Why are Intel’s manufacturing costs so high and what does it mean for the future? SemiAccurate can answer both questions with a technical history lesson.
Step back into the misty days of the distant past, a time when creatures like the striding Keller and majestic Koduri roamed the (once) marshes of Santa Clara. It was a dire time when manufacturing was in crisis and possibly faced extinction. In short Intel couldn’t make a chip on a bleeding edge, then 10nm, process to save their lives.
SemiAccurate was the first to expose this in 2016, and oh how the trolls laughed. Then we pointed out the technical problems with the process, and still they laughed. Granted we didn’t get everything right, but we are pretty good about owning up to wrong calls in print rather than silently editing old stories like some others. In any case in 2019 when we said on Twitter (Twitter killed all pre-2020 Tweets so no link, sorry) that we saw a path to Intel failing, they restarted the laughter. By the time the dire situation Intel was in became obvious to even the peanut gallery, they suddenly vanished. Strange that.
Why do we bring this up? The history is important because it is what is causing Intel’s current absurdly high production costs and cratered margins. Before you shout about Intel’s 7nm family, now called i4/i3, and 20a/18a having terrible yields like the clickbait sites regularly do, that isn’t the issue. As far as SemiAccurate can dig up, yields on both process families are just fine, more or less where they were meant to be. Nothing is currently wrong with Intel’s mainstream process families yields, nor is there a big problem with the cost of the process itself.
Since yield is the main factor affecting silicon costs shortly followed by the cost of a wafer on the process, we will ignore profit/gouging for this argument, what is the problem? Intel’s production costs are clearly too high and their ASPs aren’t great. Advanced packaging has a lot to do with it, as does the size of their cores. Not much can be done about core size in the short and mid-term but advanced packaging is a head scratcher.
Why is Intel using such a costly method of construction on low ASP devices? Multi-chip packages can bring lots of benefits to a device but the costs will kill you on smaller and cheaper chips, think server=good, consumer=bad. Look at AMD’s Strix Halo for more on this, but Intel did it anyway and is taking a financial beating for it. Why? There is actually a good reason for them to use expensive packaging at the low end of the market, and it was a good call when it was made.
Note: The following is for professional and student level subscribers.
Disclosures: Charlie Demerjian and Stone Arch Networking Services, Inc. have no consulting relationships, investment relationships, or hold any investment positions with any of the companies mentioned in this report.
Note: The following is analysis for professional level subscribers only.
