In 2021 graphics cards were hot stuff. Video-game devotees and cryptocurrency miners queued overnight to get their hands on the latest high-end offering from Nvidia or amd, two American chipmakers. And graphics processors were far from the only sizzling semiconductors. An acute shortage of chips disrupted the production of everything from smartphones to cars and missiles, just as demand for all manner of silicon-bearing devices boomed. Last year the chip industry’s revenues grew by a quarter to $580bn, according to idc, a research firm. Chipmakers’ market values soared. tsmc, a giant Taiwanese contract manufacturer, became the world’s tenth-most-valuable company.
With demand expected to grow ever more insatiable, the time-honoured semiconductor cycle—the consequence of the lag between demand and new supply, which takes a year or two to build up—appeared to be a thing of the past, prompting chip firms to spend like there was no tomorrow. tsmc and its two main rivals, America’s Intel and Samsung of South Korea, invested $92bn between them last year, a rise of 73% relative to 2019—and pledged a further $210bn or so all told over the next two years.
Now it seems that far from being banished, the chip cycle may instead have sped up. Chips of all sorts are looking wobbly. This month Samsung said that operating profit would stall this quarter, following three quarters of record-breaking sales figures. It is reportedly considering dropping prices for memory chips in the second half of 2022. In June Micron Technology, an American memory-chip maker, forecast sales of $7.2bn in the third quarter, a fifth lower than expected. TrendForce, a research firm, expects memory prices to fall by a tenth in the next three months. By one estimate, prices of graphics chips have dropped by half since January, as the cryptosphere implodes and gamers spend more time in non-virtual reality. In the euphemistic words of David Zinsner, the chief financial officer of Intel, America’s chip titan, the rest of the year is looking “a lot noisier than it was even a month ago”.
As the turbocharged boom risks turning into a supersized bust, the share prices of the world’s chipmakers have slumped by about a third this year (see chart), half as much again as the s&p 500 index of big American firms. Added to that, geopolitical tensions risk splitting up a global market and shattering complex supply chains. The pandemic’s superstar industry suddenly appears a lot less stellar.
Start with supply. One way that firms have been adding capacity is by installing new kit in existing fabs (as chip factories are known). In the second half of 2021 global spending on equipment to etch chips onto silicon wafers jumped by about 75% compared with pre-covid levels, estimates Malcolm Penn of Future Horizons, a research firm. Given that it takes about a year for such investments to translate into new semiconductors, late 2022 could see a production glut.
Another way to increase capacity is by building new fabs, which can take a couple of years. According to semi, another research group, 34 of these came online worldwide in 2020 and 2021. Another 58 are scheduled to begin production between 2022 and 2024. That would raise global capacity by roughly 40%. Intel has six fabs in the works, including a $20bn leading-edge “megafab” in Ohio, and factories in Arizona and Magdeburg, Germany. Samsung’s investment plans include a large modern fab in Texas. tsmc is building a similar one in Arizona. Most of these are expected to begin producing chips by 2025.
There was always the risk that by the time some of this fresh supply materialised demand might have faded. But the hunger for chips appears to have waned faster than expected. The clearest signs are in the market for personal computers (pcs), which account for about 30% of overall demand for chips of all varieties. Having received a boost from the pandemic as working and schooling from home became the norm, global pc shipments are poised to fall by 8% this year, according to idc. That is partly because some of those pandemic purchases had simply been pulled forward. Sales of smartphones, another 20% of demand, are expected to ebb, too. In April smartphone shipments in China, the world’s biggest market, were a third lower than in the same month last year. The slowdown in pc and phone sales will be sharper still if the world economy dips into recession.
Data centres and carmaking consume around one-tenth of the world’s chips apiece. Demand is not forecast to fall this year. But signs of softness can be seen. Chinese orders for server chips, which power data centres, have dropped off. Many panicked carmakers, for their part, have double- or triple-ordered chips to avoid the sort of shortages that forced them to cut output last year. Stacy Rasgon of Bernstein, a broker, points out that in the past few quarters shipments of automotive chips have been about 40% higher than what you would expect based on the number of cars shipped and the typical number of chips in an average car. Big semiconductor stockpiles in the car industry may mean a sudden drought of new orders.
The downward pressure on prices may be compounded by another powerful force. Political considerations, both domestic and international, increasingly influence semiconductor supply and demand. On the supply side, last year’s chip crunch spooked governments around the world and reminded those in the West that 75% of all semiconductors are produced in Asia. Many now want to bring the manufacture, especially of leading-edge chips deemed of strategic importance, within their borders. In America, Congress is wrangling over the chips Act, which, if enacted, would hand the industry up to $52bn over five years in subsidies and research-and-development grants. The eu’s version offers over €43bn ($44bn) until 2030. India, Japan and South Korea have similar schemes. China, which launched a semiconductor policy in 2014, has long subsidised the industry.
All this state largesse could lead to even more overcapacity. At the same time, greater interventionism may further dent the outlook. For one thing, a chip industry fragmented along national borders would risk wasteful duplication, driving up costs for consumers. A report by bcg, a consultancy, and the Semiconductor Industry Association, a lobby group, finds that in a scenario where semiconductor production is self-sufficient within regions chip prices would increase by between 35% and 65%.
America’s government seems intent on constricting demand in another way. It is using export controls to deny Chinese buyers access to semiconductors and the tools needed to make them. The urge is understandable: China is an increasingly authoritarian challenger to the American-led, rules-based global order. More problematically for the semiconductor industry, China is also the world’s biggest chip market.
tsmc and Intel have already lost Chinese customers as a result of American trade restrictions. Others, such as Qualcomm, note in their annual reports that Chinese clients are developing their own chips or switching to local suppliers, in part because of geopolitical tensions. American chipmakers warn that their large research-and-development budgets would be difficult to sustain if they lost Chinese custom.
Political considerations are a headache for other companies in the semiconductor value chain, too. On July 5th Bloomberg reported that asml, the Dutch monopolist in the market for the $100m lithography machines used to etch high-end chips, was under pressure from the American government to stop selling its gear to Chinese firms. China accounts for 15% of asml’s sales; its share price fell by 7% on the news. The market values of asml’s American suppliers, such as Azenta and mks Instruments, also slid. China is even more important for other American toolmakers. Applied Materials, kla and Lam Research derive a third of their revenue from Chinese customers. All are in talks with America’s government to limit their sale of high-tech tools to China.
The chip bust may be softened if the drive for greater silicon self-reliance were to sputter. That is not out of the question. Continuous subsidies may be needed to keep American fabs at the cutting edge, for example. That, in turn, would require sustained interest from easily distracted policymakers. In late June Intel said it would push back the opening of its new fabs in Ohio, blaming delays in passing the chips Act. tsmc has said it may have to slow the construction of its Arizona fab for the same reason. In April Morris Chang, former chairman of tsmc, bluntly called America’s attempt to reshore chip production an “exercise in futility”, pointing out the country’s high costs and the lack of engineering expertise.
Indeed, take out the government intervention and dips in chip cycles have been getting shallower of late, notes Ajit Manocha, who heads semi. That may be in part because the industry has become more consolidated. In the 1980s the market for memory chips had 20-odd firms jostling for custom. Today it is dominated by just three: Micron, Samsung, and sk Hynix. The situation is as stark at the cutting edge of microprocessor-making, where Intel, Samsung and tsmc are the only firms capable of churning out the most advanced kit, down from nearly 30 firms in 2001. Fewer firms control a greater share of capital expenditure and can rein it back in if supply outstrips demand. This will require the chipmakers to rediscover capital discipline—something they have not had to exercise in a while. ■