Silicon Strategies caught up with Steve Appleton, chairman, president and chief executive officer of Micron Technology Inc., at London's Savoy Hotel where Micron was hosting a conference for financial analysts. Peter Clarke asked Appleton about the challenges that lie ahead in the highly competitive memory chip markets.
Silicon Strategies: How do your sales split between DRAM, flash memory and image sensors, and how do you see that changing over the next year?
Steve Appleton: Well we don't break out the exact percentages. About 80 percent of sales is standard DRAM and the rest is made up of non-standard DRAM, flash and image-sensor ICs. But flash is growing faster than DRAM.
Where we want to be in one year is too tough to answer. In three to five years we want to have the same market share in NAND flash as we have in DRAM, which is about 15 to 20 percent.
Silicon Strategies: What is the status of your conversion to 300-mm manufacturing?
Appleton: We're ramping [Manassas] Virginia now. That's all 300-mm production and its dedicated to DRAM so we're making 256-Mbit and 512-Mbit memories. The final capacity is 40,000 to 45,000 300-mm diameter wafers per month. That's 90,000 to 100,000 8-inch equivalents.
We're already committed to take it up to about 10,000 to 15,000 wafers month by the end of the first calendar quarter, then we'll see. We have another 300-mm wafer fab in Lehigh, Utah, but that's an empty shell.
We will complete Virginia first, so Lehigh is not a 2005 thing.
Virginia is massive. By the end of 2005 it will be about 15 percent of our production and bigger percentage of DRAM. By the end of 2005 about 25 percent of our DRAMs will be made in Virginia.
Silicon Strategies: And what about the transition from 110-nanometer to 95-nm manufacturing process technology?
Appleton: Were in early production. Boise will commit to mid-2005. But Virginia won't go to 95-nm. It will go straight to 78-nm manufacturing because 95-nm will have a short life. We expect to be in production [with 78-nm] in the second quarter of 2006.
Silicon Strategies: And what about expansion of manufacturing into the far east?
Appleton: There's a difference between expansion and conversion to 12-inch wafers. We are already the most geographically dispersed producer in the world. We're in Japan, how much further east can you go? About 11 percent of our revenues already come from China. China is the second largest PC market in the world and the largest mobile phone market.
Silicon Strategies: Well what about the need to expand in China then?
Appleton: Well we do plan to have a test and assembly plant there, but having a fab in China is not necessary today. You do need to have your product flow located close to the customer but that becomes more true the closer the products get to finished goods.
Ultimately will it be necessary for us to have a wafer fab in China? Probably, but not for the next couple of years.
Silicon Strategies: Some of your competition seem to think they need to have a fab in China, such as Hynix and STMicroelectronics.
Appleton: Well that's on flash. Running a fab in China has advantages and disadvantages. On the positive side there's low cost of labor and low-cost of construction, but this is not that biggest proportion of cost. They have no advantage on the equipment side.
Silicon Strategies: Unless the equipment vendors are cutting sweetheart deals to try and get inside places like SMIC?
Appleton: If we ever caught a fab equipment vendor subsidizing our competition we would cut them out immediately; we are the big spenders not the Chinese fabs. They are paying the same for equipment.
On the disadvantage side, they have power interruptions, lack of infrastructure, there are training issues, and once you have them trained on these science parks there is a very high turnover of staff.
Silicon Strategies: But perhaps the Chinese fabs have a lower cost of capital. They can borrow money from the government and are getting tax breaks.
Appleton: Look at SMIC and others. They are going to same equity markets that we all go to.
Silicon Strategies: Perhaps rather than investing in manufacturing in China, Micron should run an asset-light model and license other fabs to make memories for you?
Appleton:That's just enabling the competition. There is no foundry model yet that has stood the test of time.
Infineon with Nanya is different. That's a joint venture where you get access to half a fab at half the cost. So why not spend twice as much for the whole fab. But with Infineon licensing SMIC, they've just enabled a competitor
Silicon Strategies: How is it that Micron doesn't need to use leading edge lithography equipment for manufacturing DRAMs?
Appleton: We've found a way to push 248-nm wavelength lithography to do a 95-nm manufacturing process -- that's the half-pitch of the first metal.
Silicon Strategies: Yes but how? Is it a matter of additional resolution enhancement techniques?
Appleton: We have our own mask shop. That's very important. And we have a unique manufacturing process. It's a combination of design, reticle and process. But we can only go so far. We'll use 193-nm wavelength lithography for 78-nm. We've already got some 193-nm tools in.
Silicon Strategies: What about your capital expenditure plans. Do you see any need to change that?
Appleton: No. We spent $1.4 billion in fiscal 2004 and we plan to spend $1.5 billion in fiscal 2005 [which ends September 2005].
Our plan was front-end loaded so we have already spent a lot of it. We won't spend that much in the second half of the fiscal year.
Silicon Strategies: Jumping back to flash briefly. One could argue that with the cut-throat competition and Samsung having a controlling influence in both the DRAM and flash markets, Micron should not be trying to get into flash or is too late in trying.
Appleton: I would put it the other way. We absolutely have to get into the flash market.
There's competition in all markets but the ASP [average selling price] in flash means there's still more profit to be made in flash than in DRAM.
We have a 2-Gbit flash memory in 90-nm that is as good as anything anybody else has today. We're not too late with the technology. We believe the ASPs will behave just like they do in DRAM. It's price-elastic and it's bit -driven.
Also the cost of NAND flash development has been supported by DRAM. Like DRAM it needs advanced processes, not the same, but similar. It costs me 1.3 times the cost of developing DRAM technology to develop DRAM and flash technology. No other product can give me that research and development reuse.
Silicon Strategies: Finally I should ask, with all the rather somber financial news coming out from semiconductor companies recently, how you see Micron's markets in the immediate future?
Appleton: Our business is contradicting what the other semiconductor companies are seeing. They are predicting a weak Q4. Our market is pretty good.
PC strength has been pretty reasonable. And IT corporate spending is also pretty reasonable. We're probably a typical corporation and right now we're spending 15 to 20 percent on refreshing the infrastructure. Thirdly, demand for memories outside computing is strong. If you just look at cell phones. That's 630 million cell phones or 14 percent [annual] growth in units but the amount of memory inside those units is increasing dramatically. The cell phone three-to-five year CAGR is 80 to 100 percent bit growth per year.