Companies are beginning to acknowledge the potential of new markets and downstream revenue opportunities as they explore a more comprehensive “silicon to services” model that spans the data center to the mobile edge. More specifically, with eroding ASPs (average selling prices) and increasingly prohibitive design costs at ever lower nodes, many companies are searching for fresh revenue streams across a wide range of verticals comprising the Internet of Things (IoT).
However, with the IoT install base expected to increase by about 15% to 20% annually through 2020, security is currently perceived as both a major opportunity and a considerable challenge for the semiconductor industry.
In addition to services, the concept of open-source hardware (OSH) and building silicon from disaggregated, pre-verified chiplets is beginning to gain serious traction as companies move to slash costs and reduce time-to-market for heterogeneous designs.
Specific strategies to unlocking the full potential of silicon and services will undoubtedly vary, which is why it’s important for us to explore a future in which semiconductor companies, along with various industries, organizations, and government offices, play an open and collaborative role in helping to sustainably monetize both silicon and services.
Acquisitions and industry consolidation continued apace in 2016 and 2017:
- Analog Devices acquired Linear Technology
- Infineon acquired International Rectifier
- ROHM acquired Powervation
- Renesas acquired Intersil
Major semiconductor players positioned themselves to better compete across multiple verticals, including cloud-based computing, artificial intelligence (AI), and self-driving vehicles. According to KPMG, many companies increasingly see Mergers & Acquisitions (M&A) as the only way to drive real revenue growth, giving a new emphasis to the “make vs. buy” question, with many selecting the “buy” answer.
Concurrently, chip development costs continued to rise and saliently affect the number of designs at advanced nodes. More specifically, the total number of Advanced Performance Multicore first-time SoC design starts has been roughly flat to only slightly up over the past five years. Although design price tags have been steadily increasing since 40 nm, analysts are most concerned with the acceleration of design costs at 7 nm and 5 nm.
As Rich Wawrzyniak, a senior analyst at Semico Research confirms, design starts beyond 10 nm will be constrained by rising development costs. While the total number of designs that migrate to new nodes may not be appreciably different than previous process geometry upgrades, Wawrzyniak says the time frame for such transitions by the majority of companies will be more protracted.
New models for both R&D and revenue are clearly needed, as heightened industry consolidation and eroding ASPs are unsustainable in the long term. This is precisely why the industry is looking toward the Internet of Things to create additional revenue streams, with McKinsey Global Institute (MGI) analysts estimating the IoT could have an annual economic impact of $3.9 to $11.1 trillion by 2025 across multiple verticals. However, with the IoT install base expected to increase by about 15% to 20% annually through 2020, security is considered to be both a major opportunity and challenge for semiconductor companies.
As such, MGI recommends creating security solutions that allow for semiconductor companies to expand into adjacent business areas and develop new business models. For example, companies could help create end-to-end security offerings, which are essential to the IoT’s success. Ideally, says MGI, the industry should play a leading role when developing such offerings, to ensure they obtain their fair share of the value chain.
From our perspective, end-to-end IoT security solutions deployed as a platform as a service (PaaS) are critical in helping semiconductor companies generate renewable, downstream revenue for specific services. For customers, PaaS offers an easy way for customers to securely develop, run, and manage applications and devices without the complexity of building and maintaining elaborate infrastructure.
Such security solutions, which could also leverage a hardware-based root-of-trust, should support device identification and mutual authentication (verification), routine attestation checks, secure over-the-air (OTA) device updates, disaster recovery and key management, as well as the decommissioning and reassignment of keys to better manage devices and mitigate various attacks, including distributed denial of service (DDoS).
Inaccessible silicon—such as chips embedded in IoT smart-city infrastructure—could offer semiconductor companies the opportunity to implement a long-term PaaS “silicon to services” model. Indeed, future smart-city infrastructure will almost certainly be designed with chips in difficult-to-reach locations, including subterranean water pipes, air-conditioning ducts, and under streets and parking lots.
Intelligent street lighting, responsive signage, and next-gen Bluetooth beacons also require future-proofing to avoid constant physical maintenance and upgrades. Therefore, silicon powering smart-city infrastructure should be capable of supporting secure in-field feature configuration, along with various PaaS-based services like advanced analytics, predictive maintenance alerts, self-learning algorithms, and intelligent, proactive interaction with customers.
The global smart-home market is projected to reach at least $40 billion in value by 2020. According to Markets and Markets, the growth of the smart-home space can be attributed to multiple factors, including significant advances in the IoT sector; increasing requirements for consumer convenience, safety, and security; a more pronounced need for energy-saving, and low-carbon-emission-oriented solutions. However, as we previously discussed, it’s critical to ensure that IoT security is implemented at the product design stage to prevent malicious actors from exploiting smart-home devices and causing service interruptions.
In addition to potentially lucrative cybersecurity opportunities for semiconductor companies, smart-home devices offer the promise of creating recurring revenue streams to support a sustainable “silicon to services” model. As an example, MarketingInsider's Christoper Dean highlights Amazon’s popular Echo devices. With at least 15 million Echoes already sold, Echo users are likely to double as active Amazon consumers, leveraging the device to monitor wishlists and discover items they’re subsequently prompted to purchase. Meanwhile, Nest is using thermostat data as a platform to offer energy-management services to utility companies across the United States, with companies paying for meaningful and actionable customer information on a subscription basis.
According to IC Insights, chip sales for automotive systems and the IoT will grow 70% faster than total IC revenues between 2016 and 2021. Specifically, IC sales for automobiles and other vehicles are forecast to jump from $22.9 billion in 2016 to $42.9 billion in 2021, while revenues for IoT functionality will increase from $18.4 billion in 2016 to $34.2 billion in 2021.
The projected increase in automotive chip sales is hardly surprising, as modern vehicles are essentially a network of networks equipped with a range of embedded communication methods and capabilities. However, this means vehicles are now more vulnerable to cyberattacks than ever before.
Potential security vulnerabilities include unprotected vehicle-to-vehicle communication, the unauthorized collection of driver or passenger information, seizing control of critical systems such as brakes or accelerators, intercepting vehicle data, tampering with third-party dongles, and altering over-the-air (OTA) firmware updates. In terms of the latter, automotive manufacturers are now focusing on providing secure OTA updates for various systems, with the global automotive OTA update market projected to grow at a CAGR of 18.2% from 2017 to 2022 and reach $3.89 billion by 2022.
Automotive manufacturers are also working to ensure that the vehicle supply chain is kept free of stolen and counterfeit components. Nevertheless, a wide range of gray-market devices can still be found powering high-value modules such as in-vehicle infotainment systems and headlights, as well as in critical safety systems including airbag modules, braking modules, and powertrain controls. Shielding vehicle peripherals and components against tampering by implementing a range of layered hardware and software security solutions have therefore become a priority for a number of automotive manufacturers.
In addition to implementing layered security solutions, the semiconductor industry would clearly benefit from adopting an IoT "as-a-service" approach to the automotive sector. As an example, companies could deploy sensor-based vehicle systems that proactively detect potential issues and malfunctions. This solution, which, in its most optimal configuration, would combine silicon and services, could be sold as a hardware and software product, or deployed as a service with subscription fees generated on a monthly or annual basis.
Medical and Healthcare
Long-life implanted medical devices will undoubtedly require a high degree of future-proofing from the semiconductor industry to avoid frequent physical upgrades and maintenance. Srihari Yamanoor, an R&D design specialist at Stellartech Research Corp., notes that medical devices will ultimately be customized to meet the needs of individual patients, thereby increasing the application of precision medicine.
Moreover, the health insurance industry is expected to leverage machine learning to optimize and reduce the cost of care, while digital health devices will also be used by the insurance industry to identify at-risk patients and provide assistance. Medical devices, especially implantable models, should therefore be designed to support a “silicon to services model” via in-field feature configuration and secure OTA updates, as well as PaaS-based services, including the collection and analysis of relevant data; proactive maintenance, advanced algorithms; and an intuitive UI for both patients and doctors.
Open-Source Hardware and Disaggregated Chiplets
Alongside services, open-source hardware offered by organizations and companies such as RISC-V and SiFive have begun to positively disrupt the semiconductor industry by encouraging innovation, reducing development costs and accelerating time-to-market.
The success of open-source software—as opposed to a closed, walled-garden approach—continues to set an important precedent for the semiconductor industry. Faced with prohibitively expensive development costs, a number of companies are opting to avoid unnecessary toll collectors while placing more of an emphasis on open-source architecture as they work to create new service-centric revenue streams.
In addition to open-source hardware, the concept of building silicon from pre-verified chiplets is beginning to gain traction as the semiconductor industry moves to slash costs and reduce time-to-market for heterogeneous designs. According to Semiconductor Engineering's Ann Steffora Mutschler, the chiplet concept has been on the drawing board for some time, although it was historically perceived as a potential future direction rather than a tangible solution in the shadow of waning Moore’s Law. This perception is beginning to shift as design complexity increases, particularly at advanced nodes (10/7 nm), and as new markets that require semi-customized solutions coalesce.
The concept of pre-verified chiplets has piqued the interest of the U.S. Defense Advanced Research Projects Agency (DARPA), which recently rolled out its Common Heterogeneous Integration and IP Reuse Strategies (CHIPS) program. In collaboration with the semiconductor industry, the successful implementation of CHIPS would see a range of IP blocks, subsystems, and chips combined on an interposer in a 2.5D-like package.
The CHIPS initiative took center stage in August 2017, when participants from the military, commercial, and academic sectors gathered at DARPA headquarters at the official kickoff meeting for the Agency’s Common Heterogeneous Integration and Intellectual Property (IP) Reuse Strategies program.
As DARPA's Dr. Daniel Green detailed during the conference, the program seeks to develop a new technological framework in which different functionalities and blocks of intellectual property—among them, data storage, computation, signal processing, and managing the form and flow of data—can be segregated into small chiplets. They then can be mixed, matched, and combined onto an interposer, somewhat like joining the pieces of a jigsaw puzzle. In fact, says Green, an entire conventional circuit board with a variety of different but full-sized chips could ultimately be shrunk down onto a much smaller interposer hosting a huddle of yet far smaller chiplets.
According to DARPA, specific technologies that could emerge from the CHIPS initiative include compact replacements for entire circuit boards, ultra-wideband radio frequency (RF) systems, and fast-learning systems for extracting interesting and actionable information from much larger volumes of mundane data.
Perhaps not surprisingly, the semiconductor industry is already eyeing a disaggregated approach in the form of SerDes chiplets and specialized low-power, application-specific die-to-die interfaces. To be sure, viable silicon disaggregation can be achieved by moving high-speed interfaces like SerDes to separate die in the form of SerDes chiplets, shifting analog sensor IP to separate analog chips and implementing very low-power and low-latency die-to-die interfaces through MCM or through an interposer using 2.5D technology.
Beyond leveraging known-good die for SerDes in more mature nodes (N-1) or vice versa, disaggregation is expected to facilitate the creation of multiple SKUs, while optimizing cost and reducing risk. More precisely, disaggregation will see SoCs broken out into higher-yielding, smaller dies and allow companies to create specific designs with multiple variants. Indeed, die-to-die interfaces can more easily accommodate diverse applications across memory, logic and analog technology. In addition, die-to-die interfaces don’t require a matching line/baud rate and number of lanes, while FEC may or may not be required depending on latency requirements.
It should be noted that multiple companies are actively pursuing SoC/ASIC aggregation for switches and other systems. Similarly, the semiconductor industry is developing ASICs with die-to-die interfaces on leading FinFET nodes, while at least one next-generation server chip is being designed with disaggregated I/Os on a separate die.
Over the past five years, the semiconductor industry has faced numerous complex challenges. These include increased development costs, eroded ASPs, market saturation, and heightened, yet unsustainable M&A activity. Through 2018, the semiconductor industry continues to seek a return to stability and organic growth within the parameters of a new business paradigm that’s both viable and collaborative. Within this context, semiconductor companies are acknowledging the potential of new markets and downstream revenue opportunities as they explore a more comprehensive “silicon to services” model that spans the data center to the mobile edge.
This includes end-to-end IoT security solutions and PaaS-based services such as in-field feature configuration, advanced analytics, predictive maintenance alerts, self-learning algorithms, and intelligent, proactive interaction with customers. In addition to services, the concept of open-source hardware and building silicon from disaggregated, pre-verified chiplets is beginning to gain traction as companies move to slash costs and reduce time-to-market for heterogeneous designs.
Specific strategies to unlocking the full potential of semiconductors will undoubtedly vary, which is why it’s important for us to explore a future in which the industry, along with various research organizations and government offices, plays an open and collaborative role in helping to sustainably monetize both silicon and services.
For more information on this topic, check out Rambus’ site.
Shrikant Lohokare, Ph.D., is Vice President and Executive Director, North America for the Global Semiconductor Alliance.