[Edited] @Petz / capacity
When I read the inventory comments in the last days or weeks,
it seems that Intel stuffed the channel, which isn't good for
both players, of course, but you mentioned this too, I'm more
worried about the TOO LOW!!! inventory at AMD. Finished goods
at the end of the quarter were 127Mio. which is extremly
low. When I assume a "GM" of 60%, the then sold inventory would
cause around 210Mio. Revs, if we take a 50% "margin", this
would lead us to around 250Mio. - hard to know, how AMD is
valueing its inventory, we in germany learned, that they
have to book at fully costs. But the number isn't that impor-
tant if we let a few millions beside, with around 1300Mio.
for CPG, this would mean around 250/1300= 0,19 * 13 weeks =
2,5 week inventory. Just to be clear, I would like it to be
at around 4-6 weeks, so I hope that AMD could DOUBLE inven-
tory, which will not happen (I assume that).
I have made a sheet with wafer ins and outs and FAB capacity
which I have to work further just to get a feeling, how more
DCs and F-Steps take new space and what this will cause for
overall capacity. Overall I'm not "sure", what the OEMs will
mainly do, just because they went mostly low-end with AMD.
We only have to look at the low ASP, which is an indikator
that Sempronsales are extremly strong, way more percentage
wise than at Intel. We will see a shift to more DCs, of course
, which is still not a huge volume driver and we will see
more notebooks units flowing through the quarter. But I have
a tough view in the desktop space whether the A64 could gain
share against the Sempron?
A few comments again to the existing inventory. When we look
at the new TLs and the AM2, we are now seeing, that many
modells are packaged in the first quarter (front), so AMD had
build inventory -> of course. They did the same with the new
Opterons. So, we get new Sempron modells + AM2 here, we get
new A64 SC with AM2, we get new X2 EE and we get new X2 with
AM2, we get new Semprons with S1 and we get new Opterons with
12xx. All will launch and ship in Q2. Some modells are low-
volume, of course, but mostly the desktop ones are in my mind
huge volume already in inventory. So when we look at the end
of Q1 (inventory) and subtract the new modells from above,
the existing VERY LOW inventory gets WAY smaller - hard to
quantify, but I would subtract MINIMUM 50Mio. from finished
goods, which would lead us to around 80Mio. Take my math from
above and we come to conservative potential Revs. from these
to 150-170Mio. which is 170/1300 = 0,13 *13 = 1,7 weeks of
inventory for "old" (754/939/940) parts. How do I call this?
I would say, produce a part, package this and instantly sell
it in the channel ... thats it. I would be pleased, if AMD
had sold more units in Q1, but the situation would make it
now even worse.
Whats the assumptions? Without FAB36 AMD shipped more in Q4+Q1
as they could produce (not that much), but volume was higher
than producing (IMHO!). With that its reasonable that we
haven't seen an extremly big DC push with 150-200mm^2 90nm
DIEs. So how the ramp will go on is, again IMHO (of course),
only a function of how many units FAB36 could put on the
street, which we will see from Q2 onwards, because they ship-
ped from the very end of Q1 - so Q2 is the first FULL Quarter
with FAB36 shippings. I hope I could make more volume assum-
ptions as I worked through the numbers in Excel and I will
try to predict how fast AMD could ramp up DC later.
Edit:
A very low inventory isn't the worst which we could imagine, but
I'm thinking further. Just look at the following quarters.
I expect Q3 to be around 15% volume surplus QoQ and Q4 could
also end up again 5-10% QoQ. When we take the around 13-13,5
Mio. number from Q1 as a base and assume 13Mio. for Q2, it
would mean additional NEEDED units of around 1,5-2Mio. for
Q3 and again around 1Mio. for Q4. Thats the low side, so
gives us 15,5-16Mio. at the end of the year. Way more DCs,
bigger F-Steps, its easy to see, why FAB36 is needed NOW!!!
BUGGI |
