ADVERTISEMENT

Intel’s Latest Chip Push Suggests the Company Has a Short Memory

Intel’s renewed bet on a volatile business has disappointed analysts who wanted a fresh look from the new CEO.

Intel’s Latest Chip Push Suggests the Company Has a Short Memory
Intel CEO Bob Swan (Photographer: Samyukta Lakshmi/Bloomberg)  

(Bloomberg Businessweek) -- It can be tough to recall that Intel Corp. invented the memory-chip business half a century ago. It gave up on the field for more than a decade, starting in the 1980s, and has struggled since returning in 2006 to match the success of Samsung Electronics Co. in the historically low-margin product category. The bigger issue for Intel, however, is that memory-chip technology hasn’t been advancing as quickly as more profitable gear. The limitations of memory chips are starting to lessen the value customers see in buying, say, Intel’s lucrative server chips, threatening a central profit center.

That’s why Bob Swan, Intel’s chief executive officer since late January, is sticking with his predecessor’s push into a new kind of chip, called Optane, which the company says doesn’t have the weaknesses of existing technology. According to Swan, Optane represents an evolution serious enough to keep pace with leaps that Intel has made in the data-center gear its biggest customers buy. “We think we have something special,” he says.

Analysts are skeptical, given Intel’s checkered history with memory chips. “In what was at one point the best market in history, they were losing money,” says Stacy Rasgon, an analyst at Sanford C. Bernstein. Swan says the ledger will start looking better soon.

There are two basic types of memory chip, each with different advantages. DRAM (dynamic random-access memory) chips read and write data quickly but can’t store it when a system is powered down; NAND (short for “not and”) flash memory chips are basically the opposite in how they function. Intel says Optane chips can permanently store data and read and write it faster than NAND, if not faster than DRAM. The product, which went on sale at the end of last year, will require more investment in manufacturing before the company can mass-produce it. But it’s been tested successfully by Alibaba Group Holdings Ltd. and Google’s cloud division, according to the companies. “There are a lot of operations where you benefit from having all the data accessible to one processor,” says Google product manager Paul Nash. “We think it is going to be commercially viable in a few quarters’ time.” Alibaba used the technology to support its massive Singles Day sales.

Still, memory chips tend to be the most volatile slice of the $470 billion semiconductor industry. Only the Paranoid Survive, a seminal Silicon Valley history written by Intel co-founder Andy Grove, spends a great deal of time arguing that it was smart to get out of memory in the Reagan era. In some ways, the company’s efforts to figure out the market still seem like a work in progress: Intel posted a $5 million operating loss in the NAND business last year, on $4.3 billion in revenue.

Intel as a whole hit a record $71 billion in revenue during that period, but it’s facing new obstacles at a pivotal moment. Delays in the rollout of manufacturing updates have undermined the company’s lead in chipmaking technology for the first time in decades, and there’s growing evidence that some of its biggest, most reliable customers are considering doing business with cheaper competitors or making their own chips. Cloud leader Amazon.com Inc., which uses a staggering number of server chips, in November unveiled a service based on its home-grown Graviton chips, saying the product would come with a “significantly lower cost.”

While he’s an experienced chief financial officer, Swan has been at Intel only since 2016, and during his six months as interim CEO he repeatedly said he didn’t want the top job permanently. Now that he has it, he’s picking up the pursuit of memory from his predecessor, Brian Krzanich, who was ousted in 2018 for a years-earlier affair with a subordinate. Some investors had hoped Swan’s background in finance would mean he’d cut back on riskier bets in favor of more stable businesses.

Intel says data-center operators spend about $15 billion a year on DRAM that would be better spent on Optane chips. And in Swan’s reckoning, the only thing holding back Intel’s memory-chip numbers recently has been all the money it’s been spending to develop Optane. “The difference over the next couple of years,” he says, “is we’ll have returns coming in.” We’ll remember he said that.

To contact the editor responsible for this story: Jeff Muskus at jmuskus@bloomberg.net, Jillian Ward

©2019 Bloomberg L.P.