What will the semiconductor industry look like in 2024?
That’s the question Pete Rodriguez, CEO of semiconductor startup incubator Silicon Catalyst, asked a panel—and a roomful—of industry veterans earlier this month. And few were shy about predicting dramatic and, for some companies, potentially catastrophic changes.
Memory is in motion
Their first warning went to the companies currently cranking out semiconductor memories, feeding the seemingly insatiable demand for solid-state storage. China has already been working hard to build DRAM factories, they indicated, and a trade war with the United States will likely push that effort into a higher gear.
Said Cliff Hirsch, publisher of Semiconductor Times: “I think the real question is what do Micron, Hynix, Samsung, Toshiba, and Western Digital do when their fabs are sitting idle because China has taken over the memory business.”
“Years ago, we [in the U.S.] were building DRAMs,” said Jim Hogan, managing partner of Vista Ventures, “and we got killed by the Japanese. Then the Koreans came in.
“It’s not going to happen immediately, but if you are the Chinese government, and you put all this money into DRAM, you aren’t going to buy DRAM from anybody else,” said Hogan. “That won’t be good for Koreans or Micron or anybody.” He predicts that “this is going to have a huge political impact.”
“Is there a value proposition in cranking out DRAM?” asked Hirsch. “Sure, Micron is getting 60 percent margins today, which is unbelievable, but what about in a few years when they are getting 10 percent margins? Maybe it’s time to think of them as the textile industry of the future, and they should be going offshore.”
The last new fab on earth…or at least in the U.S.
After conceding the future of DRAM to China, the group took a look at the process innovations that have driven node sizes down to 7 nanometers and kept the companies that developed them in the game. That path, they indicated, is coming to an end.[shortcode ieee-pullquote quote=""What does it cost to build a fab? Fourteen billion. How much cash does Intel have now? Fourteen billion. So they have just one more shot at this."" float="left" expand=1]
Today, said Silicon Catalyst’s Rodriguez, only “three companies are doing 7 nm or below: TSMC, Samsung, and Intel [which is] a distant third. Only three companies in the world can afford it; working at the leading node is very expensive.”
Intel, suggested Hogan, might not be in the game for much longer. “What does it cost to build a fab?” he asked. “Fourteen billion. How much cash does Intel have now? Fourteen billion. So they have just one more shot at this.”
That’s bad news for the industry, he indicated. “Intel has paid for all the innovation in process equipment since the 80s,” Hogan said. “So if Intel stops doing new fabs, where is the innovation going to come from? The Chinese are a generation out. The Koreans are going to get killed with the DRAM thing, so it won’t be Samsung. I don’t have the answer.”
IoT sucks/IoT will save us
So if memories all move to China and process innovation is too expensive, what path can the U.S. semiconductor industry take? There’s lots of hype about the Internet of Things right now, and maybe there’s some hope for the industry there. Maybe.[shortcode ieee-pullquote quote=""I think IoT sucks. You have these stupid tiny chips, that cost a buck or so, going into a smart building (and they) can last 30 years."" float="right" expand=1]
“No IoT companies have been very successful to date,” said Helen Li, managing partner, Needham & Company. “It’s a very fragmented market; there isn’t enough volume, but that could change. More applications are coming out of IoT, particularly industrial ones. That’s one of the sectors in which people forecast highest growth in next 5 years.”
“I hate IoT,” said Hirsch. “I think IoT sucks. You have these stupid tiny chips, that cost a buck or so, going into a smart building [and they] can last 30 years. One of the things [that drives] the semiconductor industry is turnover. Why do I want to sell a chip that lasts 30 years and costs practically nothing? What I want is to sell a disposable chip that goes in the garbage almost as soon as I sell it. We want to make chips that have value along with a turnover rate that lets us keep pushing out silicon.”
Hirsch says the answer might be in making devices that aren’t generally thought of as traditional semiconductor industry products, like MEMS, microfluidic chips, and novel displays. Hogan is betting on products that move intelligence to the edge, particularly on chips that process both analog and digital signals. What’s clear in any case, he said, is that “we are going to have to design-innovate our way out of this, using the processes that we have. It is going to be tough to process-innovate our way out of this.”
Young people wanted
In any discussion about the future, concern about the future workforce typically comes up. The panelists expressed concern about the drop-off in women entering computer science programs, and the challenge to get students interested in STEM in general. The semiconductor industry has an even tougher workforce challenge than the more software-oriented companies, they pointed out. Those companies are struggling to offer jobs that seem as interesting, and to present financial incentives sweet enough to let them compete for talent.
“Maybe the semiconductor industry has to change to be as attractive as the software industry,” Li said. “My son, who is 13, recently had a tour of Google. He was very impressed; he now thinks Google is the best company in the world. Semiconductor companies don’t give the same notion.”
Said Hogan: “We were never that cool.”
Tekla S. Perry is a senior editor at IEEE Spectrum. Based in Palo Alto, Calif., she's been covering the people, companies, and technology that make Silicon Valley a special place for more than 40 years. An IEEE member, she holds a bachelor's degree in journalism from Michigan State University.