In November, an article indicated that Intel was ready to open up its fabs to other chip designers.
AMD famously spun off their foundry capabilities into The Foundry Group, basically admitting that it was too expensive to maintain investments in fabrication while only manufacturing their own chips. The move also let AMD leverage competitive investments at other fabricators such as TSMC (Taiwanese Semiconductor Manufacturing Company). ARM paved the way for this kind of thinking many years ago, being one of the first fabless semiconductor companies. NVIDIA decided to go down this route as well. It’s just too expensive to maintain fabs yourself unless your volumes are absolutely massive.
There are some great reasons to have a fab, though. When making a chip, there are many factors that go into its profitability. Among them are size, power, performance, and cost of materials. Owning a fab and the process used therein enables you to optimize each of them. Intel has long held the performance crown by way of being vertically integrated, from architecture & design, down to the process used to lay transistors.
As transistors have gotten smaller, it has taken more and more ingenuity to cram more computing power into the same area. It takes a long time to research the materials science involved, and Intel has lots of proprietary technology in this area. So much that the rest of the industry feels the need to band together to compete.
When a contract semiconductor manufacturer allows others to use its process for making chips, it has to disclose lots of details to the chip designers. This is because chip designers need to simulate their chips on the process to understand nano-scale effects such as wire capacitance and impedance before sending the final designs to be manufactured.
By allowing others to build chips on their process, Intel is opening up a lot more than just a revenue stream. Of course, they realize this, and it is unlikely that they will end up making AMD’s chips. However, rumors this past week had Intel in a position to manufacture Apple’s A4 and A5. However, it is unclear to me what Intel has to gain from making ARM chips better than the other guy.
There are lots of wins to be had for Apple to use Intel’s process technology on these chips. The A5 is massive and could benefit from smaller, tighter process technologies with higher yields. With the relationship between Apple and Samsung souring, the move makes sense.
However, I think there is another move that could be even more interesting for Apple to make: buy its own fab. Apple always strives to be vertically integrated, and this would be one more step along the way to not relying on anyone else. If Apple could then build an innovative portfolio of process technologies, it would hold a yet larger lead on competitors.