Both semiconductors and artificial intelligence have their roots in the 1960s. But it was the chip business that took off first – initially with personal computers and later with smartphones – as AI remained in a research phase. Now the two breakthrough technologies are finding synchronicity, as semiconductor firms likeNvidia ( NVDA ) race to provide the processing power demanded by AI applications.
Makers of semiconductor equipment manufacturing gear such as Applied Materials ( AMAT ) or Lam Research ( LRCX ) could also be among those to get a lift if AI boosts demand for specialized processors as well as computer memory.
AI involves computer algorithms – software programs that aim to mimic the human ability to learn, reason and make decisions. All AI software needs computing power. That’s provided by electronic chips. AI chips enable software to find patterns and make inferences from large quantities of data.
How do makers of semiconductor gear fit in? They sell gear that cleans, smooths, bakes, chemically coats and etches electrical pathways in wafers. Wafers are the thin, round slices of semiconductor material, from which microchips are made.
Both chipmakers and their equipment suppliers stand to gain as AI comes of age. AI-driven systems create demand for processors and memory, much as PCs and smartphones did earlier, analysts say. In addition, AI chips could require specific manufacturing innovations. While the industry trend for decades has been to make chips tinier, some AI chips are actually getting bigger due to dramatic increases in computing complexity.
“There’s no question there’s going to be tremendous economic value created and shifted by AI over the next decade in health care, transportation, entertainment and other areas,” said Gary Dickerson, chief executive of Applied Materials. “And semiconductor technologies are really at the foundation of those inflections.”
“People compare data to oil in powering the Industrial Revolution. But oil sitting in a barrel isn’t worth much,” Dickerson told IBD. “You have to process it to create value. It’s the same thing with data. You have to process it. There’s going to be tremendous innovation in high-performance computing.”
A Growing, $70 Billion Industry
Contract chipmakers such as Taiwan Semiconductor Manufacturing (TSMC) buy roughly one-third of all chipmaking gear. Gartner estimated Semiconductor Capital Spending at $69.9 billion this year, headed to $75.8 billion in 2020. TSMC’s customers include Nvidia, currently the leading provider of “GPU” processors to internet companies investing in AI infrastructure, such as next-generation data centers.
At TSMC’s 30th anniversary shindig in Taiwan on Oct. 23, Apple ( AAPL ) Chief Operating Officer Jeff Williams told attendees that AI “will change the world” as software is built into an increasing number of consumer devices.
Apple’s new iPhone X features the machine-learning optimized A11 bionic chip. Apple’s iPhones account for over 5% of (DRAM) memory chips consumed worldwide. IPhones will need even more memory as AI apps are developed.
At the 30th anniversary gathering, TSMC Chairman Morris Chang contended that AI will replace many doctors in 10 to 20 years.
TSMC and other contract manufacturers – GlobalFoundries, China’s Semiconductor Manufacturing International, and United Microelectronics ( UMC ) – are among the biggest customers of gear makers. Some chipmakers – Samsung, Micron Technology (MU), Intel (INTC), Texas Instruments(TXN), and STMicroelectronics (STM) among them – still own their own factories, or fabrication facilities.
Others, including Nvidia, Broadcom (AVGO) and Qualcomm (QCOM), are “fabless” chipmakers, outsourcing their production needs to contract manufacturers.
The Electronics Semiconductor Equipment group ranked a strong No. 10 on Thursday out of the 197 industries tracked by IBD. The highest-ranked companies in the group include Applied Materials, Lam Research, MKS Instruments (MKSI), Teradyne (TER), Ultra Clean Holdings (UCTT) and Ichor Holdings (ICHR).
Electro Scientific Industries (ESIO) spiked 30% on Thursday, after reporting a surprise 239% EPS gain and 139% rise in revenue in its fiscal second quarter. That was its second quarter of triple-digit earnings growth, following three quarters of steep declines.
New, More Powerful Processors
The chip-equipment group has been cyclical, but that could be changing. When economies are strong, foundries and chipmakers order more gear so that they can supply the tides of chips being built into consumer electronics, telecom and automotive products. Personal computers have given way to smartphones as the big driver of chip demand over the past decade.
Another big market is that of the so-called “hyperscale” data centers, packed with computer servers and leading-edge networking and data-storage infrastructure rising to serve the cloud-computing units of Facebook (FB), Amazon.com (AMZN) and Alphabet ‘s (GOOGL) Google. The next-generation data centers have increased demand for solid-state “NAND” memory chips built into computer servers and data storage systems.
As AI software is deployed in such hyperscale data centers, Wall Street analysts expect memory-chip demand to increase. That’s good news for Lam Research, which sells memory chip manufacturing gear to Samsung and others, said Farhan Ahmad, a Credit Suisse analyst in a report. Ahmad says memory built into computer servers is now a bigger market than PCs or smartphones.
“As AI grows it should benefit Lam Research,” he wrote. The Credit Suisse analyst added that as AI drives demand for more powerful processors, the Dutch maker of lithography gear, ASML Holding(ASML), could get a lift.
UBS analyst Bill Lu forecasts that the AI chip market will boom to $35 billion by 2021, up from roughly $6 billion in 2016, as AI chips are built into hyperscale servers, self-driving cars, robotics, drones, smartphones, and web-connected devices (called the Internet of Things) such as security cameras.
Lu expects Nvidia, Broadcom, Intel, Samsung and contract-manufacturer TSMC to benefit.
“The democratization of AI now underway could drive a new phase of growth in semiconductors,” Lu said in a report.
Analysts expect AI analytical tools to spread into finance, health care, energy, advertising, cybersecurity and other areas.
Not all semiconductor analysts, though, are on the AI bandwagon. Joseph Moore, a Morgan Stanley analyst, attended Applied Material’s presentation on the future of computing and AI on Sept. 21.
“We remain excited for AI, and think that it’s positive for growth – but we don’t see it changing the industry cycles meaningfully,” said Moore in a report. “New technologies always contribute to growth – the last seven years saw the rise of the smartphone and the invention and expansion of the cloud business – and the wafer fab equipment (WFE) market barely grew until 2017. AI growth will be offset by other declines, and we don’t see an inflection in overall industry growth because of it.”
Applied Materials’ Dickerson disagrees. He says AI will be another factor in decreasing the “volatility” of the chip equipment making industry. During the PC era, equipment spending was often characterized by periods of rapid growth followed by sharp declines as chipmakers timed their purchases to match PC-refresh cycles. Major upcycles included 1987-1990, 1993-1996, 1999-2000, 2003-2007 and 2010-2011.
The “boom-bust” cycles started to change after 2010, when smartphones and other mobile products provided additional demand for semiconductors and equipment. Dickerson says AI – along with cloud computing and the Internet of Things – will make the equipment industry less cyclical.
He says some analysts may underestimate the impact of AI by extrapolating too much from the past. “Some people can’t see nonlinear inflections” that create new paradigms, he added.
AI Drives A New Innovation Era
While AI deployment is still in the early stages, makers of chip manufacturing gear have seen orders boom in 2016 and this year. The market for semiconductor manufacturing equipment will expand 26% to $46.6 billion in 2017, says research firm Gartner. The market rose 11% to $37.4 billion in 2016, says Gartner. The top five gear suppliers are Applied Materials, Lam Research, ASML, Tokyo Electron andKLA-Tencor (KLAC).
A year ago, U.S. regulators blocked Lam Research’s acquisition of KLA-Tencor on antitrust grounds. Lam is the largest maker of machines that deposit or etch away materials on the silicon. KLA-Tencor’s machines measure and inspect circuitry on chips.
Management at Applied Materials, Lam Research, and KLA-Tencor all indicated on September-quarter earnings calls that 2018 will be another strong year. One reason is that memory chipmakers Samsung, Micron and SK Hynix continue to ramp up capacity.
“We are entering a new innovation era,” said Goldman Sachs in a report on Lam Research’s Q3 earnings. “The last decade saw the transition from a PC-centric to a mobile-centric world. We expect the next decade to be defined by a transition to an AI-centric world.” Lam Research stock has shot up 95% in 2017.
Shares in Fremont, Calif.-based Ichor surged to an all-time high in October. Ichor makes gas- and chemical-delivery subsystems for chipmaking equipment. Its customers include Applied Materials and Lam Research.
The buildout of next-generation data centers for cloud computing has provided a boost to chipmakers, contract manufacturers and gear makers. Another plus has been the shift to “3D” NAND memory devices. Equipment makers could get a lift by 2019 from China’s buildout of new memory chip factories, analysts say.