This article was published back at the beginning of 2012. It is basically a talk by Morris Chang, CEO of Taiwan Semiconductor Manufacturing Company Ltd. (TSM) about the changing competitive environment facing TSMC. He speaks of Samsung (SSNLF.PK), Global Foundries, and Intel (INTC) as new and powerful rivals in the foundry business.
At one point Mr. Chang openly encourages TSMC customers to diversify their foundry sources:
Samsung and Global Foundries are newcomers in the industry, Chang said, and suggested that TSMC's customers should diversify their foundry sources rather than rely on TSMC only.
Mr. Chang's language gets even more grim when he speaks of Intel:
"All of our customers rely on TSMC in foundry production, and Intel relies on its own foundry plants," he said. "If our technologies are not improved enough and Intel keeps improving its technologies, our customers' products will lose competitiveness to those of Intel. It's horrible to imagine the outcome."
This is such an unusual thing for a CEO to say that I re-read the article periodically to make sure I didn't imagine the "horrible" part.
What could the "horrible" be?
Here's what I think it could be:
TSMC has never faced Intel in a directly competitive environment before. In the past, TSMC being a node behind Intel really didn't matter because it didn't build x86 processors. Now being a node or more behind and having Intel breathing down your customers' necks is a major concern.
TSMC made 28nm capacity decisions back in the middle of the great recession. It underestimated the demand and has been struggling with the repercussions of those decisions ever since. In order to have sufficient 28nm capacity to comfortably service leading-edge customers, TSMC needs to spend an immediate $5 billion on additional capacity. That is $5 billion the company doesn't have. If it had the money and wrote the checks today, it would be two years before the new capacity would produce the first wafer. By that time Intel, executing its tick-tock strategy, will be in mass production on 14nm products and TSMC could be 2 nodes plus Trigate behind. That would be an untenable position for TSMC and its customers such as Qualcomm (QCOM) and Nvidia (NVDA). Intel's chips would be 1/4 the size, 1/3 the cost, and higher performance and lower power than anything TSMC could produce.
What faces TSMC is an aroused Intel on the verge of introducing a mobile application processor with embedded LTE. Samsung and Global Foundries are also major challenges.
Since there is no way for TSMC to leapfrog to a competitive position with Intel, I would not be surprised to see TSMC "throw in the towel" on the bleeding edge of fab technology and concede the sub 28nm mobile business to Intel.
I think this is one of the outcomes that Morris Chang speaks of as "horrible to imagine." Maybe there are others. I can't think of a happy outcome for TSMC.
Who wins and loses?
Intel is the obvious winner and will double its market share of the semiconductor industry in three years or less.
Nvidia is a major loser.
Qualcomm will lose a great deal of business and be a shadow of the company it is today.
ARM (ARMH) only gets a substantial nick financially, but the stock P/E could get cut in half or more.
Short ARMH, NVDA, QCOM
Disclosure: I am long INTC.