Rumors have been swirling for some time in tech circles about difficulties the Taiwanese foundry, Taiwan Semiconductor Manufacturing Company (TSMC), is having in bringing its latest cutting-edge chip fabrication lines into full production. The stories were confirmed when QCOM warned in its latest earnings conference call that over the next quarter or two it would be unable to supply customers with all the most advanced chips they wanted. (Interestingly, in its quarterly earnings call, AAPL said it would be unaffected because it isn’t using 28 nm chips.)
Why is this important?
1. For many semiconductor chips, the history of their manufacture is one of constant attempts to make more complex and faster speed, but also smaller, less power-hungry and cooler output. One of the main ways of accomplishing all but the first of these goals has been to shrink the spacing between the lines of the chip patterns written onto silicon.
2. A nanometer is a billionth of a meter. The 28 nanometer spacing that TSMC is having trouble with is, therefore, a distance between lines of 28 billionths of an inch.
3. About twenty years ago, the foundry–or third-party manufacturing–business began to come into prominence, as several positive factors for that industry converged. The increasing complexity of semiconductor “fabs” meant that it cost $3 billion to build one. Even worse, a fab churned out $7+ billion in output, far beyond the sales of all but the largest companies. At the same time, a generation of ambitious chip designers wanted to break away from stodgier established firms and develop chip designs on their own. Many focused on customizing templates provided by ARM Holdings (ARMH).
4. The unquestioned leader in the foundry arena is TSMC.
a paradigm shift in the offing?
There are two big integrated semiconductor designer/fabricators left–INTC and Samsung. Neither is having fabrication problems. INTC is beginning to produce 22nm chips in volume, and promises 14 nm for 2013. In addition, it is using a new production technique that it calls “3-D,” that gets an unusually large benefit from its current linewidth shrink. Most important, in my view, is that the company seems increasingly concerned with providing customers with products they want, rather than just the latest engineering tour de force.
Samsung already provides foundry services to others–it builds many AAPL chips, for example. And INTC’s mammoth capital spending campaign of 2011-12 has analysts asking–and the company denying–that it intends to offer similar foundry services in the future.
ARMH, which has–with justification–been an immense market outperformer as the one-stop-shopping way to play the mobile device chips that the design firms/foundry model has been churning out. But the stock (at 55x historic eps) is down about 20% over the past year, a time when INTC shares (12x) is up by 25%. Over the same period, Samsung Electronics (5930.KS) (15x) is up 50%.
Yes, the issue with ARMH may just be the high PE multiple. And, yes, Samsung isn’t just chips. It’s a force in smartphones and dominant in TVs. And it trades in a market that marches to its own drummer. But I think the market is saying that the old integrated model has more going for it than the consensus appreciates. I also think the market is right.
TSMC’s fabrication difficulties may be the trigger that gets a wider group of investors to focus on the change.