Looking For The Next Big Thing

With slowing down or coming to an end, finding the next big thing may be very different than it was in the past. We cannot assume that more of the same will be a winner. The semiconductor industry has been blessed with two new product categories that have catapulted it through what should have been a very difficult period with barely a scratch. Those technologies were the smart phone and the tablet markets. But with those markets maturing, the industry is looking for what comes next?

Many people are looking towards the (IoT) as the next big opportunity, and it may well be a very interesting change for the industry. “It is likely to be something of a self-fulfilling prophecy,” says Bernard Murphy, chief technology officer for Atrenta. “As phone growth starts to flatten out, attention is shifting to IoT, both on the production and consumption sides, so we’ll have a herd effect.”

It is already beginning to reverse the decline in design starts, which has been looming over the semiconductor industry for the past several nodes. “The number of distinct designs is likely to be significantly larger than we have seen in the past,” says , a Cadence fellow. “At the high end (phones, tablets and servers) we will see a consolidation in the number of designs and these will be dominated by the elite few. The opposite will happen in the IoT space, and this will change the nature of the design process and the demands on the EDA industry.”

But what is the Internet of Things? “The nature of the devices is that they are small and sensor-centric,” explains Rowen. “They are gateways into the real world. In most cases the sample rates are low enough that you don’t have to run at multiple gigahertz and you don’t have to do trillions of operations per seconds in order to deliver the necessary experience.”

Gaining speed
Another important question is what is the starting point. “The momentum at the moment is around wearables,” says Ron Lowman, strategic marketing manager for IoT at Synopsys. “Health and fitness is big, and we are also seeing interest in the smart home and building automation space.”

Andrew Caples, senior product manager for the Nucleus product line at Mentor Graphics, agrees that health-monitoring devices are interesting. “They are monitoring your body’s physiological parameters but none of these devices are actually connected today. While it may be interesting to know how many calories you have burned, your physician has no means to access that data or to use it any way that turns it into something more meaningful. It becomes real when you can take this vast amount of information and turn it into meaningful data that can be used to improve our lives.”

There’s a big difference between nice-to-have health monitoring and must-have health monitoring, though—the fitness versus medical side.

“I absolutely believe the IoT will be real and important, but I also believe that fitness is not the killer app,” said Drew Wingard, CTO of Sonics. “If you think about the Gartner hype cycle, we are not yet in the trough of disillusionment. We are still in the upswing, which is the point where you ignore all the failed devices. Gartner’s presentation is that only after you go through that trough do you find out what’s meaningful. The ones that do will come out stronger because at that point you peel back the hype.”

If devices for the IoT are simpler, it is not obvious who will see the advantage and be the biggest beneficiaries. This is where significant disagreement starts. “I believe large companies will do top-down deployment of hundreds of SKUs (stock-keeping units) of IoT-class chips and that will make them quite successful,” says , president and CEO of eSilicon. “Given the relatively low cost and complexity of chips for the IoT, it will be served equally by open-source semiconductor design.”

Jonah McLeod, director of corporate marketing for Kilopass Technology is in agreement. “You can buy a complete chip solution from a variety of chip suppliers, add a sensor, packaging and software and you’re in business. This market will be largely driven by software with much of the chip design being done as application specific standard products.”

If this is true, things may not look too rosy. “The market for these IoT devices will not be a huge boon to the semiconductor market,” says Kurt Shuler, vice president of marketing at Arteris. “It will not significantly help the fabs, design houses, or EDA tool vendors. The reason is that these actual IoT device designs are quite simple: microprocessor, sensor, radio, battery. They will be cheap and very low margin commodities.”

Atrenta’s Murphy shares those sentiments. “It will be good for semiconductor folks with strength in MCUs, sensors, radios, especially with support for emerging network standards, but the margins will be low. Just more volume for the fabs, though potentially beneficial in soaking up volume on older nodes and nothing new in the mechanics for EDA, especially if cost drives design to less aggressive nodes.”

But not all agree. “They will be built in much greater variety because there will be more fundamental diversity in the features that are wanted,” says Rowen. “They will often be computationally hard relative to the power budget, and microcontrollers may not be the best way to process those signals. As a minimum an optimized DSP and perhaps a highly specialized, sensor specific DSP may be required to do it at the specified energy level. Your choice of sensor will dictate the DSP. A Microcontroller may exist that gets woken up once a second to check for reconfiguration of the network but it may not be doing the heavy lifting.”

Lowman goes even further: “We do not believe that you can take standard IP off the shelf and use it for these markets. We see the IoT as having very unique needs and vendors have to customize their IP into offerings that are well suited to that space.”

The phone and tablet markets have been helped by another trend. “Consumers have grown accustomed to paying extra for new technology devices and regularly repeating those purchases as technology advances,” points out , chief technology officer for Carbon Design Systems. “Only time will tell if consumers can be coaxed into replacing their internet-connected devices as often as they replace their cellphone.”

There are other factors that can also affect the complexity of IoT devices. Caples provides one example. “Think about a very simple application such as turning off a light remotely. What kind of silicon will we require to do this? It might require low-featured silicon that only has to accept a packet and respond to it. However, these unsophisticated nodes may need protection to stop your neighbor from hacking in and turning your lights on and off. This hacking could be extended to playing with your refrigerator, air conditioning or anything else that is connected. You have to think about security, authentication and encryption.”

That is when things can start to get more complex. “There are a lot of questions about security and everyone wants it, but there are few regulations that require it today,” points out Lowman. “It is an open book. People know the risk of not having it, but there is a cost associated with it and without clear direction in various markets there are few incentives to accept the cost.”

McLeod sees the same issue. “Every sensor that connects to the network is vulnerable to cyber attack, which may be insignificant for activity monitors, but potentially harmful for medical sensors.” Murphy agrees saying, “What we have in security today (trusted execution environments) is necessary but nowhere near sufficient.”

And yet it is only with this type of connectivity that we can expect to see the real gains. “By allowing things and people to be connected, we can truly live a life as depicted in sci-fi novels,” says Jin Zhang, senior director of marketing for Oski Technology. “Medical diagnosis and treatment that does not require you to be in a doctor’s office or hospital will have a significant impact in our daily life.”

Can we expect the industry to get behind security? “The problem is that this is an asymptotic process,” explains Caples. “You can spend a large amount to secure a device and never eliminate the possibly of intrusion. People will have to trade off how much they invest in security. I think the big device providers will take security seriously. Industrial, medical and automotive will take it seriously. Security plans will be part of the design process and the software stacks. In other areas it may not win out against time to market. Companies, and ultimately users, will accept the risks associated with a security breach.”

EDA tools may also have to go through significant changes to satisfy the demands of IoT providers. “We will see more tools that concentrate on the integration of pieces,” says Rowen. “There will be more focus on IP, both computational and interface, and more involvement in reducing the conceptual boundary between hardware and software.” Rowen also sees a growing significance for analog tools due to the increasing proportion that sensors and wireless communications take up on the complete chip.

“On the sensor technology side, MEMS demands a cross engineering discipline that doesn’t exist today,” points out McLeod. “Digital designers will need a knowledge of electrical engineering, mechanical engineering, materials science and environmental engineering, and possibly also an understanding of biology and human physiology.”

Some existing tools also may see a boost. “Verification will have to play a prominent role and hardware may be the ultimate winner here,” claims Lauro Rizzatti, an emulation expert. One thing that everyone agrees on is that power optimization at all levels will become a lot more important than it is today. While lower-level tools already exist, new ones at the system level will become important.

“We need better estimation tools that are bound into scenario analysis. This will allow the design teams to confirm how it is all going to play together,” says Rowen.