Hi Bilow, let's carry this a bit further.
"It's not trivial to compare these things."
- True, but fortunately to arrive at a conclusion never requires perfect knowledge.
"You have no idea what the product mix of inventory is at Micron nor what the product mix of inventory is at Infineon."
- True, but a variation in DRAM product mix cannot explain such a large variation in inventory write-down as prices for different DRAM parts move too much in sync for that.
"In addition, both companies make a lot more than just DRAM"
- MU was 87% DRAM in FY2001. The inventory write-down number I used for IFX was only the Memory Unit, which is almost 100% DRAM.
"Infineon isn't a US based company and doesn't file reports with the SEC, as far as I can tell. Because of this, you really don't know what the accounting rules are."
- They publish information according to US GAAP, e.g. see their web site.
"Even if the two companies did have the same GAAP rules, we all know that there is massive variablility in how those rules are applied. The two companies could have different reasons for choosing slightly different accounting policies."
- True, but again I don't think that this can explain such a massive difference in write-down. But it could, I'm not an accountant.
"You're looking at quarterly figures, and these are very dependent on the pricing for DRAM at the end of those particular quarters."
- Yes, but since DRAM price went down even further in September, this would mean that IFX' write-down would have to be larger, not smaller than Micron's. It works in my argument's favour.
"In fact, the figures for September 31 for IFX were 142 million Euros."
- I used the figure only for the Memory Unit, which is more accurate, s.a.
"If the figures for Infineon include their share of ProMOS, then Infineon makes a little more DRAM than Micron. If they don't include ProMOS, then Micron is 64% larger."
- Memory Unit sales for IFX in FY2001 (ending 9/30/01) were about Euro 1.5 billion, practically all of which is DRAM.
- MU sales for FY2001 (ending 8/31/01) were about USD 3.9 billion, of which 87%, i.e. USD 3.4 billion is DRAM.
- I.e., Micron's share is more than twice that of Infineon, and this is consistent with all the market share figures I've seen in the last year
- For the argument, it doesn't matter whether the number includes ProMOS or not. Important is that the inventory write-downs were based on these sales figures and therefore are consistent with them.
"Oh, and this link tells which companies are paired together:
tice.com.tw;
Good link, thanks.
"They're not selling the same products. For example, I would bet that Infineon's product line averages out to larger bit sizes. Micron's product line includes SRAM and Flash which Infineon doesn't even sell, while Infineon sells a whole bunch of ASIC parts that Micron doesn't get near. As far as comparing the write down costs for the two companies, why don't you explain what the ratio of inventory in DDR SDRAM to PC100 was at Micron and compare that to the same ratio at Infineon. The fact is that you can't because you don't know."
- I've used DRAM numbers, s.a. You don't need to have perfect information. These things can maybe explain some difference but not this big a difference, see my comment on product mix above.
"No real ASIC engineer in the world is suffering under the delusion that cost is proportional to die size."
- Yes, but this is not the ASIC business, it is the DRAM business.
- Besides ASPs, fab output is by far the dominant driver of profitability. And fab output is driven by dies per wafer plus a series of yields. Amortizing your capex over 10% less dies all the time is significant.
- Carl, as an engineer I think you really shouldn't use loaded terms like "delusion" in an argument :-)
"Micron doesn't push the envelope of technolgy, so their equipment is cheaper. Cheaper equipment makes cheaper semiconductors."
- Micron does push the envelope for shrinks, as can be seen by their die size and by the line widths they use. They just don't push the envelope in wafer size transitions because they believe that material and equipment costs are too high for early adopters.
- True, cheaper equipment makes cheaper semiconductors. But better designs, smaller line widths and larger wafers all make more output over which to amortize the equipment.
- Of course, there is an optimum timing out there somewhere, but getting that right consistently is what we're all dreaming of, right? :-)
"Basically, you're taking an extremely complicated situation, and trying to simplify it."
- There is never perfect information. Fortunately, it's never required to support a conclusion.
"If the die size difference is only 10%, then how did you manage to already "prove" that Micron has a 2x cost disadvantage? Give me a break."
- Keep cool, Carl. I was never saying, let alone trying to prove that Micron has a 2x cost disadvantage. Clearly, that's not true.
- What I am saying is that the sweeping and unreflected statement "it is acknowledged in the industry that Micron is the lowest cost producer" is not exactly a strong argument, and it doesn't become more true by repeating it many times.
- What I am also saying is that there is a first tier and it consists of Samsung, Micron and Infineon, all of which are close enough in terms of cost basis to make other factors, such as resources and commitment, more important in comparing them. Everyone else is just opportunistic and won't be in it long-term.
- Actually, not even IFX and Samsung may be in DRAMs long-term because it's basically not a very attractive business. IFX may exit in a few years. In the meantime, however, they are well positioned for the next 2 years.
"Maybe you didn't notice, but all the Taiwanese just announced that they're making money on DRAM."
- Good, then the first tier will be making even more money :-)
- Also remember, Taiwanese are mostly selling in the spot market and spot market prices rise before contract prices. Unfortunately, they also fall before contract prices. But of course they will never announce that they have begun losing money at the beginning of the downturn as quickly as they announce that they have begun making money at the beginning of the upturn, especially if they want to raise capital :-)
"How are they doing this with those big dies? The fact is that die size has little to do with cost."
- That is definitely not true, s.a.
- ASP is by far the dominant driver of profitability. ASPs have been rising. Spot market rises first. Taiwanese become profitable first. Taiwanese want to raise capital. Taiwanese announce they are making money.
- If the Taiwanese are making money, first tier will also make money, maybe with a delay of a month or two until contract catches up with spot.
- Also, do any of the Taiwanese publish info according to US GAAP? How can anything they say be used for a good argument? :-)
"While it's true that reducing the die size for a particular semiconductor line will reduce the cost of the parts manufactured on that line, it is not true that a line that uses a smaller die size is necessarily lower in cost than another line, with completely different equipment, that happens to use a larger die size."
- "Not necessarily lower", yes, but in most cases the line with smaller die size will be lower cost. That's why everyone, including Micron, is always staying at the cutting edge concerning line-width. Those that fall behind ultimately have to exit.
"If the cost of silicon were the dominating factor in DRAM prices it what you're assuming would be true, but it's not. The dominating factor in DRAM costs are equipment related, not silicon related."
- ASPs are the dominant driver of profitability. Output is next, with a big gap. Then, capex, again with a big gap. Thereafter, all the various opex.
"Shrinking is how you get more out of your equipment, but these companies do not have the same equipment, and when they do happen to have the same equipment, they didn't buy it at the same time, or at the same price."
- True, and that is why it is, of course, important to spend as little as possible on your equipment. Still, fab lifetime profitability analysis shows that output is a much more important driver of fab NPV than either capex or opex, s.a.
"This is all very fascinating, but it doesn't explain why Infineon has failed to take over the DRAM market. In fact, Infineon has repeatedly announced that they are going to reduce their exposure to the DRAM market as much as possible. If their costs are really so low, why do you think that is?"
- Gee, at least I could fascinate you ;-)
- They don't have any cash, mostly because of the balance sheet they were spun out of Siemens with. Also, they want to limit their exposure to what is basically a not a very attractive business.
"Why no compare the lines on the basis of how much their electricity costs? Why not compare labor costs? Why not compare taxes? Why not compare how much retirement benefits they have to pay? Why not compare what kind of environmental rules they have to follow?"
- Because all of these costs are minor portions of total costs.
"Why not compare their ratios of capital equipment cost to sales? That last will give you a better estimate of who has the lowest cost production, but even then, as I've said before, the two companies are not comparable. Micron is a pure DRAM play, Infineon is not."
- That's a good metric. Why don't *you* do that comparison for a change? I'm willing to accept the imperfect comparability and am looking forward to the results :-)
"If IBM and Toshiba are such cheap makers of DRAM, then how come Toshiba is exiting the business, and IBM pretty much only makes parts for internal consumption?"
- These two exited the business not primarily because of cost reasons but because of commitment reasons: 1. It's not a core business for them, and 2. It is a high-volatility business with low margins that is not very attractive if you can do other things. I.e., they have better options. At 87% DRAM, Micron doesn't have a lot of options but to be successful in this business. Which is one of the reasons they are so strong and aggressive and why they will continue to be among the leaders.
"How come Infineon didn't ramp up their production during the good times? You'd think that the guys with the lowest cost production would have taken advantage of their advantage."
- I'm not sure as to which times exactly you are referring to. Moreover, I'm noy saying that IFX has historically been the lowest cost producer. What I am saying is that things are changing, they have moved into the top tier during the last 2-3 years and they have secured their cost position for at least another 2 years.
- I believe they are also more reasonably valued and have more upside in the stock price than MU. Look at the ratios and historical stock prices.
"How come Infineon couldn't come up with a deal to buy Toshiba's plants?"
- Because they don't have any cash. But this is not due to a worse cost position but to the fact that they had to bow to get out of Siemens.
Next, I would kindly request that you come up with more convincing arguments as to *why Micron is the lowest-cost producer* besides the statement that "it is acknowledged in the industry" to be so, which is the only constructive argument I have heard from you so far - and that doesn't exactly fulfill this engineer's requirements for a good argument :-) |
