The outlook for the 2019 semiconductor market is unclear given trade tensions and memory shifts...
SUNNYVALE, Calif. — Amid big uncertainties over China and European trade and the bottom of the memory cycle, veteran analyst Bill McLean predicts that the semiconductor industry will slow down over the next two years before picking up steam again. Long term, don’t worry about China, he advised in his annual Silicon Valley talk.
Specifically, the president of IC Insights forecasts 1.6% revenue growth this year and a 0.9% contraction next year, followed by three years of rising growth in the 7% to 13% range. His views are in the ballpark of analysts from Gartner and VLSI Research, who expressed hopes earlier this month that the industry’s current slowdown will avoid a recession.
IC Insights shares forecasts at a dark moment
The end of a two-year super-cycle in memory will drag average selling prices on chips down 6% overall this year, IC Insights predicts. Hopefully, issues around both the U.S./China trade dispute and the U.K. Brexit will settle down in the spring, he said.
What happens around March deadlines for both Brexit and raising U.S. China tariffs to 25% “will tell a lot about this year,” McLean said, noting that both China and the U.S. have “plenty of incentives” to settle their trade war.
A recent proposal in Congress to ban sales of chips to both China’s Huawei and ZTE “would be an economic Armageddon … I don’t think it will happen, but it may be a negotiating tool,” he said.
The likely extradition of Huawei’s CFO to the U.S. to face charges over sales to Iran adds to wild cards of “what we are overlaying on the world economy,” he said.
These days, macroeconomic changes are a dominant influence on the semiconductor forecast. GDP figures are expected to trend down a few tenths of a percent in almost every country this year with the exception of India, up a tenth of a percent.
The net for a chip industry still working off some excess inventory from the holiday season is that “no one is looking for a good first half, but they all look for a pretty good second half, and if the trade issues fall in place, the second half could look good,” he said.
A slowdown and a minor contraction may be ahead in semis. (Source: IC Insights)
Seeking the bottom of the memory super-cycle
The good news is that unit growth in the $430+ billion chip industry is on an average 11% annual growth trend, up from 6% and 9% in the recent past. The bad news is that the two biggest parts of the market are the most cyclical — $100+ billion in DRAMs and $60+ billion in NAND flash.
“We’ll see this year if the three major DRAM companies can put a lid on capacity expansion to keep a lid on declining prices,” McLean said.
The cycle was so bad last year that Nanya, a small player in Taiwan, watched its monthly DRAM sales rise 45% in the first half of the year, then decline 44% in the second half.
Memory chipmakers spent about $7 billion too much in DRAM capex last year and $10 billion too much in NAND, chasing the growing markets. Overall, chipmakers will cut capex by 13% this year, in part to shore up prices, IC Insights predicts.
Memory leader Samsung, which doubled its capex spending to $24 billion in 2017, is the wild card.
“I have never seen a company as aggressive as this … it was to show China it did not have a chance to catch up,” he said, estimating that the South Korean giant will scale capex back to $18 billion in 2019.
Falling memory prices are dragging down what would otherwise be a decent year in chips. (Source: IC Insights)
A skeptic on China’s rising chip prowess
When it comes to China’s threat as a rising producer of semiconductors, McLean calls himself a skeptic. It may be an understatement.
China’s planners, noted for their optimism, set a goal with the Made in China 2025 program of supplying 40% of the country’s chip needs by 2020 and 70% by 2025. IC Insights estimates that the actuals will be closer to 17% and 24%, respectively.
“I think they may have some success, but I don’t think it’s going to be anything near what they are hoping for,” he said
Fabs in China produced about $23.7 billion in chips last year, about 5.5% of the world’s total. But most of those chips (more than $16 billion worth) came from large DRAM and NAND fabs run by Intel, SK Hynix, and Samsung, IC Insights estimates.
China’s share of the worldwide foundry business stands at 9.2% today. IC Insights expects that it will grow to just 10.3% by 2023.
China clearly has money to invest. However, it has been quiet since May about the next tranche of the estimated $170 billion that it plans to invest in semiconductors, McLean noted.
China’s chip market is large and growing; it’s chip production, not so much. (Source: IC Insights)
IC Insights makes an exception for Yangtze Memory Technology Co. (YMTC), which reportedly got as much as $24 billion in investments to become a leading NAND player. The market watcher projects that YMTC could leap from sales of $300 million last year to $5.4 billion in 2023, making it the third-largest chipmaker in China, surpassing SMIC.
McLean admits the forecast is speculative. He expects that Micron, Samsung, SK Hynix, and Toshiba will sue the company for infringing their NAND patents as soon as it ships competitive chips.
He also casts doubts on whether it will produce leading-edge chips given that it is currently “five years behind with 32-layer 3D NAND” that it is starting to ship today. He laughs at the company’s ambitious plans to skip 96-layer parts and go directly to 128-layer ones.
“Ultimately, they plan to go to DRAM,” he said. “I guess they thought, ‘We might just as well add that, too.’”
McLean is even more dour on Fujian Jin Hua Integrated Circuit Co. (JHICC), which got about $5.6 billion in investments and won a patent dispute with Micron but faces an equipment ban from the U.S. As a result of the equipment ban, UMC has backed away from its role as a technology partner, leaving JHICC “in limbo now,” he said.
Likewise, Innotron is getting an estimated $7.2 billion investment for DRAM but lacks a leading-edge technology partner — or even a working website. “I don’t see how they will make much of a dent,” he said.
Two other projects — Shanghai Huali and a reported $9 billion fab near Shenzhen with support from Foxconn — include plans for 28-nm foundry services. However, McLean said that the 28-nm node is likely to be in oversupply for years and new China foundries will only steal business from existing players such as SMIC.
— Rick Merritt, Silicon Valley Bureau Chief, EE Times.