As more things become computerised, and demand for more powerful computing swells, the semiconductor industry is booming and also transforming.
In the past five years, the market cap of listed semiconductor firms has grown fourfold, topping $4 trillion worldwide. With employees working remotely and consumers ravenous for streaming video and gaming, the pandemic has brought a bounty to chipmakers. Mega-deals have come out of this. In October, American multinational AMD announced a $35 billion deal to acquire Xilinx, which makes programmable logic devices (PLDs). The following month Nvidia, another big American firm, said it will purchase British chip designer Arm for $40 billion.
Flush with cash and expecting demand to continue rising, semiconductor firms are also ratcheting up capital expenditures. South Korean giant Samsung said it plans to pump more than $100 billion into its chip business over 10 years. Taiwan Semiconductor Manufacturing Company (TSMC), which produces chips based on blueprints from firms like AMD and Nvidia, adjusted this year’s planned capital spending from $17.2 billion to as much as $28 billion.
The boom in chipmaking is the result of several other trends, unrelated to the pandemic. One is an increase in competition and innovation. Many are eager to meet the demand for more advanced chips used to power artificial intelligence (AI), network and other specialized applications. Thus new chip designs, both from startups and from behemoths such as Amazon, Google and China’s Baidu, are competing with the more established designs of companies like AMD, Nvidia and Intel.
Greater diversification in chip design is not good news for designers of general-purpose chips. Until fairly recently, technology companies used off-the-shelf chips in their devices. Apple, for example, has over the decades sourced the microprocessors in its desktops and laptops from MOS Technology, Motorola, IBM and Intel. This changed after it launched the first iPhone in 2007. The company started designing its own chips, and it used these in subsequent generations. Eventually Apple began using them in its laptops and desktops as well, and in 2020 ended its partnership with Intel. Its earliest supplier, MOS Technology (later known as CSG), has been defunct since 2001.
While chip design is diversifying, manufacturing the silicon chips based on those designs is becoming increasingly concentrated amongst a handful of players. This is due to extremely high production costs. To keep up with each technological advance, companies must keep investing. Few are willing and able. The Economist reports that currently only three firms in the world, Intel, TSMC and Samsung, can make advanced processors. There has also been a geographic shift. According to the Semiconductor Industry Association, 80% of global chipmaking capacity is now based in Asia.
Semiconductor manufacturing requires factories packed with ultra-advanced equipment. A 2011 report from McKinsey estimated the average cost of such a factory at $3 billion to $4 billion. TSMC’s new 3nm factory, completed less than 10 years later in 2020, cost $19.5 billion. Demand is being fuelled by the proliferation of the Internet of Things (IoT) as well as electrification. As this trend continues, the industry could become dominated by even fewer companies.