Predictions: Markets And Drivers

Part 1: What advancements can we expect to see in 2018, which markets will drive the industry, and what are the major challenges that have to be addressed?


Semiconductor Engineering received a record number of predictions this year. Some of them are just wishful thinking, but many are a lot more thoughtful and project what needs to happen for various markets or products to become successful. Those far reaching predictions may not fully happen within 2018, but we give everyone the chance to note the progress made towards their predictions at the end of the year. (See Reflection On 2017: Design And EDA and Manufacturing And Markets.)

Predictions are divided into four posts this year. This one covers markets and drivers. Part two will look at manufacturing, devices and companies and part three will cover methodologies and tools. In addition, the outlook from EDA executives will be provided in a separate post.

Many of the drivers for the semiconductor industry are tightly interwoven. The , automotive, machine learning and others are coming together to set the direction for technology and tools and each area is leveraging the advancements in the others. , safety and power are common factors that influence designs in all of these categories. It is thus difficult to fully separate these issues.

The industry is in a rapid growth and learning phase. “As we head into 2018, it’s worth remembering the humble beginnings of three of the most visible companies setting the agenda for the semiconductor industry,” says , managing director of Lanza techVentures. “Google was an online search engine, Amazon sold books online, and Facebook started life as an online dating service. Their initial strategic views are nothing like the path they are on today. These technology giants reinvented themselves as they went along and, in their latest incarnation, are maximizing their influence to change society in ways we haven’t begun to imagine.”

Looking forward, Lanza says that reinvention is crucial. “The semiconductor design ecosystem has proven time and again, without an assumption, it could predict the future. And yet it made possible the IoT, creating ubiquitous connectivity and the digital society where physical continuity and dimension are no longer important. The need for hardware components and solutions to support these emerging markets will become obvious in 2018. It’s up to the semiconductor design ecosystem to set a path to respond, as it always has.”

What is causing the industry to enter this learning phase? “After the introduction of the iPhone in 2007, the smartphone segment experienced triple digit growth for many years,” recalls Tom Wong, director of business development for Cadence‘s IP Group. “However, in the past five years, we have seen the growth rate slow to single digits. The excitement in smartphones and applications processors has diminished, having been replaced by the emergence of automotive SoCs, automated driver assistance systems (ADAS) chips, neural network chips and artificial intelligence.”

Two product categories are attracting most of the attention. “Machine learning will be a hot market, but it is not clear that end-user applications are there yet to sustain revenue expectations to maintain current level of enthusiasm,” says Charlie Janac, president & CEO for ArterisIP. “The one market that will sustain the hype is automated vehicles, where investments will continue to be large. This investment will be supported by strategic considerations and ever-growing automated vehicle system revenue. Yes, it may take 25 years for the full impact of the new transportation economy to unfold, but major hardware decisions will be made in 2018.”

Automotive was the fastest growing segment on the application side. Semiconductor content is expected to rise from $300 to more than $3,000 per car for consumer automated driving vehicles and more than $30,000 per unit for automated taxis, platooned trucks and geo-fenced self-driving vehicles. “The gains from the transportation economy revolution will tally into hundreds of billions of dollars by reducing the loss of life, improving asset utilization and enabling novel transportation business models,” says Janac. “I believe automated vehicles and their associated economic benefits will transform society more than smartphones did.”

The other category that gets everyone’s attention is consumer devices. “In 2018, we’re expecting tremendous growth in smart home devices, particularly in the smart speaker market,” notes Darin Billerbeck, president and CEO at Lattice Semiconductor. “As in the recent release of the devices from Amazon, Apple and Google, voice-enabled assistants are anticipated to be adopted in wearables, home appliances and autonomous vehicles. Our expectation is that, in 2018 and beyond, the industrial and automotive industries will experience meaningful growth as auto infotainment, ADAS and factory automation, to name a few, continue to leverage FPGAs for sensor bridging, embedded vision and machine learning.”

Consumer devices will keep evolving. “This holiday season brought us connected home devices galore: security cameras, lighting controls, digital assistant portals, and environmental monitoring/control,” says Marc Greenberg, group director of product marketing in the IP Group at Cadence. “At tens or hundreds of dollars, those were gifts under the tree this holiday. In 2018 some of these connected home devices will retain their price but get a lot smarter with artificial intelligence – recognizing their users, authenticating new users, and identifying threats. The connected home devices without AI will get picked up as stocking stuffers for a few dollars by the holiday season of 2018.”

Big changes for IoT
The IoT has been a hot topic for a few years now, but there is a significant shift happening in its architecture. “Centralization is reaching its limit,” says Mike Fitton, senior director for product planning and business development at Achronix. “The volume of data that will be needed to drive the next wave of applications is beginning to force a change in direction.”

Fitton provides an example. “Today, the intelligence in ADAS is largely self-contained, with scenes captured by the built-in cameras and radar systems processed purely within the vehicle. Only a tiny proportion of this data is relayed to the carmaker’s servers, where it may be used to help with predictive maintenance and collect statistics on the performance of the ADAS software. A fundamental limitation of today’s cloud-centric computing model lies in the laws of physics. The transmission of data by photons through fiber-optic cables is restricted to 70% of the speed of light in a vacuum. A distance of 1,000 km adds 10ms of round-trip delay. There are also the delays incurred by multiple switch-routers and other networking infrastructure that needs to be added. To support the millisecond-scale response times, compute resource need to be brought much closer to the point of delivery — to the network’s edge. Cloudlets.”

Pushing more computing to the edge, rather than all the way into the cloud, appears to be gaining steam as a design direction because of the growing amount of data that needs to be processed. “The IoT continues to drive innovation in areas such as artificial intelligence and human computer interaction, where devices at the edge (and not the cloud) need to capture, connect, analyze, infer, learn and make decisions,” explains Billerbeck. “More intelligence at the edge also solves latency, privacy and bandwidth issues.”

Edge devices are about to get smarter, too. “Smart IoT devices will push the edge further and further from the enterprise expanding the size of the core network,” says Andrew Dauman, vice president of engineering at Tortuga Logic. “Their volume will increase by 10-100 fold as this segment continues to accelerate. It is perhaps unbounded as everything can be made to be a connected device. Huge investments in the end-to-end ecosystem will support this expansion.”

“IoT will be used extensively by various industries to improve their bottom line by lowering operating costs, increasing productivity and developing new products,” says Ranjit Adhikary, vice president of marketing for ClioSoft. “Sensors and external data gathering implements are becoming an essential catalyst for IoT industry growth.”

Some of this continues to piggyback on the smartphone developments. “Growth areas will include low power and secure edge devices, faster data connectivity between the edge and cloud and possibly the beginning of the 5G roll out,” adds , CEO for Mobiveil.

Machine learning
“How can any prediction for 2018 not mention AI, which has displaced IoT as the hot topic?” asks Gregg Recupero, CTO at Performance-IP. “The biggest reason is the drive for autonomous vehicles fueling the progress made in AI. Continued improvements to development tools are allowing designers to create and implement new and better network topologies. The propagation of these network topologies is opening new markets for AI.”

As we have seen processing moving to the edge for IoT, it is also likely to happen for machine learning. “Even training is likely to move to the edge, again because of the gravity of data,” says Fitton. “It will often prove impractical to upload enough data to enable good training even with high-ratio data compression.”

Machine learning also presents many new challenges. “Potential implications include issues with safety and security, which still need to be analyzed, understood and solved,” says , president and CEO of OneSpin Solutions. “Questions of accountability and transparency of machine learning models will give rise to new legislation in Europe. Spoofing of road signs, misleading ADAS systems, demonstrate the inherent limitations of current approaches and they will ask for engineering solutions. Finding solutions for these new challenges in 2018 will spawn technology development, more startups, and the need for additional tooling.”

Safety and security
Safety and security are two design attributes that may bring about significant changes to the industry. “One of the biggest market game changers for 2018, having come to the forefront during 2017, will be safety and security,” says Frank Schirrmeister, senior group director for product management at Cadence. We are fast approaching a connected world, and without security and safety built into all aspects of these complex systems, consumer adoption will be at risk.”

Safety and security are closely linked in automotive. “Crystal-gazing into 2018, functional safety seems a safe bet, no pun intended,” says Srikanth Rengarajan, vice president of products and business development at Austemper Design. “Rising cost structures in automotive electronics driven by Level 3+ autonomous driving systems and stringent ASIL-D requirements will force semi vendors and Tier 1s alike to move away from duplication models of safety analysis to system-level safety mechanisms. This will be coupled to rigorous verification techniques that rely on fault injection and/or formal-based techniques. The arrival of the next rev of the automobile safety standard in the form of ISO 26262-2018, with its all new Part 11 devoted to semiconductors, will change practices and tools across the board. Together, they promise to make 2018 the year that automotive Functional Safety comes of age, paving the way for Industrial IoT and beyond.”

The tide appears to be turning toward hardware as the location where security starts. “We are starting to see a big shift in the investment for hardware security, where the root of all trust begins in the silicon,” says Tortuga’s Dauman. “Hardware threat scenarios must be verified before products are released and deployed in the communications infrastructure. There already has been significant investment in development of secure silicon architectures and foundation building blocks. The verification of these concepts must now catch up to the complexity of the SoCs that implement them. We are seeing the shift from conversation to action, as silicon providers have felt the impact of gaps in security exposed in deployed products.”

ClioSoft’s Adhikary sees a similar trend. “As the IoT ecosystem continues to evolve, security will continue to be a major concern and companies will transcend toward a platform approach, where the security can be updated as necessary instead of trying to secure every device.”

Security is a complex issue, and there are multiple approaches to address it. “Integrating security functions into the hardwired logic that surrounds an eFPGA makes it possible not just to support encrypted uploads of virtual circuits into the fabric, but continually monitor them for potential breaches,” adds Fitton. “The hardwired logic can ensure separation of programmable functions that may be uploaded by different users and prevent them from spying on each other. Having both security and programmable logic integrated on-chip makes it extremely difficult, if not impossible, for an attacker with physical access to the system to eavesdrop on communications.”

The other concern that continues to affect all of markets and products is power. “The race to get to Level 5 autonomy will continue to heat up,” says , CEO of Silexica. “Still, questions remain about how you can do that without having a trunk full of power-hungry computers. Nvidia and Intel will need to deliver with their boards, but they can only do so much if the software isn’t fully optimized for heterogeneous platforms.”

New architectures will emerge. “In-memory processing could solve some power budget restrictions of deep learning if logic and memory process technology can be combined,” says Brinkmann. “A different example is huge machine learning chips composed of small processing elements, local memory and interconnect, that drive capacity requirements for EDA tools, including formal, for chip-level connectivity.”

“For today’s battery-powered electronic devices, energy per operation and reliability are key factors determining product planning,” says Lauri Koskinen, CTO for Minima Processor. “An increased number of features brought about by increased performance are still important, but now must be balanced against the need to prolong battery life. This can already be seen in product markets such as Bluetooth. Once the AI algorithms mature, it will be seen increasingly in other market segments such as voice wakeup audio.”