|
Don't bet the ranch: WWK's study weighs the possible economic impacts of 450-mm fabs |
|
|
|
Feb 19, 2008 at 03:05 PM |
Anyone following the industry debate about whether and when the semiconductor manufacturing community should move to the next larger wafer size knows that economics, not technology, will be the driving force behind the ultimate decision.
But trying to get a handle on the specific economic impacts of 450 mm on chipmaking remains elusive, with some studies lacking the kind of unbiased objectivity or proven cost-of-ownership methodologies that the subject demands.
Wright Williams & Kelly has just published its own independent study on the topic, using the company's cost modeling and simulation capabilities to compare the economics of 300- and 450-mm fabs. The WWK team got a wake-up call when it tallied the results of an April 2007 survey about the "expected arrival of 450 mm." To no one's big surprise, most respondents said that the larger substrates would come into play no earlier than 2013, if not later. But when nearly 40% of those surveyed believed that 450 would never happen, that really caught the attention of the surveyors and pushed them to take a closer look at the possible economic impacts of those big ol' slices of silicon on the fab world.
WWK's David Jimenez sent me the intriguing 16-page report, but pleaded with me not to share too many details, since his firm wouldn't mind selling a few copies. The study starts off by looking at the historical drivers for wafer-size transitions in the 450-mm context, discounting die-size as a "driving force," balancing considerations of prior-node learning and capacity and technology requirements, and coming to the conclusion that ultimately, cost per bit will be the "bottom-line justification" for moving to 450 mm.
The report really kicks into gear with the section describing the estimated manufacturing requirements for a 450-mm fab and goes into overdrive with the two chapters on economic simulation inputs (tool, building, wafers, materials, and consumables costs) and the results of the simulations. It continues with possible motivations for going to the larger wafer size and a speculative chapter titled "What about 2025?" then comes to a close with the key conclusions. Although it doesn't have all the answers, the study should help bolster the arguments of those involved in the industry's 450-mm discussion who have been gesticulating and shouting "slow down!" or "stop!"
Without giving away too much, here are a few random morsels. The study estimates that the cost of those 2 million none-too-cheap 450-mm test wafers that will be needed for fully qualifying the first equipment sets will come in at about $7 billion, up significantly from the billion or so needed to get the first 300-mm sets ready for bear. It also states that because of much higher equipment and starting materials expenses at the larger wafer size, "even if R&D costs associated with 450 mm were free, our economic models show that 450 could be as much as 36% more expensive as 300-mm manufacturing at the same node."
At the end of the motivations section, the question is posed, "will CMOS run out of capability before 450-mm equipment is fully depreciated?...WWK expects that it is very unlikely that the 450-mm equipment installed in 2016 will be capable of manufacturing 13-nm devices ---devices that are about one half the minimum geometry as the processes for which the equipment was originally designed." As for 2025, that's the year in which Jimenez, Dance and Co. believe it "makes a lot more sense" to introduce 450-mm manufacturing.
After reading over the study, I contacted WWK's Daren Dance, one of the coauthors of the report, and sent him a list of questions. Here are some of the Q's and A's from our email conversation:
What were some of the biggest surprises in the findings of the report?
The size of the cost difference. We were not surprised that there was a cost difference, as there generally is for a new wafer size introduction. The other surprise was how strong the 200-mm market remains.
What are some of the key lessons/findings in the report?
For wafer fabs, the keys to making 450-mm work are controlling wafer cost and equipment cost. For equipment and material suppliers, don't overlook the revenue opportunities at 200 mm and 300 mm.
Are there any recent developments that might "tweak" the findings of the report, such as recent tech breakthroughs or changes in the new edition of the ITRS?
Areas that we are watching include die-size trend changes, lithography and mask/reticle costs, and capital availability for semiconductor fabs. The wafer-size decision is more economic and less technology driven.
How might efforts by ISMI and others lead to possible changes in the findings scenarios and 450's advent in general?
We are excited to see ISMI look at 300 mm again. The industry has learned a lot about 300-mm manufacturing that can be used in continued productivity improvement. Success with 300-mm Prime may tend to postpone 450 mm even further. The capital-intensive nature of semiconductor manufacturing means that almost any productivity improvements from 300-mm Prime or other innovations will have significant economic benefits for manufacturing.
Might there be potential wildcards that could significantly alter the outcome of your study, or change the course of 450-mm development? Examples might include EUV lithography's ultimate fate and other possible alternative litho strategies, polysilicon shortages, introduction of post-CMOS approaches before the CMOS runs out of steam (i.e., carbon nanotube interconnects, 3-D interconnects, post-hafnium metal gates, etc.)
Any of these factors will impact the economics of 450 mm, but we think that one of the largest unknowns will be 3-D device technologies which could continue the trend toward smaller die sizes and further postpone 450 mm. We will be following up on these areas in our April 2008 survey. We are also planning studies on some specific areas of potential 300-mm productivity improvement.
Are there any arbitrary assumptions used in the study that might be deserving of a more specific, granularized approach in future studies?
Two specific areas of the 450-mm question need further study: the cost of developing prime single-crystal 450-mm wafers and the costs of developing prototype equipment for 450 mm.
In your opinion, is it possible for the tool and materials suppliers to stand firm, if they so choose, and refuse to move forward on 450 until and unless the IDMs and foundries offer certain financial guarantees?
Any equipment or material supplier can only afford to spend so much on R&D. With the slowdown in capital spending forecast for 2008, there will actually be less money for R&D at just the time when the potential for 450 mm is increasing demand for R&D investment. Suppliers must understand whether the bulk of their customers---not just a vocal few---want this money spent on technology improvements or productivity improvements. And if customers choose productivity improvements, will productivity be best delivered through 450 mm or higher productivity versions of 200- and 300-mm equipment?
If 450 mm does not appear to offer better returns on R&D investment than other options, then the R&D funds---and industry momentum---will flow in other directions. Any other decision would not be fiscally responsible. The 450-mm R&D decision will be less of "standing firm" and more of trying to meet Wall Street's expectation of fiscally responsible management.
|
Comment by GUEST on 2008-02-25 11:40:53
Is it determined that 450mm is the desiderata?
What about 400mm or up the odds to 500mm? |
|
|
