Shares of Intel Corp. dropped more than 16% the day after company executives admitted they would be six- to 12-months late delivering a processor that would compete directly with those already shipping from Advanced Micro Devices Inc. and other rivals.
The problem is not that it is taking longer than expected to produce chips built on a 7-nanometer process, analysts said. Rather, the kind of problems Intel CEO Bob Swan described during the company's earnings call July 23 sound too much like the difficulties the company had in delivering its last generation of chips, which are still not up to promised levels of performance, according to Linley Gwennap, president and principal analyst at the Linley Group.
The technical difficulties are understandable with a product as complex as a multibillion-transistor processor built at a new, smaller process size, but the same was true during the effort to deliver processors built on a 10-nanometer (nm) process, which took more than twice as long as expected and is only ramping up to full sales volume this year, according to Matt Bryson, senior vice president at Wedbush Securities.
"No one really expected this, in part because it looked as if Intel was finally getting past the problems it had with 10-nm, so there was more optimism that they'd gotten past all the problems that plagued them," Bryson said.
Even if the problem is short-lived, analysts say rival AMD will have another year or two to make the most of its current technological lead, a position it gained by having its chips fabricated using the 7-nm process developed by contract chip foundry Taiwan Semiconductor Manufacturing Co. Ltd. while Intel struggled with its own manufacturing process.
"There are OEMs and customers more loyal to Intel than they might be because of marketing incentives and discounts and such, but at some point, you begin to ask if you should diversify your risk profile because you don't know what to expect from Intel and don't want to be in the position of not having the right product to sell, or not having the best technology available," Bryson said. "Even if it's only a six-month delay, you have potentially lost a lot of customer confidence."
The shocking part of the announcement was the news that Intel would consider handing over an unnamed part of its production to an external foundry, but that is ground Intel has walked before, with products from MobilEye and other companies that used external foundries to make their chips before they were acquired, though not with products it built from scratch, according to Mario Morales, program vice president, Semiconductors and Enabling Technologies at IDC.
"We're getting close to the point that 20% of Intel demand is being served outside of Intel's capacity, but not necessarily for leading-edge products," Morales said.
But going outside for production expertise is also not a fast solution, he said.
"Even if they decide to use an external foundry, it takes time to take a complex design like Intel's and port it over to another foundry, so that's definitely not something easy to do. Having the kind of volume Intel would provide would mean a shuffle of a lot of the customer base that are all essentially competing for capacity on a single process node, so there would be an impact up and down the supply chain," he said.
Making the decision, implementing it and getting a 7-nm process to work with an external foundry is likely to take two to three years, given Intel's current roadmap, which would create the kind of delays that both customers and investors hoped they were done with when Intel appeared to have fixed the production problems that delayed its move to 10-nm.
"There was one problem and then another one and the expectation was that 10-nm was an anomaly and Intel had gotten beyond that and progress with 7-nm would be fine," Morales said. "Intel did try to mask the news under some very strong earnings for the second quarter to hopefully diminish the blow, but people caught on to that very quickly because it sounds very much like what happened with 10-nm."