[Fwd: The Costs of Making a DRAM Chip]

Dave Long dl@silcom.com
Mon, 27 Jan 2003 09:56:11 -0800


For a 32 MB DRAM, we are told we use:
	32   kg H20
	 1.6 kg fossil fuels
	  .7 kg lN2
	+ various chemicals

For which some local prices are:
(assuming that the fuel provides the
energy source for filtering)
	H20:            $17.00/acre-inch
	fossil fuel:    $  .50/L
	lN2:            $  .30/L

and using 102800 L/acre-inch and densities
of 1, .8, and .8 respectively, we get:
	H20:            $  .005
	fossil fuel:    $ 1.00
	lN2:            $  .26
	various:          ??

So unless those "various chemicals" are
rather spendy, the formal costs[0] for
a final cost $25 DRAM far outweigh the
material costs.

Of course, the water price reflects an
area where we haven't had fabs sitting
above the aquifer.  Wouldn't a better
question be "how much would a fab pay
per DRAM if it always used the _same_
water, nitrogen, etc.?" [1]


:: :: ::

[0] or, continuing with aristotelian
accounting, perhaps also amortization
on the efficient costs.

[1] In a pseudorelated vein, it's true
that bothering with memory allocation
doesn't often seem very informative to 
those who write programs, but swapping
continuously doesn't often seem very
amusing to those who run them.