Sunday, October 1, 2017

"The Natural Evolution of Artificial Intelligence"

From Barron's Tech Trader, Sept. 24:
As everyday items get “smart,” the technology around artificial intelligence gets more real. The rising stars in this drama include Xilinx, Synopsys, and Cadence Design.

One thing that often fuels technological innovation is the ton of money that gets thrown around.

Case in point: Japanese conglomerate SoftBank Group’s purchase last year of ARM Holdings, a British computer-chip company, for $32 billion in cash. That deal, which came at an astounding premium of 43% above ARM’s value at the time, has no doubt fueled the animal spirits of tech to dream up more companies that can be taken out.

Masayoshi Son, the 60-year-old chairman of SoftBank (ticker: 9984.Japan) who delights in stirring up a commotion, joked at the time that he spends like crazy because he “still feels young.” This past February, Son said, in no way jokingly, that the real reason he’s in such a hurry to spend is that 2018 is the year artificial intelligence surpasses human intelligence, and he wants to control the technology when the machines take over.

Whether from youthful exuberance, or fears of mortality, Son and others like him will most likely continue to fund anything that smells of artificial intelligence, the tech topic du jour. The market is already seeing waves of AI companies, and the chip industry is at the epicenter.

Nearly two years ago, Barron’s wrote that the rise of cloud computing—and with it, aspects of artificial intelligence such as “machine learning”—would spark a wholesale shift in the nature of computer chips (see “Watch Out Intel, Here Comes Facebook,” Oct. 31, 2015).

SOME THINGS WE PREDICTED have played out. Mobileye, a company we featured, got bought by Intel (INTC) this year for $13.7 billion. Two other chip makers we pointed to as potential winners, Nvidia (NVDA) and Advanced Micro Devices (AMD), have seen their shares soar since the article, while shares of incumbent chip giant Intel have lagged, as we predicted.

The market has come to appreciate that Nvidia’s sales of so-called graphics processing chips, or GPUs, are central to the expansion of AI computing by Alphabet’s (GOOGL) Google, Amazon.com (AMZN), and many other giants.

Advanced Micro Devices took on new relevance last week with the rumor that it may be selling chips to Tesla (TSLA) to help with autonomous-car functions. Tesla is already a customer of Nvidia’s. Whether true or not, the speculation stirred some positive words on Wall Street about AMD’s AI potential.

One topic we didn’t pay as much attention to, but that’s starting to get some AI-shine, concerns programmable chips sold by Xilinx (XLNX). Xilinx competes with Intel, which two years ago bought Xilinx’s main competitor, Altera, for $16.7 billion. The third name sometimes lumped in with Xilinx and Altera is Lattice Semiconductor (LSCC), whose sale to a Chinese investment firm was blocked last week by the administration of Donald Trump on grounds of national security.

Xilinx’s parts, called “field-programmable gate arrays,” or FPGAs, aren’t yet synonymous with AI like those of Nvidia. But it has been selling more and more parts to Amazon for its Amazon Web Services. When chips start getting used in cloud computing, it’s predictable they soon get the imprimatur of AI as a side effect.

Xilinx shares have trailed the market this year, rising just 14.8% compared to the 19.4% rise in the Nasdaq Composite Index. That’s despite the fact that its profile as a cloud supplier has grown, and despite its being an oft-mentioned acquisition target. Its shares trade at 25 times next year’s projected earnings per share. That is reasonable for an earnings growth rate of just 10.5%, and cheap compared to the multiple of almost 37 times ARM Holdings got from SoftBank.

Other companies that could get swept up in the fervor include Synopsys (SNPS), a company we profiled in that 2015 story, and Cadence Design Systems (CDNS). The two make software used to design chips. In a world of chips customized for artificial-intelligence tasks, it makes sense that the tools to make those chips will play an important role.

Synopsys founder and CEO Aart de Geus, one of the industry’s deep thinkers, told Barron’s that the next big thing that will propel chip development is when all objects become “smart.” He considers that a better term than AI, given all the misconceptions and stigmas attached to AI.

Synopsys and Cadence may have an opportunity to help with the design of all those custom chips that will increasingly vie with the chips and other parts made by Nvidia, AMD, and Xilinx....MORE