The semiconductor industry lies at the intersection of several high-growth sectors that are now reaching commercialization. The electronic components that are manufactured by the various players in this industry are essential for the manufacture of the billions of devices comprising the Internet of Things and mobile device ecosystems.
High demand exists for the chips that power our smartphones and for the complex servers enabling cloud services. The push for machine learning is also fueling strong growth for well-positioned players and attractive opportunities for start-ups. All this leads to positive growth in the semiconductor industry, with ~12% expected growth predicted in June 2017 for the coming year.
The Internet of Things:
As discussed previously, many organizations operating within the semiconductor industry are diversifying their operations to capture value across the technology value chain. Expanding from hardware design to software and systems integration, and designing for niche applications like wearables allows these firms to capture the growth driving multiple attractive sectors.
The small form factor and miniaturization of IoT sensors is also driving requirements for smaller chips with low power consumption, higher memory count and wireless connectivity (the later being accelerated even further by the coming 5G revolution).
However, the high demand does offer some drawbacks. The emergence of new players and growth of already established players means supply prices for raw materials are increasing, and will impact the bottom line and make it harder to compete without securing a long term raw material supply. Additionally, a diversification of devices is coming which will make it more complex (from a logistical and operational standpoint) for one player to compete in all relevant markets.
The Automotive sector:
A 2016 market research study by Markets and Markets found that the Automotive Electronics component of the Semiconductor M&A sector would grow at the highest rate from 2016 to 2020, followed closely by industrial electronics (both at ~6% CAGR). The automotive sector is undergoing significant changes at the moment: it too is trying to integrate new digital capabilities with a desire to electrify vehicles as a response to climate change.
Management consulting firm McKinsey cites the evolving demand of automotive chips as one of three key developments to provide opportunities for increased performance:
“Automotive chips now account for about 8 percent of total semiconductor sales, and current projections suggest that they will see about 6 percent annual growth through 2020—higher than the 3 to 4 percent growth predicted for the sector as a whole. That would put yearly revenues from automotive semiconductors in the $39 billion to $42 billion range.”
Though this presents significant opportunities, it does present a technical difficulty stemming from the necessity to couple numerous complex modern technologies together, ranging from wireless connectivity and artificial intelligence to cloud services and IoT.
Artificial intelligence and machine learning:
The push for AI is already creating true giants within this sector – NVIDIA being a prime example. The structure and operation of the neural networks that form the backbone of most of the artificial intelligence services and products we see today can operate faster and more efficiently using the architecture and hardware previously used for gaming and video processing cards. Coupling the strong manufacturing competency with a push to develop relevant partnerships in most sectors has been a fantastic strategy for NVIDIA thus far. Most people up and down the AI commercialization chain, from academics and researchers to startups and established companies, are using NVIDIA products to operate and to continuously develop new offerings within the sector.
However, though neural networks can do a lot of amazing things, we are still far from a true “sentient” artificially intelligent system. Such a system will most likely have to rely on other techniques besides neural networks to function at the levels imagined by science fiction writers and futurologists. This will likely require a different architecture – and presents a significant opportunity for those that are able to discover which architecture is best.
Ultimately, we may stumble on the right answer and not even know why this is – already today, some answers obtained through deep learning techniques (though correct) are not understood. For example, Deep Patient is a tool that can be used to diagnose schizophrenia accurately before doctors are able to – however, it cannot tell the doctors how it determined this, and the doctors are left with high uncertainty: should they rely on a “black-box” technology and prescribe potentially dangerous drugs to these patients when they do not even know how the AI system is making these predictions?
Conclusion:
Ultimately, we find that the semiconductor industry is currently in a great position – it forms the backbone of many high growth technologies that will be fundamental to our future lives and activities. Beyond the Internet of Things, the automotive sector and artificial intelligence, it can benefit from a myriad of other technology trends, such as the increased digitization and emergence of mobile devices or infrastructure requirements for the soon-to-come roll out of 5G networks globally.
