A few days back, Intel CEO, Bob Swan tweeted that the company’s foundry business had doubled its 10nm and 14nm capacity over the last three years. Considering that 10nm is still at least a few months away from mass deployment, it’d be fair to assume that we’re mostly talking about 14nm wafers here.
You can watch the video above, but if we’re being honest, there’s nothing interesting or practical in it. It’s mostly just marketing hogwash. The fact that Intel was able to double its foundry capacity over the last three years is impressive, but despite doubling output, the company’s revenue increased by only 13.5%.
The reason being that yields haven’t improved as much as expected and demand has outpaced the supply. This also led to the 14nm shortages last year and the introduction of the F and KF series chips to salvage as many wafers as possible. You can read more about that here:
The gist of it is that over the last few generations, Intel had to significantly increase the die size of its CPUs in order to increase the core counts and keep up with AMD, but as the die size grew, the density (process node) remained the same. As the die size grew, the number of chips produced from a single yield decreased. Basically, as the wafer size or density didn’t increase, the number of CPUs obtained from each wafer decreased significantly. At the same time, market demand increased. This meant that within just a couple of years, Intel had to increase its 14nm production by more than 50% to avoid shortages. Only, that didn’t happen.
The die sizes of Intel CPUs over the last few generations (Sq. mm):
- Kaby Lake: 126
- Coffee Lake: 152
- Coffee Lake Refresh: 178
Furthermore, as newer foundries were started, the yields weren’t quite as the existing ones. This meant that a good chunk of the factories was showing yields worse than the existing 14nm capacity. One of the reasons we’re seeing so many F series CPUs (Core i3-9100F, 9400F) was introduced because Intel couldn’t afford to discard chips with faulty iGPUs as the production was already falling short. This was just a measure to utilize as many (otherwise obsolete) wafers as possible.
Therefore, while Intel’s capacity has grown significantly over the last few years, the per chip cost and die area have also more or less doubled. The 7th Gen Core i7-7700K topped out at four cores and eight threads while the 10th Gen Core i7-10700K has twice as many while the Core i9-10900K packs an additional two. The price, on the other hand, has largely remained the same. The chiplet design has allowed AMD to keep the profit margins wide enough but the monolithic design means Intel can’t do the same. The profit margins for the contemporary Core i7s and i9s is likely smaller than any other products Team Blue has ever sold.
The lack of a new core microarchitecture or a new node made this even worse and cost Intel the efficiency crown as AMD started regaining lost fortunes. According to sources, the 10nm yields are still far from what the 14nm node offered a few years back, while the 7nm node has also been delayed by at least another year. All this means that Intel’s foundry and supply issues aren’t going anywhere and the only way out is to outsource its flagship offerings. That’s not necessarily a bad thing, but as expected it’ll sure send the stocks tumbling.