Will a steep ramp in global IC manufacturing capacity result in a glut?
When Mark Liu, chairman of Taiwan Semiconductor Manufacturing Co. (TSMC), suggested to CBS journalist Leslie Stahl that the global semiconductor shortage may extend through 2022, Liu’s comment raised questions about the current steep ramp of new capacity to meet demand, the potential for rising chip prices and eventual oversupply.
Chip shortages and possibly over-exuberant capacity expansion come as planners in China and the U.S., recognizing that semiconductors are critical for economic growth, have launched projects to build local, integrated semiconductor supply chains.
The end of shortages in 2023 may be followed by overcapacity in 2024, according to worst-case scenarios offered by industry analysts interviewed by EE Times.
“It depends on how many gigafabs get built,” said Dan Hutcheson, CEO of VLSI Research, who predicts overcapacity by 2024, noting that governments tend to be heedless of the profit-maximization mantra that has driven the semiconductor industry in recent years.
“When you are government-funded, profit doesn’t matter to you. Your job is to gain market share. That will be the problem, probably with China,” Hutcheson added. There’s also the tendency for politics rather than market dynamics to drive decisions on capacity expansion. That’s what causes the gluts, Hutcheson said, noting a string of boom-bust cycles going back decades.
Demand for leading-edge chips at 7-nm nodes and below will remain tight over the next couple of years, according to Brett Simpson, senior analyst with Arete Research.
The main factors behind shortages are the migration of 4G (12- and 16-nm chips) to 5G (7nm and below). High performance computing will grow exponentially, and the automotivemarket will, for the first time, use leading-edge silicon, according to Simpson.In the short term, companies using older technology to make chips on 8-inch wafers for customers such as auto makers will be hard pressed to meet demand.
“For the lagging edge, there is clearly a lack of equipment — and it’s super expensive to add capacity here — to solve this capacity shortage,” Simpson said. “This gets largely resolved as migrations from 8-inch to 12-inch get finalized.”
Sometime in the first half of 2022, full production should be underway for “migrated” chips such as power management ICs, display drivers and contact image sensors. They will ramp into mass production on 12-inch lines, Simpson said.
There is no foreseeable threat from oversupply, according to Simpson.
“It depends on the extent of subsidies. We won’t see a meaningful U.S. fab footprint until the middle of the decade. China is unlikely to build extreme ultraviolet fabs. So, for leading edge, we do not see this as a major risk. For nodes like 12 and 16 nanometer and 22 to 28nanometer, there could be some long-term risk of oversupply as this is where China can add capacity,” as will GlobalFoundries and United Microelectronics Corp.
Substantial investment in new capacity creates risk, particularly with those fabs starting production in the 2023-24 timeframe, according to Wedbush Securities Senior Vice President Matt Bryson.
“I also think that semiconductor content is slated to increase, a result that could temper the length of the inevitable downcycle.”
For commodity parts, oversupply is possible, but less concerning, according to Simpson. “While demand is hard to predict, the growth in silicon content across key end-markets is a strong growth trend for the industry.”
Product replacements such as electric vehicles for internal combustion engines, accelerated processors taking the place of general-purpose CPU servers and 5G overtaking 4G are all factors behind an increase in silicon content. For the remainder of this decade,semiconductor growth will be higher than the previous decade, Simpson predicted.
For the time being, shortages have afforded chipmakers an opportunity to raise prices.
“This is uncharted territory for some foundries, and we see a broad trend by foundries to secure long-term contracts with strategic customers at premium prices,” said Simpson. Smaller chipmakers will face significant inflationary challenges. TSMC has the strongest bargaining position, but they are the least aggressive in driving up prices, particularly at the leading edge, according to Simpson.
For some lagging-edge nodes, industry analysts see wafer prices rising 20 percent over the same period last year. For example, at 28 nm, many foundry customers are paying $2,500 or more per wafer. Quotes for 8-inch wafers run as high as $600 each, a price that is expected to remain into 2022.
Chipmakers are applying the leverage provided by IC shortages to secure more favorable pricing on mature and mainstream parts, according to Bryson, who also expects higher pricing to last through 2022.
Still, there’s a difference between possessing pricing power by obtaining a higher value per wafer and pricing power that comes from the chance to jack up prices in the short term, Hutcheson notes. Some foundries are boosting prices with small customers, he added.
“If…I want to buy 100,000 8-inch wafers or 5,000 8-inch wafers, I’m going to pay a huge premium. But if I’m Apple, I’m going to get the price that I negotiated, and the Apples of the world negotiate long-term wafer prices. If there’s a glut, they usually renegotiate those prices, but the prices usually go down. They don’t go up.”
Expanding the supply on 8-inch wafers will likely be more difficult because of production tool sourcing difficulties, according to Bryson. He also foresees “elongation” in the parts supply from 7-nm fabs.
Automakers like BMW and Volkswagen will probably need to change their chip inventory management from “just-in-time” to “just-in-case,” according to Hutcheson.
“The auto industry is thinking on fractions of a day of inventory, and it takes three to four months to get a wafer through when you run the current utilization rates this year in the 96- to 98-percent level.”
That high utilization rate means cycle times have increased to four to four-and-a-half months. Companies that urgently need chips simply have to stand in line.
“If I need to build a car, or I need to build a missile or a fighter jet or an aircraft, if I’m missing one chip, I can’t finish it,” Hutcheson said.
Automakers who have idled production due to chip shortages understand that for every dollar they’ve saved by keeping IC inventories low, they are now bleeding hundreds of dollars in lost car sales.
Beyond the automotive sector, chip shortages are “pretty much over”, Hutcheson said. Based on a weekly metric that tracks shortages and gluts, he added: “In the second quarter, we moved from shortage to tight conditions. In autos, we are still in a shortage.
“But all the other areas are in balance. When we look at DRAM, NAND, logic, analog and power, they’re all either tight or still in balance. But they’ve been loosening up quite a bit.”
This article was originally published on EE Times.
Alan Patterson has worked as an electronics journalist in Asia for most of his career. In addition to EE Times, he has been a reporter and an editor for Bloomberg News and Dow Jones Newswires. He has lived for more than 30 years in Hong Kong and Taipei and has covered tech companies in the greater China region during that time.