Monday, September 11, 2006

Deja Vu - 2002 all over again?

I found this old article from 2002 on "thestreet.com" and it got me thinking:

AMD Sees Likely Loss of Market Share

Is it 2002 all over again. Here we are in the middle of an intense price war initiated by Intel. AMD's next generation is 9 months away. AMD's market share is around the same level as it is today - perhaps a delta of 200 basis points. AMD has profitability but we've already seen them scale back Q2 and Q3 is not looking hot. Now, with the added debt of borrowing to buy ATI, I think it's a matter of time before they recede into the red again. There are two major variables that are different from last time that I can see:

1. AMD now has entry into the high margin server business and is getting a toehold into mobile. This should buffer ASPs.
2. Intel has a kick ass new product line that gives AMD a run for it's money on almost every front except perhaps MP.

So the question is will history repeat itself. Will AMD lose market share and slip back into a loss. Here's my prediction:

- Q3 will be a mess for both Intel and AMD. AMD will begin to lose ASP and market share. Intel will have a bad quarter because they have driven the floor of Pentiums down into the ground while Conroe is still not going to have enough supply or impact to balance this. They'll both be lucky if the hit their Q3 projections. However, AMD will have a relatively better Q3 than Intel compared to their forecast because even as the impact of the July 27 price cuts start to change the market share dynamics, it takes time for such swings to happen.

- Q4 will be better for Intel as Conroe/Merom/Woodcrest help them regain market share (though reports show they are already gaining back in places like US retail and as much as the AMD fans may knock Woodcrest availability, DP server deals are starting to be won). The server stuff will help ASPs. Merom is coming in at the same price points as Yonah so no ASP relief there and Conroe will help a bit. More importantly is the cost savings of moving over to Conroe simply because the die size is smaller. For the disbelievers - here's a handy comparision on Tom's Hardware. In fact, I wouldn't bet against Intel dropping prices and bringing Conroe close to mainstream pricing in Q4 as yields stabilise or at the least introducing a SKU below the 6300. AMD will be in a better position to fight price drops with 65nm production. However, until K8L arrives, AMD is in a difficult position strategically. How low do they go on X2 pricing as they get sandwiched between PDP and Conroe? If they start to lose money, do they give up market share to salvage ASPs or vice versa. My guess is they will do what they can to keep the factories filled to maximise margins and profit in order to generat cash. With the incremental debt to purchase ATI, the more cash they generate the better it is for them. It's kind of funny - AMD is now behaving like the incumbent and Intel like the challenger.

Of course, all this makes the assumption that K8L will blow away Conroe but until we see product nobody knows for sure. I'm in the camp that this will happen but not across every application and benchmark. Then Penryn comes in 2H07 and I'm thinking that Intel may be faster to 45nm than what is their public position. They have been focussing on execution and so far have managed to pull in almost every major product launch by 1 quarter in the last 6 months. For now, my observation continues to be that the microprocessor market is evolving and becoming like the US car market. Highly segmented and different products will excel at different applications. The day of one processor to rule them all is over.

18 comments:

Sharikou, Ph. D. said...

DELL goes AMD full force, that's one thing what is different today.

Intel BK is inevitable.

Read my analysis at http://sharikou.bogspot.com

Scientia from AMDZone said...

You bring up two or three good points but you overlook a lot of things. Most of the points you bring up overlook other factors that diminish or even reverse the effects you talk about. I'm trying to decide if I want to post a long, detailed reply here or create an article for this in my own blog. If you are mistaken like this then probably others are as well.

180 Sharikou said...

Sharikou - you're right, Dell has gone AMD and I should have mentioned this specifically. But I did factor it into my thinking if you read my earlier post. As you point out, Dell is floundering - their screwdriver strategy is no longer a sustaining advantage. They now have accounting problems. Lenovo yesterday said they would give Dell a price war like they've never seen before in China. Their senior management in Asia is defecting to Lenovo.

I think Dell has lost it's way. They do not know what their next step should be. While AMD will gain share through Dell, Dell's impact on the global PC industry will decline over the next couple of years...if their market share losses are any indication this will be sooner than later. In the next 2 years a significant part of AMD & Intel's destiny will be driven by HP. Once again, Intel is placing their resources back into the DIY channel because as a group, they can balance their over-reliance on any one OEM.

Having said that, I hold with my prediction on how Q3/Q4 will play out. Declining revenues/margins for both Intel & AMD in Q3 - AMD less impacted than Intel. Intel starts an upswing in Q4 with a reversal of revenue margins. Ttime will tell if I'm right.

180 Sharikou said...

Scientia - I'd advise you to write your own article and send me a link. I'll pop it up. I'll be happy to have you cure me of my ignorance. However, pls do state your source of info and explain how you arrived at the conclusions you did. Most of your posts are simply you stating your opinion as being contrary to the ones being expressed here. Unfortunately, they lack credibility unless you do the two things I ask above and hence don't necessarily need discussion. For example, if you make a claim "Intel will be at 60% margin by divesting all non computing businesses" then show us how you arrived at that number.

One last thing - please try and look at the big picture which is what I'm trying to address here. It's easy to take pieces of my post out of context and then slice them but again, not helping the overall debate. Make your own prediction on Q3/Q4 and tell us why you think it's right. That's what I'd like to see - that's the intent of this blog.

thanks...

Anonymous said...

Even in MP, AMD has lost performance leadership. Tulsa is ratching up impressive numbers in transaction performance, and database performance compared to AMD, which are important in MP than floating-point and integer. AMD's top-bin MP part is 120W, while Intel's is 150W. But the top Tulsa variant is probably $600-700 cheaper.

Plus corporations can buy a Tulsa platform today with DDR2, not sure about the same with AMD.

Anonymous said...

"Even in MP, AMD has lost performance leadership. Tulsa is ratching up impressive numbers in transaction performance, and database performance compared to AMD, which are important in MP than floating-point and integer. AMD's top-bin MP part is 120W, while Intel's is 150W. But the top Tulsa variant is probably $600-700 cheaper.

Plus corporations can buy a Tulsa platform today with DDR2, not sure about the same with AMD."

Are you that desperate for a heater this winter? Netburst is over, no matter how well it performs.

Anonymous said...

No one should question you because they deep down know your right. All hail the Sharikou's! AMD's rise = Intel's fall!

Is 57% loss in proffits and still loseing enough proof? Intel is dead.

Scientia from AMDZone said...

For example, if you make a claim "Intel will be at 60% margin by divesting all non computing businesses" then show us how you arrived at that number.

Changing the profitability or divesting of the non-computing sections would not actually change the gross margin. However, the higher the gross margin, the more money Intel makes.

In the second quarter of 2006 Intel lost $805 Million in the non-computing areas. Raising the gross margin to 62% would gain an extra $805 Million which would exactly offset the non-computing losses and change their influence to zero.

With a real divestiture, the actual gross margin would not actually go up, but it would be the same margin without a -$805 Million adjustment. In other words, divesting of the non-computing sections would be the same profitability as having the current losses in the non-computing sections and having an overall gross margin of 62%.

Having said that, I hold with my prediction on how Q3/Q4 will play out. Declining revenues/margins for both Intel & AMD in Q3 - AMD less impacted than Intel.

Intel starts an upswing in Q4 with a reversal of revenue margins.


First of all, I don't know if you are talking about total revenue for Intel or just processors. Margins are only given for total revenue so that is clear. And, since AMD no longer has flash memory their income is basically all processors. So, when I talk about revenue and marketshare for Intel I'm only talking about processors, not chipsets or other areas however, the margin figure has to be for total revenue.

I don't know if you are comparing with the same quarter in 2005 or the previous quarter. Basically, the 2nd quarter tends to be less revenue than the 1st quarter. The 3rd quarter tends to be about equal to the 1st quarter and the 4th quarter revenue tends to be the highest. Therefore I would expect market revenues to increase in both the 3rd and 4th quarters from the previous quarter (and this should be true for both companies). In other words, cpu revenues for both AMD and Intel should be Q4 > Q3 > Q2. So, for year over year:

Q3 06
Intel: Lower cpu revenue, marketshare (both revenue and volume), and gross margin than Q3 05.

AMD: Higher revenue, marketshare (both revenue and volume), and gross margin than Q3 05.

Q4 06
Intel: Lower cpu revenue, marketshare (both revenue and volume), and gross margin than Q4 05.

AMD: Higher revenue, marketshare (both revenue and volume), and gross margin than Q4 05.

2006
Intel: Lower revenue, average margin, and marketshare (both revenue and volume) than 2005.

AMD: Higher revenue, average margin, and marketshare (both revenue and volume) than 2005.


The reduced die size of Conroe is helpful to Intel but at 30% volume this isn't enough to make much difference in 2006. This will be more of a factor in 2007. However, in 2007 AMD gets double cost reductions because of increased 300mm volume and 65nm volume. Further, AMD's die size at 65nm is less than Conroe's. Therefore, AMD should see a lot more cost reduction than Intel.

Scientia from AMDZone said...

Even in MP, AMD has lost performance leadership. Tulsa is ratching up impressive numbers in transaction performance, and database performance compared to AMD

I'm not sure where you are getting your database information. Perhaps you could post a link. However, I'm at www.tpc.org and looking at their data right now for transactions.

Typically it is meaningless to talk about total tpc numbers since the more servers cost the better they tend to perform. Higher cost buys faster memory, hard drives, networking, and even external caches. The price/qphH is a better indicator of efficiency which is what most business want. Also, keep in mind that these numbers are divided into four categories of 100G, 300G, 1000G, and 3000G. Also, Sun, IBM, HP, and Dell all recently expanded their AMD server lineup and I don't think these are showing up yet. What there is in these charts is pretty good for AMD.

For 2 cores opteron systems have the highest price/tphH for both 100 and 300.

For 4 cores there is no Opteron equivalent for 100 so Xeon wins by default. However, Opteron is highest in both 300 and 1000.

For 8 cores the Dell Xeon just edges out the HP Opteron for 100. This is impressive for AMD on an HP since the Dells tend to be the best price. However, the Intel HP equivalent is worse than AMD.

AMD's top-bin MP part is 120W, while Intel's is 150W. But the top Tulsa variant is probably $600-700 cheaper.

Plus corporations can buy a Tulsa platform today with DDR2, not sure about the same with AMD.


Currently, the 2.8 and 3.0 Ghz Opterons on socket 940 draw 95 watts. The 2.8 Ghz Opterons on socket F and AM2 which use DDR2 draw 120 watts. These will be back down to 95 watts when 65nm versions are released.

Anonymous said...

You are plain wrong. As described in the article, AMD expected loss of market share and lowered ASP in 2002 due to its losing on high-end desktop. High-end desktop was the only thing that made AMD good money at that time. Things however is very much different today:

1) AMD's price-performance are still competitive up to X2 4600 / E6400 level, which is high-end enough for 95% of consumers.

2) AMD now has a foothold on 4P Opteron server market, the highest profit margin for x86, with better scalability than any Intel offering.

3) Intel is yet to ramp up Core 2 Duo enough to lower its (street) selling price. This was not the case with Northwood in 2002.

4) The Pentium-4 beated Athlon-XP market-wise partly because its much higher clockrate and the public's obsession with megahertz, which is not the case anymore.

Anonymous said...

That ONE MILLION conroes that are said to be available will not shake the market anyway. Some enthusiasts will get one .... .

Until end of year (conroe = 25% of intel) amd's cpus are as well positioned in the market as in the years before.

As for longtime costsaving i would suggest for 10-24 / 7 computing any EE flavor amd cpu. On this i could save arround 30 Euros here in germany per year.

The 20-30 % performance advantage of conroe might have a bigger influence if availability in broad.
This time may arrive in 2Q 07.

But starting with end of 2006 the _ALL_OVER_PERFORMANCE_LEVEL which is STANDARD (the typical 100 $ cpu, which is sold most) will have got a level, so the consumer will say : Pentium4 / Celeron is OUT".

Intel then might sell these to Nokia to put it in a handheld.

At end of q4/06 Conroe lefts 75 % of the market, at end of q1/07 50%, in q2/07 25%.

Even if all conroes were better,

(and amd's share a broad range in the upper performance field -and might increase 200-800 MHZ in the comming 3 Quarters to their cpus as they ever did),

that what is left on the market might not -as said before- nor Must filled up with "Celerons".

That said for sure amd will ON THE ACTUAL (of both) cpus take their portion.

At the cost side we see amd's reduced cache cpu's -X2 3600 /...-
that feeds very well a lot of consumer / and corporate needs.

65 nm too is on the way.

The first time, which could be critical is, when 100% - "conroe-%"
are greater than amd's capacity.

Might be in beginning Q3/07.

Then we have a new Landscape with K8L !

On the corporate side least with dell / today with lenovo we can expect %ages of marketshare being sold.

On the Server side i dont think anyone doubts on further success of amd.

The mobile area is still lean i think, but might get some upgrades (but must not be today).

In the neartime we do not see what comes to bith at DAAMIT.
I guess we will see a ALL in ONE Chip (cpu,gpu,controller).

Conclusion:
Fighted by intel,
got round by round,
will do the next too (into 2-3 Q 07)
Then get CPU enhencement into 65 nm duals and quads
-might be up the Athlon 10000 area (on duals) Otereon/Athlon 20000 (on quads)

regards
-wont post it on crappykous

Scientia from AMDZone said...

Okay, I finished the article with the counter argument to this topic on my blog. I had talked about this before at AMDZone but when you wrote this article I decided that I needed the information in a more permanent format. It is quite long and fairly detailed. It counters pretty much every argument that I've ever heard about the supposed AMD crash in 2002 and potential repeat in 2007.

Anonymous said...

"Currently, the 2.8 and 3.0 Ghz Opterons on socket 940 draw 95 watts. The 2.8 Ghz Opterons on socket F and AM2 which use DDR2 draw 120 watts. These will be back down to 95 watts when 65nm versions are released. "

True, but the 95W Opterons you are talking about are in 2P space. In 4P AMD's 2.6Ghz is at 95W while the 2.8Ghz -3Ghz is at 120W.

Tulsa has some skus that are at 95W as well. These wont have as high performance. But the line with Tulsa is that AMD is not the supreme in MP, and you do have a decent alternative. Plus the price-per performance that Tulsa comes in at is pretty competitive. The highest Tulsa costs $1976 while the highest 4P Opteron costs $2600.
AMD will continue to make margins on Opteron 4P but wont outright gnaw away market share like when Intel was underwater with Paxville.

When 65nm comes by, its a different story, and we dont know anything. It may come back down to 95W, or like the Inq speculates there may be voltage issues with AMD's 65nm process.

180 Sharikou said...

Link to Scientia's analysis for those who want to read it:
http://scientiasblog.blogspot.com/

Scientia - I scanned your post, it's a good analysis and I'll come back and read it in more detail. However, you're talking apples and I'm talking oranges. Market share dips and swings are not necessarily proportionally responsible for revenue ups and downs. You're just talking about revenue. I'm talking about market share. AMD went from 20%+ market share down close to 10% during that down cycle. Of course the entire semiconductor industry was hammered but no connection to AMD's market share loss which benefitted Intel. I repeat, Intel used pricing to win share.

Scientia from AMDZone said...

Market share dips and swings are not necessarily proportionally responsible for revenue ups and downs. You're just talking about revenue. I'm talking about market share. AMD went from 20%+ market share down close to 10% during that down cycle.

Actually, we see the same thing with revenue share. AMD went from a high of 12.5% down to 5.5%. However, as I said in the article the high in Q4 01 and Q1 02 were temporary (caused by Intel's production problems). And, the low in Q3 02 was temporary (caused by AMD's production problems).

The revenue share average was 9.3% for AMD before Intel's decline and 8.2% after its own decline. So, you are suggesting that Intel cost itself $13 Billion in revenues to gain back 1.1% of the revenue share. I would say that wasn't very cost effective if that was Intel's plan.

BTW, you are making the same mistake that most people make in comparing market share with a temporary high of 23%. This is not the same as an actual market shift.

Of course the entire semiconductor industry was hammered but no connection to AMD's market share loss which benefitted Intel. I repeat, Intel used pricing to win share.

Actually, you are just repeating common misconceptions. This is from ChipGeek Jan 2002:

"AMD's Athlon XP cuts mirrored Intel's, with the 1900-1600+ versions of their chips lowered by 14-19% and Durons dipping by 7-17%. "

Intel's price cuts were matched by AMD's. AMD's actual problem was that they had trouble with the 130nm transition to Thoroughbred and fell behind temporarily. This caused AMD to lose about 12% of its net revenue. AMD bounced back just two quarters later in Q1 03 but then fell slightly because of low yields on the 130nm SOI process as they began K8 production. By Q1 04 the yields were back up and AMD had regained half of the lost share. By Q3 04 they had surpassed the 9.3% average before the decline.

Sharikou, Ph. D. said...

Dell is floundering - their screwdriver strategy is no longer a sustaining advantage. They now have accounting problems.

Dell is floundering on revenue and profits because of competition. But its market share is actually growing, it's approaching 20%. AMD capacity is expanding, and DELL eats that capacity. One more AMD sold is one less Intel sold. Simple math. I expect DELL to be 50% AMD by the end of 2007. Intel's unit share will drop to about 50%. Coupled with price crash, Intel revenue will be about $4.8 billion/q instead of $8 billion today. BK inevtiable.

Sharikou, Ph. D. said...

See my analysis

AMD 3Q06: battle won

Anonymous said...

Intel lost more then 60% of its proffit today. AMD keeps riseing. Intel can not compete.