For years, chip giant Intel (INTC 0.29%) has touted the fact that it develops its own chip manufacturing technologies and builds its own chips using those technologies as a significant competitive advantage over most of its peers.
Intel's peers, by and large, rely on third parties to manufacture their chips. Companies that don't manufacture their own chips are known as fabless semiconductor companies.
Image source: Intel.
When everything goes right, owning and operating one's own chip manufacturing plants can be a huge advantage, because it can allow a degree of collaboration between chip design and chip manufacturing teams that's difficult to achieve otherwise, which can potentially lead to better products.
Additionally, third-party chip manufacturing companies don't work for free -- they typically generate significant gross profit margins on the chips that they sell to their customers. A company that builds its own chips effectively gets to keep those margins to itself.
However, a company that relies on itself for chip manufacturing also faces substantial execution risk. If a company both designs and manufactures its own chips, then issues on the manufacturing side of the house can lead to severe issues for the chip design part of the company.
This risk factor is playing out in front of our very eyes with the issues that Intel has been facing in bringing its 10-nanometer technology into production. Allow me to explain.
A fabless company can go elsewhere; Intel cannot
A fabless semiconductor company generally has the option of working with multiple potential chip manufacturing partners on a given product. It does cost more for a fabless semiconductor company to build versions of its chips for two entirely different manufacturing technologies, but such a move substantially reduces the risk to its own business.
Indeed, suppose that a fabless semiconductor company -- Company A -- is working with two chip manufacturing companies (Company X and Company Y) but it becomes clear at some point in the middle of the product development cycle that Company X just isn't meeting its goals. Say, its manufacturing yield isn't improving at the expected rate, or the performance of the manufacturing technology just isn't where it needs to be to support Company A's product.
Then, Company A can still get its product out on time by indicating to Company Y that it will be shifting most, if not all, of its orders to it and then Company Y can adjust its planned manufacturing capacity plans to accommodate the shift in orders. Company X can continue working to get its technology in order in the hopes of capturing some of Company A's orders, but in this case Company A won't suffer a delay or setback as a result of Company X's failure -- only Company X and its shareholders suffer.
With Intel, it's different. Intel's product teams develop their main product lines exclusively to be built using Intel manufacturing technology in Intel's factories. In recent years, Intel's manufacturing group simply hasn't delivered its manufacturing technologies on time. This has led to substantial product delays as well as the last-minute insertions of stop-gap products on older manufacturing technologies that were far less competitive than the originally planned products on more advanced manufacturing technologies were.
And, in many cases, those originally planned products were outright canceled.
This has all had the effect of weakening Intel's competitive positioning in the markets that it participates in.
A cloud of uncertainty
Right now, Intel is facing what is arguably the most intense competitive environment that its business has seen in many years. Its fabless semiconductor peers continue to improve their execution and are going aggressively after Intel's core markets, and those fabless semiconductor companies are backed by several large, well-funded, and highly competent chip manufacturing companies.
Against this backdrop, Intel's chip manufacturing division is executing poorly. The company's 10-nanometer technology, which was originally supposed to be ready for production by the end of 2015, still hasn't gone into mass production, and it looks as though the first high-volume products built using Intel's 10-nanometer technology -- processors intended for laptop computers -- aren't going to go into production until very late in 2018 or early in 2019.
Image source: Intel.
Given that laptop processors tend to be relatively simple compared to data center processors, the issues that Intel is facing with its 10-nanometer technology could make it difficult for Intel to launch processors for its important data center business using this technology until, perhaps, the second half of 2019.
The situation doesn't look good right now, and Intel's apparent opacity in sharing its 10-nanometer product introduction plans with investors only makes it worse.
Perhaps this situation will force Intel to reconsider owning and operating its own chip manufacturing plants and will become the next big chip manufacturer to ultimately become a fabless semiconductor company.
