Friday, September 15, 2006

AMD 3Q06: battle won

Intel's BK is now inevitable. Intel's BK is a result of colossal strategic and tactical blunders by Intel management and their total inability to adapt to a changing climate. They will go extinct just like the dinosaurs. The time of AMD64 has come.

The situation today is totally different from 2002. The differences are

0) AMD now has gathered all the greatest minds on computer and system architecture. Chief architects of Opteron, Alpha, Power, UltraSparc, MIPS, Itanium and PA-RISC are all in AMD camp.

1) AMD now has the most advanced FAB process through its partnership with IBM and its APM 3.0 FAB automation technology excels.

2) AMD firmly controls the high end server with 8P 16 core Opteron technology at an affordable price.

3) 95% of Intel's inventory and 70% of Intel's production are more than 20% slower than AMD64 and Conroe and thus become totally obsolete.

4) AMD is doubling its capacity by the end of 2006. There is an oversupply of CPUs. AMD's capacity will be able to supply over 80% of the market by the end of 2008.

5) AMD's average cost per CPU is at least $40 lower than Intel's.

6) The AMD anti-trust lawsuit is tying Intel's hands. Intel found it impossible to impose exclusion agreements on OEMs.

7) The 65nm K8s will be here any time soon.

8) The next gen 65nm Rev G core with 60% integer performance increase and 200% FP performance increase is just six months away.

9) AMD+ATI will be a killer enterprise.

10) The next gen Bulldozer mobile paltform based on the same Rev G core will shatter competition.

11) The OLPC project will equip the children of the world with the $100 laptops, which will establish AMD Inside as the standard for furture generations.

12) The OEMs all see (0)-(11) above and are embracing AMD like there is no tomorrow. Former Intel strongholds such as DELL and IBM have suffered enough and are now introducing their AMD64 business and consumer offerings. Traditional AMD alliances such as HP, SUN, Lenovo and Gateway are also expanding their offerings.

The DELL-AMD alliance will expand AMD's footprint in the commercial client market and beyond. I projected that AMD will exit 2006 with 40% market share(run rate), leaving Intel at 57%.


I pointed out to Intel execs in 2005 that their share loss is inevitable due to FAB36 ramp and AMD's advantage with Direct Connect Architcture. In view of that inevitability, I suggested Intel to cut production and hike ASP to maintain constant revenue.

But, Intel miscalculated.

From the 2Q06 conference call transcript, Intel truly believed their loss of unit share in the previous quarters was due to the shortage of low end 865 chipsets. Andy Bryant actually blamed himself for making the suggestion to cut 865 chipset production. To remedy the situation, Intel actually ramped production of those chipsets and their companion CPUs, and now Intel is in a deep pile of unwanted legacy chips.

This was a tragic miscalculation -- as Intel execs refused to recognize that it was AMD64's strength that pushed AMD up.

But that's not all. Intel made the next strategic blunder which will prove fatal. It again failed to adapt to the changing situation, instead of thinking anew, they resorted the tried and old tactics. "Opereation Crush", "Operation checkmate", those old Andy Grove campaigns were recycled and reused in an AMD64 era. Intel's whole calculation was based on the assumption that C0nroe will take the high end, Pentum D will take the mainstream, while Pentium 4 takes the low end, thus pushing Athlon 64 out of the market. However, AMD has forseen this coming (even I have predicted this), and they quickly changed their battle plan: Athlon 64 X2 becomes the mainstream. The result is now Athlon 64 X2 takes the sweet spot, and Pentium D becomes totally unwanted. In the low end, Athlon 64 and Sempron frag Pentium 4 and Celerons with more performance and less power consumption. Now, if you consider that legacy Netburst and IA32 chips are still the majority of Intel's inventory and production, you know Intel is in a terrible situation.

With FAB36 and Chartered FAb7 cranking like crazy and DELL's insatiable appetite for low price and high performance desktop CPUs, AMD will rule and Intel will be pushed out.

AMD has won the decisive battle in 3Q06. History will record this as the true turning point in a war that will soon end with Intel's final capitulation.

24 Comments:

Anonymous Anonymous said...

"5) AMD's average cost per CPU is at least $40 lower than Intel's."

Link please:)

"7) The 65nm K8s will be here any time soon."

So?

"10) The next gen Bulldozer mobile paltform based on the same Rev G core will shatter competition."

I thought Bulldozer would be radically different, in the same way as Pentium M vs Pentium 4.

"1Former Intel strongholds such as DELL and IBM have suffered enough and are now introducing their AMD64 business and consumer offerings. Traditiona"

IBM and Dell also have little retail penetration.

11:55 AM, September 15, 2006  
Blogger Pop Catalin said...

After AMD ramps 65 nm Intel will take a steep plunge! But I don't think it will go BK.

In the era of value based computing wich has just arived ($500 PC's have 51% of the market and rising compared to 37% last year) Intel is ramping up expensive CPU's and it does a poor job to reduce it's overall costs.

Untill Conroe replaces P4 event in the lowest of low end Intel will suffer from the P4 mistake! If it gets any worse for Intel, it will financiary colapse under it's own weight, wich BTW, isn't entirely not posible when AMD launches K8L at high capacity.

12:22 PM, September 15, 2006  
Blogger Sharikou, Ph. D said...

After AMD ramps 65 nm Intel will take a steep plunge! But I don't think it will go BK.


The math is simple. Intel needs to generate $7 billion revenue to break even. I project Intel's revenue will fall to $6.0 billion and below, resulting a $1.0 billion loss per Q.

12:34 PM, September 15, 2006  
Blogger Pop Catalin said...

"The math is simple. Intel needs to generate $7 billion revenue to break even."

Yes I know!, there's no more room for Intel to loose market share and also no more room to lower it's prices.Wich is a shame because it IS and will be loosing market share due to AMD's new deals including the one with DELL, and higher than ever AMD capacity.And I also believe that by the end of the year Intel will be red zone.

But when a company wants to layoff 5000 emploees it will start by laying of 1000 then 2000 more and then the rest! Intel started with 10.000+ so I think it won't bk because it has plenty of room for layofs and business sale. It can make it for another thow years without going bk and by then Intel won't be tied to the P4 mistake anymore and will hopefuly manage to drasticaly reduce it's costs. Failure to reduce cu cut the costs down will indeed lead to bankroupcy for Intel.

12:49 PM, September 15, 2006  
Anonymous Anonymous said...

http://explodingunicorn.blogspot.com/2006/09/college-is-waste-of-time-money-and.html
http://img232.imageshack.us/img232/193/omgze1.jpg
Muhahaha!

2:43 PM, September 15, 2006  
Blogger Sharikou, Ph. D said...

I think it won't bk because it has plenty of room for layofs and business sale.

Well. Intel could avoid BK by laying off 80% of its workers, but I alsom ointed out laying off 80K workers costs $1.6 billion. Intel can't afford that.

2:45 PM, September 15, 2006  
Anonymous The Sheepshagger said...

0) AMD now has gathered all the greatest minds on computer and system architecture. Chief arhcitects of Opteron, Alpha, Power, UltraSparc, MIPS, Itanium and PA-RISC are all in AMD camp.

Plausible. Although I do recall you hating on the Itanium something fierce. So really, can you call that a win? Only time will tell on what these great chief architects can come up with.

1) AMD now has the most advanced FAB process through its partnership with IBM and its APM 3.0 FAB automation technology excels.

Opinion or fact? I’d like to see something from sematech.org that can qualify that statement. Additionally, it almost sounds like AMD is riding IBM’s coat tails instead of being innovative. But that’s just my opinion.

2) AMD firmly controls the high end server with 8P 16 core Opteron technology at an affordable price.

Remains to be seen. I say we just wait and see what the Q3’06 reports for both companies say about that as I’m sure each has a counter opinion.

3) 95% of Intel's inventory and 70% of Intel's production are more than 20% slower than AMD64 and Conroe and thus become totally obsolete.

I think you are hung up on Q2’06 information. Again, I’d say we’re a few weeks away from knowing for sure both how Intel has handled the inventory issue and how widely accepted the Core 2 Due is (Merom and Conroe) and how that translates to Intel’s production.

4) AMD is doubling its capacity by the end of 2006. There is an oversupply of CPUs. AMD's capacity will be able to supply over 80% of the market by the end of 2008.

Doubling one’s capacity in the midst of an oversupply affects everyone’s bottom line, meaning that neither company can enjoy the high ASP’s. I’m really curious how you really think AMD can supply 80% of the market by the end of 2008 if they will still only have 2 fabs running material and are using a foundry like Chartered. Which market are you specifically taking about? Do you mean server, desktop, mobile? Or are you attempt to lump all of those together and saying that AMD can supply 80% of the world market? Or are you sampling stating that AMD will supply 80% of one geographical market such as Japan, or Asia-Pacific, or Europe, or The Americas? I’d also like to you to back up this 80% figure in whatever market with a non-AMD or non-INQ link. I’m not saying you’re wrong, I’m just saying it sounds like a lofty pipedream when you consider 80% is a figure that Intel has needed 9 to 10 fabs to hit.

5) AMD's average cost per CPU is at least $40 lower than Intel's.

This one I’d really like to see researched properly. There is a difference between ASP’s in desktops, mobiles, and servers. I won’t dispute that might be lower, but $40 lower seems like a stretch. Hector Ruiz has gone on record that AMD has to have a $100 ASP to break even. That been said, if AMD is making a profit then they must be selling chips for more than $100/cpu. How much more is what is really in question. I have a follow up question that I’ll save for later should you reply to this one.

6) The AMD anti-trust lawsuit is tying Intel's hands. Intel found it impossible to impose exclusion agreements on OEMs.

Hard call to say one way or another. This one is in your pro-AMD favor as what you’ve overlooked is that AMD has made inroads with product and OEM’s have jumped onboard. I dare say you didn’t pat yourself on the back for this one. I’m surprised.

7) The 65nm K8s will be here any time soon.

Still waiting… and waiting… and waiting…

8) The next gen 65nm Rev G core with 60% integer performance increase and 200% FP performance increase is just six months away.

Still waiting… and waiting… and waiting…

9) AMD+ATI will be a killer enterprise.

Remains to be seen… Still waiting… and waiting… and waiting…

10) The next gen Bulldozer mobile paltform based on the same Rev G core will shatter competition.

Remains to be seen… Still waiting… and waiting… and waiting…

11) The OLPC project will equip the children of the world with the $100 laptops, which will establish AMD Inside as the standard for furture generations.

I’m all for the OLPC project. I’m curious, tho… exactly how much profit is there for a CPU company when the entire laptop is $100? It is a platform after-all and the ATI acquisition may help. However, a cpu and a graphics chip do not a laptop make. My guess is a true OLPC computer with have banks of flash microdrives. I currently have a 4 GB microdrive for my new DSLR camera. Imagine the technology or demand for drives 10x-100x that size being just a few years away. It is almost a shame that AMD spun off Spansion.

12) The OEMs all see (0)-(11) above and are embracing AMD like there is no tomorrow. Former Intel strongholds such as DELL and IBM have suffered enough and are now introducing their AMD64 business and consumer offerings. Traditional AMD alliances such as HP, SUN, Lenovo and Gateway are also expanding their offerings.

I would really call IBM an Intel stronghold. When was the last time you say a real IBM-PC? In servers they’ve been split for many years between their own chips, Intel’s chips and AMD’s chips. I really wouldn’t call this newsworthy.

Lenovo purchased the IBM desktop and mobile products group within the last year or two and I think they came out of the chute at nearly 50% mix of cpu vendor. I’ll give you HP and Sun, but really… where is Gateway on the list of OEM’s? No offense, but they were a solid #3 before they switched to AMD.

3:01 PM, September 15, 2006  
Anonymous Anonymous said...

Will 65nm rev g's be compatable with existing AM2 mobo's? It would be a nice upgrade path for me.

3:14 PM, September 15, 2006  
Blogger S said...

If it is true that X2 is now mainstream, then AMD's average cost must have nearly doubled with X2 being dual core. Sharikou, you have said before, based on Q2 numbers that AMDs Avg cost is around $40. For Q3, the Avg cost would be definitely >$60-70. Assuming ASP remains same, that will be a big decline in Gross margins. If interest cost on the $2 billion and odd debt kicks in in Q3, chances are that AMD will slip back into Red - where it has been for most of its recent history.

Note that I haven't even considered the margin shrink due to Dell business.

Time you stop worrying abt Intel BKing and think abt how to shore up AMD.

Regards to your other points, agreed that AMD is doing a lot in product development. But you are too naive to assume that Intel will not succeed in its efforts at bringing in product enhancements.

BTW, do u really believe Dell, IBM, HP etc would let AMD to become the only source of X86 CPUs. Remember AMD exists today only because IBM did not want Intel to be sole supplier of CPUs for their PCs.

3:32 PM, September 15, 2006  
Anonymous Anonymous said...

And I predict that GM is would go BK in next 4 Quarters....


Sharikou, what planet do you live on? Do they have chinese food?

3:36 PM, September 15, 2006  
Blogger Sharikou, Ph. D said...

If it is true that X2 is now mainstream, then AMD's average cost must have nearly doubled with X2 being dual core.

No, Silicon is just a small part of the cost. Packaging, testing add to the cost. Expect AMD's average cost to stay the same as it increases production scale.

The $2 billion loan is no problem. At 8.5% interest rate, annual interest is $170 million, or 43 million per quarter. AMD needs to sell 20,000 more Opterons to Google to make that profit. The DELL business will generate about $150 million profit/quarter.

Also, as Intel goes BK, AMD will be able to sell at a better price as the premium brand. Only suckers will buy Intel.

Also, AMD will soon get $15 billion damages payment from Intel. Basically, expect AMD to own Intel's FABs sans the workers.

6:20 PM, September 15, 2006  
Blogger 180 Sharikou said...

If it is true that X2 is now mainstream, then AMD's average cost must have nearly doubled with X2 being dual core.

No, Silicon is just a small part of the cost. Packaging, testing add to the cost. Expect AMD's average cost to stay the same as it increases production scale.


Interesting how silicon becomes a small part of the cost here. But on Conroe it will be more expensive to manufacture for Intel because of the larger cache - even though the die size is significantly smaller.

http://www.tomshardware.com/2006/07/14/core2_duo_knocks_out_athlon_64/page4.html

In fact, even smaller than Athlon. And please don't tell me AMD will be on 65 nm soon as they have recanted the report on the Inquirer themselves saying first orders ship by end of year and substantial volume by mid 2007. And how much benefit they get is only your conjecture right now.
http://www.theinquirer.net/default.aspx?article=34282

http://sharikou180.blogspot.com/
(A balanced point of view)

8:25 PM, September 15, 2006  
Blogger Sharikou, Ph. D said...

Interesting how silicon becomes a small part of the cost here. But on Conroe it will be more expensive to manufacture for Intel because of the larger cache - even though the die size is significantly smaller.


No. Intel cost is higher because its yield is only a fraction of AMD's yield. I analysed in one of classic articles. Also, Intel has large overhead and is much less efficient than AMD in every way.

8:29 PM, September 15, 2006  
Anonymous Dr. Yield. PhD, MBA said...

Intel cost is higher because its yield is only a fraction of AMD's yield. I analysed in one of classic articles.

Hogwash. You proved nothing. AMD may indeed have higher yield than Intel, but nothing you have done proves it or not. I posted a detailed analysis of why K8 AM2 die size meant it needed to have a substantially better line yield to have lower incremental die cost and you could not disprove it. Hell, you even admitted that you didn't read it, either because your superior intellect couldn't decode it, or because it didn't support your world view.

I repeat- AMD may indeed have higher line yields than Intel. I won't claim they don't- but you have posted nothing of substantive value to prove it. Post provable facts, with real math (not your vaunted logic) backing it up. You just claimed overhead isn't the primary factor, so leave that out of your equation. I eagerly await your evidence, Dr. Sharikou.

10:05 PM, September 15, 2006  
Blogger core2dude said...


Intel cost is higher because its yield is only a fraction of AMD's yield.


Intel's yields are typically in high 80s. They can't be a fraction of AMD's yields unless AMD yields are 160%+

10:17 PM, September 15, 2006  
Blogger aravind said...

I’m just saying it sounds like a lofty pipedream when you consider 80% is a figure that Intel has needed 9 to 10 fabs to hit.

they make mobo chipsets, graphic chips and embedded ones called Xscale,

the actual no of fabs that produce Pentium or Conroes may be three or four

10:24 PM, September 15, 2006  
Anonymous Anonymous said...

If Tom's logic is a broadcast network they would be ABC, if sharikou's logic was a broadcast network it would be PAX.

You never "analyze" anything. You apply suspect "facts" with faulty logic that draws from unreasonable inferences.

The only thing classic about your articles is that they are classicly silly.

11:14 PM, September 15, 2006  
Anonymous wise investor said...

core2dude said:

"Intel cost is higher because its yield is only a fraction of AMD's yield.


Intel's yields are typically in high 80s. They can't be a fraction of AMD's yields unless AMD yields are 160%+ "

Dude... 1/100 is a fraction, 84/96 is also a fraction. Even 99/1 is a fraction. I am saddened, to see someone who could affort a C2D but failed to manifest a adequate high school mathematic education...

12:42 AM, September 16, 2006  
Anonymous The Sheepshagger said...

No. Intel cost is higher because its yield is only a fraction of AMD's yield. I analysed in one of classic articles. Also, Intel has large overhead and is much less efficient than AMD in every way.

I'd really love to know where you guys get your information on AMD's yields. That being said you can add 18+ percentage points on to the speculation of Intel only reaching 80% yield on Conroe.

80% + 18% = Conroe at 98% and change.

12:55 AM, September 16, 2006  
Blogger S said...

If silicon was only a small part of the total cost, yields should not matter. The fact that chip industry strives so much to reach high yields speaks against this.

Cost of raw silicon may be cheap. But the silicon we r talking about here is the processed silicon which has spent weeks inside fab and has taken up valuble machine time. It will cost significant amount of money.

When you fabricate 2-core cpu instead on single core cpu on a wafer, yields, in terms of no. of CPUs per wafer is down 50% straightaway. Can't you see that?

1:00 AM, September 16, 2006  
Blogger hyc said...

Seeing text like this, the only thing that comes to mind is "BK: the home of the whopper."

(Burger King Whoppers, and whopper as in a tall tale, greatly exaggerated story, or outright lie. Interpret as you wish...)

2:36 AM, September 16, 2006  
Anonymous Graham said...

I could spend time "fisking" every one of the points Sharikou made in this post. However, I do have better things to do with my time. I would just like to point out that there is not a single link to any of these assertions except to his prior posts. Therefore this list of "facts" is just a list of opinions of an angry little man.

11:36 AM, September 16, 2006  
Anonymous Anonymous said...

Yo PhD pretender..

Yields seem to be a topic right now..

Back it up, tell me ballpark what you claim AMD yields are and why you claim or belive INTEL yields are...

Numbers.. remember you're a PhD.. you use data.. or is your PHD earned in virtually with virtual data and conclusions..

Back up your claims

Not a pretender

12:09 PM, September 16, 2006  
Anonymous Anonymous said...

80% accurate? IF you guess intel BK right, Sharikou write a book now.

1:57 PM, September 25, 2006  

Post a Comment

Links to this post:

Create a Link

<< Home