Intel Acknowledges It Was ‘Too Aggressive’ With Its 10nm Plans
When Intel finally ships 10nm for laptops later this year, it’ll hopefully close the book on a uniquely difficult period in its own history. Originally, 10nm was intended to ship by 2016, following delays to the company’s 14nm process. Intel has since been forced to push the date back multiple times, though the company has repeatedly stated that it will not miss its own promise to have Ice Lake on store shelves by the holiday season of 2019.
According to Intel CEO Bob Swan, that delay was caused by the company’s overly aggressive strategy for moving to its next node. The delay, Swan said at Fortune’s Brainstorm Tech conference, was “somewhat a function of what we’ve been able to do in the past, which in essence was defying the odds. At a time when it was getting harder and harder, we set a more and more aggressive goal. From that, it just took us longer.”
According to Swan, Intel is prepared to put its problems behind it. “The short story is we learned from it,” Swan said. Intel has reportedly reorganized itself for greater transparency and better information exchange between business units to avoid these problems in the future.
This explanation fits with what is publicly known about the situation. Back in 2017, Intel unveiled its 10nm node with promises of aggressive scaling, higher performance, and leading-edge density improvements. Features like contact over active gate were declared to be core components of how Intel would take a leadership position in the semiconductor industry. Intel’s 2017 slideshows on the topic are shown below:
Even in 2017, it seemed like a risky bet. Intel’s 2017 argument, essentially, was that it had taken the extra delay on 10nm and turned it into a win by delivering a more aggressive node shrink. But there was always a risk of biting off more than one could chew and that kind of mistake can harm companies for a very long time.
At Architecture Day last December, Intel shared details of how the last four years have logjammed development processes. Intel’s entire design and development strategy hooked architectural improvements to new nodes. Once the company realized it wasn’t going to hit its initial 10nm projections back in 2015, it would have had to make a decision: Did it try to de-couple process node and architecture to pull 10nm design changes back into 14nm, or did it keep going and bet on 10nm being available in the near-term future?
What we’ve heard from various sources is that Intel has pulled certain features originally intended for 10nm CPUs into 14nm parts, but that this linkage between new nodes and planned architectural improvements hit the firm hard. It’s possible that Intel made a decision to stay the course in 2015, believing that 10nm would be available by 2017. This obviously didn’t happen, but it’s possible that by the time Intel knew it wouldn’t, it didn’t make sense to try and pull 10nm chips into 14nm process nodes. The company might well have finished the work right around the time 10nm finally becomes available.
Putting the Miss in Context
There’s no arguing the fact that Intel missed, and missed badly, on 10nm. It’s unquestionably going to impact the company’s finances and competitive position against AMD, which is enjoying a resurgence of its own. But when you step back and take a look at the entire semiconductor market across multiple companies, these kinds of things happen. It isn’t always driven by foundry issues, but when you toss in the possibility for designs to under-deliver, the metaphorical board lights up with examples.
Samsung’s first implementations of big.Little were broken and the resulting SoC’s had high power consumption. Qualcomm’s Snapdragon 810 was viewed as a disappointment, though the company argued that some of this was the result of its own partners failing to implement power management properly. AMD’s Bulldozer architecture nearly killed the company. Jump back in time a little farther, and you’ve got the poorly received ATI Radeon HD 2000 family (developed mostly before the AMD acquisition). Nvidia, of course, had the infamous GeForce FX. Intel had Prescott. The point is, most major semiconductor companies have weathered significant silicon misses. This is something that happens to even the best-run companies.
We’ve known for years that each new node brought greater challenges and tougher problems than the previous; that’s why fewer and fewer foundries bother operating on the leading edge at all. Intel might not have been the company anyone expected to slip, but the chances that somebody would have been rising every node.
Intel is going to face a more competitive AMD and a tougher environment for its own products than it would have had it kept to its original plans. That’s a given. If it can’t bring 7nm to market in 2021 as is currently planned, the situation will get harder. We don’t even know when or if desktop CPUs will make the jump to 10nm, and they might avoid that node entirely until 10nm+ or 7nm are available. But Intel’s fundamental position across its markets is strong. It may lose a few points of market share to AMD it probably will but it isn’t going anywhere.