450 mm: One Man's View

We've been writing about the wisdom and challenges of the transition to a 450 mm wafer size for most of this year. In support of the transition are some very large companies (Intel, Samsung, Toshiba and Taiwan Semiconductor Manufacturing Co. Ltd.). Other large IDMs (IBM, Micron and Hynix) say they would prefer that the industry focus on solving the many productivity challenges that remain in 300 mm production.

Recently, our intrepid West Coast Editor Alex Braun sat down with Tom Caulfield, executive vice president for sales, marketing and customer service at Novellus (San Jose), to talk about 450 mm.

Prior to joining Novellus, Caulfield spent 16 years at IBM. In his most recent assignment as vice president of 300 mm Semiconductor Operations, Caulfield was responsible for the financial management and operation of IBM's premier semiconductor facility located in East Fishkill, N.Y.

Here's what he said:

"Our industry has been experiencing modest growth in the capital equipment sector because of high productivity. Much of this is attributable to significant productivity improvements on existing 300 mm tools. Sematech's 300mmPrime also focuses on this approach by investigating ways to make existing 300 mm applications 30–40% more productive. This chips away some of the value-add 450 mm would provide if 300 mm had remained at previous levels of productivity. It further reduces whatever momentum there is for the move to 450 mm.

"The debate is ongoing as to whether there is enough business to generate the OEM ROI from the dollars required to do 450 mm development — about $36B. Also uncertain is the volume needed for that ROI. If the 450 mm equipment users are a subset of the 300 mm customers today, there will be even fewer ROI dollars. The 300 mm wafer was a winner because the productivity was there and everybody could use it. I don't know whether these economics also hold true for 450 mm. Not every design will work well on a 450 mm wafer.

"While 450 mm is being debated, we continue to offer productivity improvements in all our main 300 mm products. By splitting this effort's cost with our customers, we can still earn the revenue needed to support this progress. But we don't have the resources to undertake 300mmPrime and 450 mm at the same time.

"Intel is doing a solid job in bringing the subject of 450 mm to the table and proposing ideas on how the transition might be carried out, such as more commonality in wafer handling in tools and process modules. The problem is that not every tool configuration lends itself to that approach. No one thing works for everybody. I don't know how there could be enough commonality to avoid having the industry come up with four versions of a process and spend four times the amount. Frankly, I don't think the current proposed efforts will reduce development costs much. Maybe they'll cut 10 or 20% off a very big dollar number — not nearly enough to improve the economics and dynamics of the situation. So we must continue the discussion to find a solution that works for all sides.

"If 450 mm gains momentum, the transition might come around 2015, depending on the economics of the situation. I see the first 450 mm fabs going up after 22 nm. Again, this has nothing to do with Moore's Law or our ability to push electrons and atoms, and everything to do with affordability.

"It also depends on whether unit demand continues to grow at a 10% CAGR. With the exception of three years, since 1970 the electronics industry has grown at an average annual rate of 5% to become a $1.4T industry. That's close to the run rate for the Gross Domestic Product. That growth has driven unit demand because of the increasingly higher semiconductor content in each product, which increases at a 10% CAGR. If that rate falters, perhaps due to a reduction in demand for consumer products, it would put even more pressure on the economics behind the move to 450 mm.

"The economics of our industry can remain intact only if the semiconductor value proposition that drives the consumer to purchase electronic gadgetry, and the semiconductor manufacturer to buy capital equipment, continues."

Thank you, Dr. Caulfield, for shining some light on this difficult issue.