Intel, TSMC: This Is Where The Real Battle Is Taking Place

Includes: INTC, TSM
by: Russ Fischer

Intel (NASDAQ:INTC) is the largest semiconductor manufacturer in the world. TSMC (NYSE:TSM) is the largest semiconductor foundry in the world.

All of the other players in the mobile semiconductor business are more or less irrelevant in the sense that all of them are dependent on the ability of TSMC to provide the appropriate technology at a competitive cost to the dozens of fabless semiconductor companies.

Basically any end customer who buys products from any of the dozens of fabless semiconductor companies is indirectly buying their products from TSMC, and is therefore, sole sourced to TSMC. To understand this defect in the fabless model, please go to this link and look at slide 18. End product companies might spend thousands of engineering hours deciding whether to use a Qualcomm (NASDAQ:QCOM) application processor or a similar device from Nvdia (NASDAQ:NVDA), but in the end they are buying from TSMC since both Qualcomm and Nvdia get the vast majority of their wafers from TSMC.

I remember a comment that I will attribute to Andy Grove of Intel many, many years ago. From memory he said, "If you are going to compete with us, you better run as fast as we do because we are tearing up the pavement behind us." This comment was made back the days of 80286, 80386, 80486, and Pentium. For example, when Intel was making the 80386 and working for 3-4 years to upgrade to the 80486, a handful of companies would work very hard to design and produce a microprocessor that was compatible with the 80386. They would cut the price and begin to take some business from Intel. Then one Monday Intel would release the 80486. That simple act caused the 80386 to instantly become obsolete.

This kind of obsolete wasn't the kind where the business declined in and orderly fashion and the competitor could prepare for the change. Oh no, this was the kind of obsolete where Friday you had a business in 80396s and Monday you had a warehouse full of junk. Even Intel's 80386s were junk, as in "… tearing up the pavement behind us."

A marketing manager for one of these Intel wannabes tried to describe to me the feeling he had on one of those Mondays. He said it was not much different than the feeling one might have to coming home to see your house burned to the ground and every material thing that you care about is gone forever. I've never been party to such a depressing conversation.

Well, the same thing is going on today. The only difference is that instead of new parts obsoleting old parts, we have new processes utterly obsoleting old processes.

When Intel shipped the first Ivy Bridge processor using their 22nm Trigate process, while TSMC's customers were complaining about insufficient capacity of 28nm from TSMC the message went out to the world that Intel was tearing up the pavement again. TSMC's 28nm process was in the same place as that wannabe 80386 was back in the 1990s.

The demoralizing thing to TSMC and their fabless customers is the fact that they know that Intel is demolishing 32nm capacity, capacity that TSMC would dearly love to have, to make room for the Intel 14nm ramp up later this year which only makes their problem worse.

The simple fact is that node reduction in size reduces cost by 50%, or, in the alternative, nearly 100% more transistors (another word for functionality) can be added at the same cost. This year Intel will, at least for a while, be two nodes ahead of the best in class competition, TSMC. This means that Intel can manufacture four times the same chip on the same amount of silicon as TSMC can. That means the Intel's cost will be 25% of the TSMC cost. That means that if a fabless competitor wanted to sell at the same price that Intel sells with 60% gross margins, they would be selling product at approximately a 40% NEGATIVE gross margin. I think it is obvious that his business model can't work for very long.

The very important point here is that being 1 1/2-2 nodes behind is functionally not much different than being too far behind to EVER catch up. What would there be to catch up to? By the time, if ever, that TSMC could get to high volume 14nm all the interesting business would have migrated to Intel, primarily through adoption of better, cheaper Intel proprietary (X86) devices, or through foundry activities for the hard-nut OEMs like Apple. And to pour salt in the wound, Intel would likely be ramping 7nm by that time.

You will notice that I have not included Samsung (SSNY.OB) in this article. The fact, confirmed this morning, that Samsung is using the current (old) version of atom for use in their new line of android 10" and 8" tablet computers is very clear handwriting on the wall. I could write another article speculating why Samsung would do this with no scenario favorable to the TSMC and their fabless clients.

The conclusion is clear: Intel will win; all other mobile chip sellers will lose to a more or lesser extent. Mark June 3, 2013, the release of the new Intel Haswell processor, as the inflection point for Intel in mobile.

Disclosure: I am long INTC. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

About this article:

Author payment: $35 + $0.01/page view. Authors of PRO articles receive a minimum guaranteed payment of $150-500. Become a contributor »
Problem with this article? Please tell us. Disagree with this article? .