Above the level of the software and the hardware, the general business environment of the semiconductor industry also impacts the ecosystem. With the current cycle of consolidation occurring, the semiconductor ecosystem is starting to feel the impacts.

“There is an impact of all of this consolidation,” observed Drew Wingard, CTO of Sonics. “You can’t combine $100 billion worth of silicon companies without seeing some impacts to the ecosystem. The question is how does everybody operate in this coming new world? Do the resulting semiconductor companies end up having the same stuff because if you don’t have this asset, you go buy a company that has that asset. Does everyone now have the same capabilities and they are all just slugging it out as very general market providers? That’s exactly opposite of what’s going to happen. What you’re seeing is people going deeper into specific areas of expertise so resulting companies will be, in many ways, competitive on fewer fronts than what was happening before. In the long run, economically for the semiconductor industry, that’s a good thing. Many times semiconductor companies are their own worst enemies. They build these incredibly expensive chips and then then end up competing based upon the number of square millimeters of silicon inside the package, which is crazy. It means design cost was meaningless in the selling price of the finished product. That will change. As the semiconductor companies become more specialized, it probably does mean the ecosystem companies become more specialized, as well.”

At the same time, specialization comes at a cost, and the question is still open as to the resources to focus down each rabbit hole. Who is going to pay for all of the specialization?

Wingard doesn’t believe complexity is reduced. If the semiconductor companies can get control of their worlds, he believes their prices and profits should go up and therefore the amount they can afford to share with their ecosystem would go up. “How they choose to do that is interesting. You could argue if the ecosystem gets sufficiently specialized, you’re better just buying the ecosystem and consolidating that, as well. That really boils down to how few major players are at the end. We watched it in the systems space back in the rah rah days when if it wasn’t Cisco doing a billion-dollar acquisition at least once a month, you could imagine a world in which that kind of consolidation happened. But even in those days the companies relied upon innovation coming more from without than from within. I don’t see a model by which innovation isn’t still valued.”

He added that even Intel CEO Brian Krzanich is making a bet that the world of these makers is the future. “What’s different about that world? That world is enabled by ecosystem. The way that world attacks complexity is by trying to use what’s there, and by relying upon the ecosystem to provide enough there that it works. Traditionally silicon companies have not been very good at that. They are used to selling complex products to experts. The part of the semiconductor industry that has been pretty good at it is of cours the microcontroller segment, so it won’t surprise me to find out that we end up in a world where there’s a whole lot more of the digital chips that start to be designed more and more like microcontrollers.”

Conclusion
Each player must have an understanding of how and where they fit into their chosen ecosystem(s). Understanding what their role will be, and how to adapt to the evolving nature of product development is still a work in progress. Consolidation continues to disrupt the market, and the IoT/IoE is still just getting going.

But there is no doubt that ecosystems will continue to play a role, and most likely a growing role, for years to come. But what those ecosystems ultimately look like, and how they will change, is anyone’s guess.