By Milind Gandhe, Chief Program Officer at MINRO at IIIT, Bangalore
The semiconductor industry is facing some interesting times today. For almost a decade, the industry after several rounds of consolidation, as a notoriously fragmented industry, looked to gain some pricing power. The past few years have been marked by multi-billion-dollar deals such as NXP’s acquisition of Freescale, and Softbank’s acquisition of ARM. Some of these deals also ran into a storm of regulatory scrutiny – Broadcom had to drop its acquisition of Qualcomm. Product mix shifted from digital processors needed by the communication vertical, towards analog/mixed signal products that gave far better margins with limited R&D costs – witness Renesas’ relentless acquisition push for analog companies such as IDT and Intersil. With these acquisitions, the company has tried to position itself as a leader in Analog, instead of its traditional positioning as a leader in MCUs.
Pressure on margins meant that investment was tightly controlled – both in R&D and in fabs. This under-investment was already making itself felt towards the end of 2018, when memory chip went famously in short supply. Several top executives from product OEMs made it a practice to visit Taiwan and South Korea regularly in order to secure supply chains. The COVID-19 pandemic over the past year and a half has brought the situation to a head now. The news cycle is dominated by delayed car launches due to shortage of chips. A generation of consumers used to impulse purchases of laptops and mobiles is now getting used to waiting times of weeks, if not months. Where is all this demand coming from? Or is the issue merely broken supply chains, as we are often told?
Amidst all the noise about semiconductor supply chain, we must not lose sight of a central fact: Semiconductor market is growing rapidly! World Semiconductor Trade Statistics, in its Spring 2021 Market forecast has said that it expects the world market to grow 19.7% to US$527 billion in 2021, and that the momentum will continue into 2022. World semiconductor sales are expected to hit a record US$573 billion in 2022, a further growth of 8.8%.
The growth is being powered above all by increasing demand for memory, a category often dismissed as commodity. Memory is the only device category that is expected to grow in double digits into 2022. Mordor Intelligence claims that the primary driver within the memory space is increased demand for non-volatile memory. In the past few years, one big driver for memory demand was the trend towards replacement of bulky mechanical disk drives in laptops with solid state drives, as well as replacement of tape-based back-ups by cloud back-ups. Going forward, there are multiple potential trends that would drive demand for memory. The move to cloud, and increased adoption of big data and AI would drive demand from the BFSI vertical. In automotive vertical, traditionally memory lean applications (like body and chassis control, drive train etc.) are likely to be replaced with memory rich applications like infotainment and autonomous driving.
Demand for Analog and mixed signal chips as well as sensors is showing solid, though not spectacular, growth. Sensors are expected to grow at around 10%, but on a very small base. The analog/mixed signal market will show a small growth at 3%. The big surprise has been the resurgence of demand for digital processors – this market should grow at around 8%.
What are the trends in the end market that are driving this demand mix?
As it is well known, the pandemic has forced a rapid migration to work from home across industries. This in turn has led to increased demand for cloud infrastructure and personal computing devices such as laptops. Demand for greater bandwidth is also ramping up expectations of rapid deployment of 5G – most users in advanced markets can hope to see significant deployments in the next couple of years. The computing and communication verticals have for long been key volume drivers for both digital processors as well as memory. Each of these verticals form roughly 30% of the world semiconductor market, and increased demand in these verticals bodes well for continued growth. The automotive vertical has historically been a much smaller share of the market at just around 11%. This sector is expected to show steady growth over the next several years as the car transforms from an electro-mechanical device into a computing platform. The automotive industry will, of course, drive demand not just for processing capability and memory, but also sensors. Interesting, KPMG predicts that some hyped sectors such as medical devices, robotics, drones and AR/VR will show relatively moderate growth.
This demand picture tells us that the demand supply mismatch in the semiconductor industry is not just a case of supply side constraints. Indeed, some of the problems are caused by robust demand in certain verticals created by long term changes in consumer behavior driven by the pandemic.
What does all this mean for the industry in India?
The tech industry in India has an indirect dependence on silicon for its compute needs. Lead times for both server infrastructure as well as client devices such as laptops are lengthening at a time when the industry is seeing strong demand. The industry will need to take stock of its IT infrastructure and plan ahead to ensure that demand fulfillment is not hampered by challenges on the infra side.
The global semiconductor industry is also a major customer for the Indian tech industry. This in turn means that there is strong demand for semiconductor talent. IESA estimates that the country needs around 500,000 engineers skilled in VLSI, and around half of that are currently available.
Historically, much of the work done in the country has been on digital verification. However the changing industry-wide from digital to analog/mixed signal designs has meant that the industry has had to shift its emphasis over the past five years or so. Training semiconductor talent is a very time-consuming process, and it is not uncommon that the student needs up to 18 months of apprenticeship before s/he is productive. Professional finishing schools are taking up some of this slack, at least in digital verification, but results are patchy and inconsistent.
What the industry needs today is a closer industry academia collaboration, so that academic research is better aligned to industry requirements. We also need stronger system of apprenticeships, so that the industry can groom fresh graduates on specific hands-on know-how needed for executing live projects. The next few years are going to be an exciting and hectic time for everyone in the industry – the Indian tech sector needs to move fast in order to capitalize on the opportunities that present themselves.
