Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: Elmer Phud who wrote (138169)	10/29/2004 8:01:13 PM
From: pgerassi	Read Replies (1) \| Respond to of 275872

Ephud: Youir numbers do not multiply up. The wafer costs are for a 30K WPW fab with 95% up time (high according to others who should know). 30K WPW * 612 GDPW * 13WPQ = 238.68 million good CPUs per quarter. Since Intel doesn't sell that many even in 1 year and they have 4 300mm fabs, the production rate is way overblown. Also GDPW does not take into account saleability as I get reamed out by process guys for using that for "yield". If it runs at 1GHz, a P4 is a good die. It works, but who would buy one? And even then, what would they pay for it? And how much lower would it need to be after everyone got their needs satisfied to pony up for another one? Intel may have a 1.4GHz bin, but they need 1.5GHz at nominal voltage to sell. Intel had to cut production this quarter because they were making too many to sell. That raises the bar for saleable bins. So first we should cut your WPW down to 10K or so, boosting your wafer costs to $6-7K. Your saleable bin yield to 60% instead of the 76% you used (and it might be even lower still). We also should use the Prescott die size of 112mm2 and kick the process from that used of "6M CMOS 0.13um Cu" to "7M 0.09µm CMOS Cu Low-K Inter-Layer High-K Gate Strained Si" which is quite a bit higher cost. That puts us to over $25 before packaging. That means that $30-35 is the after packaging cost which is where I placed it. Boy, the MPR reference wasn't far off. And the 2M version of Prescott is going to make it worse still. Doubling that to add in the rest of COGS not yet included gets us to the $60 COGS/CPU cost cited previously. One speed bin of ASP drop (between $11 on the sweet spot and $300 on the upper end) would knock 3% off a 60% GM to 57%. Two bins down knock about $57 off of ASPs at the sweet spot reducing GMs to 27%. At that point, Intel loses about $2.7B in revenue wiping out all of their operating profits. They would likely show a $200 million loss, without the reduction from the buyback as they normally don't take it out. And in Q3, they showed a 3% loss in GM. For Q4 they guided to another %3 loss (a 15% production cut back would do that). So it is emminently feasible that they are already feeling the speed bin drop wrt AMD. If Q4 makes for another half a bin drop, Intel's GMs could go below %50 feeling a real pinch in their operating profits. This would be realized by AMD getting $300 million in additional CPU sales (1M * $200 90nm A64s + 200K * $500 Opterons) which hit Intel by $600 million (1M * $400 high end P4s + 200K * $1K Xeons) or about $15-20 in ASPs. So COGS/CPU rises to $69 and ASPs fall to $130 making GMs at 47% (of course I do not think it will be this bad quite yet, Q1 maybe). $900 million hit pre tax and $630 million hit post tax or about $0.10 EPS drop. Given that Intel gets sheared 10% for a $0.01 miss, what would a dime drop do? Cut it in half? Pete

To: Elmer Phud who wrote (138169)	10/29/2004 8:07:53 PM
From: TimF	Read Replies (1) \| Respond to of 275872

If you have $2500 per fully processed wafer and 612 gross die per wafer that would give you a manufacturing cost of a little over $4 per gross die. But doesn't gross die include all die including those that don't work or that bin too low? To know the cost per net die you would have to know the yield (counting thoe that bin too low as not yielding a good die). Then you would have to take those good die and package them. I think that would cost more then making them. I forgot to add test costs, and there is also shipping/ distributiion (pretty low I think but I don't know how much), marketing and general overhead (marketing, R&D, management and administration, fab construction costs ect.) Maybe it only is $4 manufacturing cost per gross die produced (and maybe $4.50 to $7 per net, but I really don't know Intel's yields so this is just a guess) but I think manufacturing cost per net die is only a fraction of the total cost (including overhead) Edit - Message 20704670 Maybe $15 is reasonable for variable costs (might even be less at least for the smallest dies) but there is the fixed costs of the fab and their are other overhead costs. I have no way of estimating them but I doubt they would be negliable. Tim