Taiwan Semiconductor Thinking Too Far Ahead

Although the current 12″ (300mm) wafers on which most chips were produced were initially used as early as 1998, they didn’t enjoy widespread adoption until 2003. Given that it is normally a decade or more between wafer migrations (as opposed to the 2-3 year intervals between transitions to smaller transistors on each wafer) we were a bit surprised to see this InfoWorld article:

Semiconductor industry looks ahead to 18-inch wafers

Taiwan Semiconductor Manufacturing Co. (TSM), is working with an industry group led by the International SEMATECH Manufacturing Initiative [ISMI], to figure out how to make 18-inch silicon wafer technology more economical, as well as overcome some technical challenges. It’s important that they find a solution to the problem, because without the new wafers, chip manufacturers won’t be able to reduce costs at the pace users and gadget makers have come to expect.

Our response to that was that maybe it is time for chip manufacturers, users and gadget makers to adjust their expectations - even if only slightly. The article continues:

The trouble is that a factory designed to make chips on 18-inch wafers could cost between $12 billion and $15 billion to build, nearly triple the price of an equivalent 12-inch wafer factory. Not many companies could afford to build such factories, according to Stanley Myers, president of Semiconductor Equipment and Materials International, in a letter on the industry group’s Web site.

Chip makers will only buy into the new technology if it makes sense financially.

If only that were true. Luckily most chipmakers won’t have this opportunity to dig themselves into a hole for a few more years. By then they should have clawed their way out of the current one.

Comments

[Brian Bandsma:]

They want to upgrade to 16" wafers for the same reason they are going from 8" to 12". The per chip costs goes down by 30%. Just like most Taiwanese hardware manufacturers the business model is to bring costs down. As mentioned very few companies are willing or able to make that kind of investment, especially US based ones. Hence more IDM outsourcing to foundries, even more so if you can no longer compete with the cost structure. Although TSMC is 40-50% of the foundry business, depending on how you measure it, it is only 7% of total global semiconductor manufacturing.

I would also not be too quick to criticize Morris as his past investment decisions have built a company that is unmatched in the industry. As I recall, TSMC has an R&D budget equal to the next three largest foundries combined. So pushing the technology curve forward is a critical competitive advantage for the company.

Even in this seasonally weak quarter the company still generates an annualized ROE of 24%, ex-cash. And it is the only foundry that is consistantly profitable. Even when the tech bubble burst and orders dropped off a cliff they still generated a 10% operating margin.

2007 Apr 27 04:46 PM