I've been busy and only had time to get back to the blog today. It's a nice segue to my last post however. AMD pre-announced a big revenue miss in Q1 = revenue coming in at 1.23 billion on a projection of 1.6-1.7 billion:
http://biz.yahoo.com/ap/070410/advanced_micro_devices_outlook.html?.v=1
They also announced a re-structuring with more details to come shortly. In simple English - cost cutting. As I said earlier, AMD cannot go from being a company of 10k people to 16.5k people with lower revenues and no profits. The pre-announcement says a head count freeze is going to happen. My prediction as below continues to be lay-offs are coming:
http://sharikou180.blogspot.com/2007/03/two-strikes-for-amd.html
I'm always sorry to see hard working employees who have to pay the price for poor management decisions with their jobs. Just as Intel's management should have seen heads roll last year during their re-structuring, I think so should AMD's. It's sad to see that they have gone from being on a 40$ high a year or so ago to this position so rapidly.
Anyways, in the next days we shall see how Intel and then AMD's quarters turn out and how much of my predictions in my earlier posts were true or not. I still have this quarter to see Intel begin to re-take market share from AMD.
Subscribe to:
Post Comments (Atom)
11 comments:
Take a crack at the stock price?
Ok - though I will start by saying don't buy or sell on what I write here. I think this can play out in a couple of ways:
Scenario 1:
AMD presents a robust re-structuring plan and Wall Street likes what it sees. Stock holds for now but starts to slide downward with any signals that Q2 will be as bad or even worse than Q1. Which is what I think it will be so with the Q2 results we will see a serious descent heading close to the 10$ mark that's been bandied about earlier.
Scenario 2:
AMD presents their re-structuring and Wall Street thinks it's either not enough...or the changes leave AMD competitively exposed over the medium to long term. At which point the stock heads south faster than Scenario 1.
(P.S. - also remember, they promised savings of about 300-400 million on efficiencies from the ATI acquisition. If the 500 million in the pre-announcement is not incremental to that, they are probably heading to the lower end of the stock price range I'm predicting below.)
In either case, over the next few months I think we're looking at something north of 11.50 but south of 13. The simplistic reasoning for this is that even K10 will not have a sufficient impact on their revenue/financials this year and that's the only bullet they have right now. As a result, they will continue to lose share and be forced to drop prices and hence sacrifice margins to try and sell out all this capacity they've built up. Rising inventory and unused capacity will hit their balance sheet hard.
If they are forced to raise new capital either through a share placement of further loans, then that will place additional pressure on their stock price and I think south of 11.50 is even a possibility.
"If they are forced to raise new capital either through a share placement of further loans, then that will place additional pressure on their stock price and I think south of 11.50 is even a possibility."
I'll just add one thing through AMD's management genius(?), if they issue new stock, ~1/2 will have to go to paying the new debt they took on from the ATI purchase. (Part of the loan terms)
Thus for any new stock offering done to raise cash, AMD will get ~50cents on the dollar - of course the other 1/2 will go to paying down debt which is a good thing for the long term, but means they will have to significantly dilute the stock to raise an appreciable amount of money in this manner. To raise ~500mil in cash would mean $1Bil in stock which would be ~14% stock dilution. And of course 500mil in cash is not that much - they already pushed out capex spending by that much....
Your comments on K10 are dead on - folks who are thinking that this will help them this year (wrt stock price) are loopy, it may give them a short term stock price bump from PR perspective but it won't impact fundamentals until 2008 at best.
On a side not I just read an analysis on Scientia's which is a bit wrong (I think). He assumes a ramp to 30% by end of year - which assuming it goes from 0-30% in 1/2 year equates to 7.5% of total yearly production. Only issue with his analysis is I doubt AMD will be at 30% ACROSS ALL PRODUCTS! While this is very likely achievable in server space, as desktop is not slated until Q4'07 (and who knows about mobile?), I think it unlikely AMD is 30% K10 across the board by end of year.
It'll be interesting to compare Intel's C2D ramp rates (which were called slow by all of the AMD fanboys). They were I believe at ~70% sevrver conversion by year end!
I wouldn't be sad about going from $40 to $13, the value of the stock when it was at $40 was ridiculous and folks who were investing back then weren't doing there homework. When the stock was at $40, the P/E was well over 100, this is not a rational multiple for a company as mature as AMD (perhaps for a startup or growth company this may be reasonable short term). This was just one indicator of several indicators that said the stock was overvalued... I'm sad when I see things like Enron , where to the average investor the publicly available financial data looked reasonable to support investments, but folks who purchased AMD @40 had access to a bunch of data which was screaming overvalued. They were either investing with their hearts or speculating - either way I don't see why you should have remorse for a stock price correcting to what is (now) a fair valuation.
On Barcelona - I think in a minor PR slip, even Hector admitted limited impact this year in the following interview:
http://www.crn.com/hardware/197007770?pgno=2
Sharikou180: I've been busy and only had time to get back to the blog today
I suppose just in time for the AMD bloodbath. It was pretty quiet anyway, except of course AMD's double warning.
As for Barcelona... any semicon proffesional will tell you that its insane to run a quad-core with such a die size at 65nm. AMD knows this which is why it said it only plans to get design wins in 2007 with very limited volume.
Roborat - could you elaborate a little more around the issue of Barcelona die size on 65nm.
thanks...
"Roborat - could you elaborate a little more around the issue of Barcelona die size on 65nm."
While I'm not Roborat... yield is proportional to defect density, which obviously is therefore dependent on die size (the larger the die the greater the chance of a random defect occuring on the die). The quad core version of Barcelona is fairly large, even with 65nm tech - I believe it is bigger than the 90nm versions of the X2. I think I saw ~283mm2, which is quite large. Thus yields would likely be worse than dual core 90nm X2's and that's assuming AMD's 65nm process is yielding as well as their 90nm process (which I would speculate is NOT the case as AMD has been running 90nm for some time and primarily on 200mm while they have been running 65nm in volume for much less time and process improvements are still being implemented)
This is why MCM approach is far better from economic manufacturing perspective - your die are much smaller (as you are only making dual cores on the wafer) and any random defect only effectively takes out 1/2 of a quad core as the dual cores can be matched up for the MCM after sort yield. Also as die size is larger you tend to lose a bit more potential die at the edge of wafer due to geometry issue of filling a round wafer with rectangular die. (More partial die on edge of wafer)
The final thing to factor in is how good the bin splits will work as all 4 of the cores on a native quad core must be over the rated speed; this will likely impact the percentage of high speed bin parts for a given wafer. If you have 3 cores at 2.5GHz and 1 at 2.3GHz, you have yourself a nice 2.3GHz part. With MCM approach you can mmix/match dual cores thus you are limited by slowest speed of 2 cores as opposed to slowest speed of 4 cores.
So while "native quad core" sounds cool from a marketing/PR perspective, it is far less economical and manufacturable than an MCM approach. On top of that there is minimal/no data that I have seen to show the benefit of a "native" design over an MCM approach. I have seen lots of AMDroids saying the FSB is saturated so theoretically native is better, but i have yet to see ANY actual benchmark showing performance delta between a native vs MCM approach.
I do however see the potential for the native approach to be better from a power perspective, due to potential to more independently vary individual core sleep states.
180 Sharikou said...
Roborat - could you elaborate a little more around the issue of Barcelona die size on 65nm.
bigger die, lower yield due to probability that a defect kills the product. The magnitude yield loses is hard to guess unless we know what AMD's D0 curve is. But one thing is universally true, 2 AthlonX2s have a better chance of working than 1 Barcelona just because of defect density.
As for factory output, a die size of ~290mm2 also suggests that it has to sacrifice about 3 AthlonX2's just to make 1 Barcelona chip.
AMD just doesn't have enough capacity to produce Barcelona without sacrifing volume outs unless of course it wishes to screw OEM's this time around. Ofcourse it is manufacturable but ramping Barcelona at 65nm is just plain suicide.
BTW, if you're getting too busy, i suggest leaving your comments unmoderated just to allow the discussion on this blog to continue while you're away.
"I've been busy and only had time to get back to the blog today."
No worries - if you're blog was partly to refute the ridiculousness (is that a word?) of Sharikou's blog, he has been doing that on how own of late.
What's really scary is that Scientia now appears to be heading in the same direction as Sharikou. He keeps getting hung up on this "they can't go above 3 GHz so their 65nm process and/or design must be limited" kick. I think he's now done several articles on this. He conveniently neglects to mention that while there may not be >3 GHz parts there is:
A) no business/market need for this as AMD has nothing to compete against the current 2.93GHz/3.0GHz parts
B) Intel has managed to go from a 2.93 GHz DUAL CORE chip to a 2.93 GHz (or is it 3.0?) QUAD CORE chip - this is a far more significant thermal challenge then going from 3 to 3.2 GHz on dual core.
C) Intel (right or wrong) wants to position the quad core as it's top of the line desktop part, so releasing a higher clocked dual core chip, which will likely outperform a lower clocked quad on many benchmarks, would undercut this strategy.
D) a look at AMD's Barcelona roadmap (albeit speculative ones that have been seen on the web) show no real top end K10 clock speed increases through mid-2008, does this mean IBM's, uh, I mean AMD's, 65nm process is broken too? (The answer is no)
The other interesting thing is Scientia presents his argument/position and then states "We've now seen" this is exactly what Sharikou used to do, as if stating his opinion/argument by using "We" instead of I might make people think this is a generally accepted fact. Much like Sharikou's crazy capacity calculations that he keep referring back to as "we already know...."
Scientia's arguments MAY be right, but they are certainly not conclusive, and he has now devolved into rejecting all other possibilities that don't coincide with his own views.
I understand things look bleak for the AMD fanboys, but these things tend to be cylical so rather than destroy a reputation as being "balanced", they should present the data, offer their opinions (and state them as opinions not as FACTS), take their lumps and wait for the next cycle. FYI the same thing could (and should) have been said for the Intel fans pumping the P4 design.
It will be interesting to see if K10 is that next cycle, unfortunately we will not likely know this until late Q3/early Q4 - unless of course people think the likely selective server K10 benchmarks next week will provide a good idea of DT performance.
The other interesting thing is Scientia presents his argument/position and then states "We've now seen" this is exactly what Sharikou used to do, as if stating his opinion/argument by using "We" instead of I might make people think this is a generally accepted fact. Much like Sharikou's crazy capacity calculations that he keep referring back to as "we already know...."
Typical "I'm not biased" tactic.
Post a Comment