Unless Intel Wins Apple, Foundry Is a Waste of Time

Intel (INTC 0.61%) has been pretty loudly telegraphing its intentions to become a major, high-end semiconductor foundry player. Its highest-profile win to date has been FPGA vendor Altera, and it is presumed that the company is working to bring aboard more customers interested in the leading edge. Some investors believe that Intel's future is as a semiconductor foundry, but if you really dig deep into the numbers, there's just not enough business there to make it worth Intel's while.

How big is the market for a leading-edge foundry?
In the foundry landscape, the latest technology node is known as the 28-nanometer node. (This is actually a bit misleading as there are several flavors of 28-nanometer out there -- 28 poly silicon for cheaper, lower-performing devices, and 28 high-K/metal gate for higher-performance devices.) To get a sense of the market opportunity for a leading-edge foundry, it's worth looking at just how big the market for a 28-nanometer foundry is today.

Taiwan Semiconductor (TSM 1.60%), the world's leading semiconductor foundry, saw its 28-nanometer node begin to ramp into volume by late 2011 to early 2012. Here's a look at 28-nanometer revenues at TSMC by quarter:

In 2013, TSMC did roughly $6 billion in leading edge 28-nanometer revenue. If we assume that TSMC's share of 28-nanometer was about 84% in 2013 (per TSMC's own statements), this implies a leading-edge total addressable foundry market of about $7 billion. Note, however, that the second largest 28-nanometer player in 2013, Samsung, saw most of its business come from Apple, as it builds Apple's A-series processors, and the third largest player, Global Foundries, was likely mostly revenue from Intel rival Advanced Micro Devices (AMD 1.36%).

This business isn't worth it without Apple
Intel is unlikely to build chips for direct PC-chip competitor AMD, nor is it likely to build chips for Qualcomm (QCOM 1.62%), the world's leading smartphone chip vendor and by far TSMC's largest leading-edge customer, as that would eliminate Intel's biggest competitive advantage for its own processors. So we're talking about a $7 billion-per-year total market, most of which is silicon from Qualcomm, AMD, NVIDIA (NVDA 4.35%), and Apple.

Since AMD, NVIDIA, and Qualcomm would be extremely poor strategic choices for Intel, the only real "large" customer that Intel will be able to win at the leading edge would be Apple. Even then, if we assume that Samsung gets between $10 and $15 per chip from Apple for its very leading edge, this wouldn't amount to more than $1 billion of leading-edge business per year. (Remember, not all of Apple's phones are the latest one in any given year, so the portion that is leading edge is much lower than the total Apple business.) This just isn't all that much and would be more of a strategic win (i.e., to keep that revenue out of the foundry landscape) than it would be a way to meaningfully increase revenues.

Foolish bottom line
The leading-edge foundry business is worth about $7 billion a year, which just isn't all that much to a company like Intel, which does $33 billion a year in PC chips and an additional $11 billion to $12 billion a year from server processors. The foundry business works for pure-play folks like TSMC because foundries don't need to spend a dime on actual chip development. Intel, on the other hand, is investing nearly $3 billion a year in mobile processor R&D, so taking competitor business (from Qualcomm and NVIDIA, for example) doesn't make sense, and even something like an Apple deal, which -- if Apple went long-term exclusive with Intel -- would be worth about $1.5 billion to $2 billion, with even less than that at the leading edge.

Intel is better off developing competitive mobile products and capturing much more value that way.