In a world where software updates occur every day and hardware upgrade cycles are speeding up, the ability to make the right bet in silicon is becoming both more important but also more difficult. The people heading the major chip companies can’t just look 18 months into the future, they have to see out five or even ten years in order to design and manufacture the right chips for tomorrow’s hardware. Their designs will make the future, by defining in hardware what is even possible for software to accomplish.
Many of the men leading these companies, and they are men, are aware of their need to seem visionary, and tend to speak with grand rhetoric and bold statements about the future. But Simon Segars, the incoming CEO of ARM isn’t even inclined to big hand gestures. Instead, he speaks in a low, even voice and shies away from making grandiose statements or predictions. As a current resident of Silicon Valley, he is a reserved Brit working among the frenetic futurists of the Bay Area’s technology entrepreneurs and executives.
Next month Segars, currently the president of ARM, will take over as the CEO of the chip licensing firm, replacing Warren East who has steered for the last 12 years. Segars, who has actually been at ARM longer than East, has been the president of the company since January of this year, and has moved around a lot within the company’s business units as well as its offices.
He’s currently based out of the company’s San Jose office, where he plans to stay for the next two years while his daughter completes high school. In a recent interview, Segars noted that ARM only has one customer in the U.K., and even East spends only a third of his time in England. “The job of ARM’s CEO is a roving one,” Segars says. However, he said he doesn’t have any plans to move ARM’s headquarters to Silicon Valley.
Changing captains in calm waters
That fits with the ARM mentality. The company, which was formed in 1990, is a pragmatic and steady member of the semiconductor world. Unlike more flamboyant executives, ARM’s East isn’t one to insult the competition — although because of its licensing model, its competition is hard to define.
ARM designs processor cores that companies ranging from Apple to Qualcomm use to build the chips powering the mobile phones that have so change our world in the last seven years. Companies can either license the ARM architecture as is and pay a fee on every chip sold, or they can take a deeper architectural license that allows them to tweak the core to build new functions and features for their particular market. With its emphasis on power efficiency over performance (although it’s pushing the envelope there as well) ARM has maintained a dominant position in the growing mobile market, while rival Intel has failed to place one of its x86-based chips inside a mass market handset.
East has said he’s leaving at this point because remaining could act as a brake on inovation. Plus the company is at a high point, making a shift in leadership less fraught.
|Application||Market share percentage|
|Mobile BT and Wi-Fi||75|
|Digital TV/set-top box||45|
|Mobile computer apps||40|
Meanwhile, it has also made bets — some as far back as nine years ago — on the data center and networking processors as well as the rise of graphics processors. Much like Intel drove the x86 architecture to prominence with its PC chips in the late 80s and early 90s (and then pushed those into the data center and other realms), ARM’s power-sipping architecture is taking on more and more workloads as computing and connectivity are embedded in more and more places.
In the first quarter of this year ARM was the technology behind 2.6 billion chips. It also has added several new licensing partners from surprising areas, such as AMD, which has taken a license to make chips for the data center.
From mobile first to data first
So what does Segars plan to do as the next captain of ARM? Graphics cores will be one important area. While GPU cores are still not a huge business for ARM, Segars pointed out that licensees sold less than 50 million of its Mali cores in 2011 but 150 million in 2012. ARM estimates its GPUs comprised about 13 percent of the GPU market in 2013.
Other forward-thinking moves were the launch of the ARM Cortex A15 chip in 2010, which helped push ARM into the telecommunications networking equipment world, and the decision in 2010 to invest in Calxeda, design a 64-bit ARM core and give an architectural license to Applied Micro. Both moves were integral to bringing an ARM processor into the data center server market, where Intel has dominated but where power was becoming a concern.
This is what Segars sees for the future of ARM: More devices with computing and connectivity trying to share data on a variety of networks. You can call it the internet of things or just the obvious direction we’re heading as a technology-inclined society.
But Segars sees more people getting smartphones, more sensors and all of those billions of nodes sending back data to the cloud or even to local hubs for processing. Thus ARM is pushing it’s application processors and GPUs for all kinds of mobile phones; fancier, but low power chips for networking boxes; ARM cores for inside the data center for data analytics, compute and storage and microcontrollers for those myriad sensors.
Software is eating the world, but ARM has to stick with chips
Thus ARM’s big challenges for the next few years are less related to identifying the next big thing than building out this vision of the world. When it comes to the internet of things he’s trying to figure out the lay of the land: how do we stay open to innovation while addressing privacy and security? On the data and cloud computing side, he’s adapting ARM to a world where software is eating everything.
“Everything is virtualizing, where you can run everything in software on top of the hardware so you have more flexibility,” Segars said. “That trend is catching up on the networking side.” He noted that while running applications directly on the hardware will always offer better performance, ARM is hoping to win in these areas by being open to the virtualization trend while still offering power-efficient performance.
But the shift to a more software-centric world leaves Segars walking a thin line. He sees the importance of software to ARM’s success as an architecture, but it cannot develop its own software because that wouldn’t serve its chipmaking customers. In many cases, ARM’s customers (and other chipmakers) are trying to respond to this same challenge of software gaining in importance (and value) by building their own software.
Intel, for example has developed its own distribution of Hadoop as well as its own software-defined networking strategy that will put it in competition with switch makers. That’s a story playing out across the chip ecosystem, as virtualization means one could run a variety of chip architectures in one data center.
While ARM can’t try to build an enticing software layer to add value to its architecture, it can join partnerships to help make sure software that runs on the ARM architecture exists and can match features with Intel. It has joined partnerships such as the Linaro effort, as well as a security joint venture with Gemalto and Giesecke & Devrient to offer its customers an ecosystem that supports the ARM architecture without stepping on their own software development efforts.
Is ARM the architecture for the internet of things?
As ARM attempts to break the x86 hegemony in the data center it’s also extending it’s business further down the computing chain to microcontrollers — the tiny chips that act as the brains in washing machines, smart sensors and other devices. While mobile phones and tablets are still a growing market for ARM, albeit one that Intel is also targeting, the possibility of a microcontroller in every appliance or a sensor on every bridge is one that ARM can’t ignore. Microcontrollers may not have the margins that application processors have, but they will be everywhere.
According to Gartner analyst, Adib Ghubril, the global MCU market is roughly $15 billion but ARM’s cores are limited to the large 32-bit cores, where ARM has about 50 percent of the $5 billion global market. Ghubril says ARM’s going to go after the whole market, but to go for the whole thing means getting a 32-bit core to act like an 8-bit core. That’s possible and some vendors are doing that. Plus, thanks to the internet of things and the automotive markets, Gubril expects the 32-bit MCU market to grow, “like gangbusters” over the next five years.
And as Segars discusses the various ways one might implement security on networked devices, it becomes clear that ARM might find a new market for application processors inside home and industrial hubs where data from sensors is aggregated and processed before getting sent up to the cloud. He also discussed some of the subtle problems of the emerging market in data around connected devices that might find an answer in silicon.
If we want to use connected devices for setting prices or even medical monitoring, authentication will be important. One thought is to employ biometric security to authenticate the person wearing a device, which then would require smarter microcontrollers and more sensors. From Segars’ perspective, everywhere he looks he sees a potential home for ARM’s architecture.
That may not be realistic, but much like Intel rode the wave of cheaper general purpose computing, back when the IT world saw the value in taking “good enough” chips at a lower price, ARM is hoping that today’s focus on lowering the energy consumption of devices, as well as the benefits of having a wide ecosystem of silicon providers based on its architecture, will benefit it in more markets in the years to come.