Thursday, May 10, 2007

Intel Is Doomed!

I constantly get accused of being slanted/biased/pro/fan in favor of AMD. So, I'm going to try to see the world through the eyes of an unabashed AMD enthusiast. This is not an easy point of view to maintain though because Common Sense keeps getting in the way.

Intel is an evil company!

CS: There is no doubt that Intel has taken advantage of market conditions and even stepped a bit over legal boundaries in their struggle with AMD. However, you should be aware that Hallmark and Coke have also done this. Hallmark reps have been known to threaten to pull entire Hallmark displays if the stores didn't stop selling competitor's cards. Coke was well known for creating exclusive contracts not only with customers but with vending machine manufacturers as well. I can't exactly give Pepsi a pass on this either since, when Coke settled with them out of court, Pepsi readily joined in a dual monopoly targeting smaller bottlers like Royal Palm, Faygo, Shasta, and Canada Dry.

Just wait though, Intel will end up having to pay Billions to AMD in compensation when the lawsuit is resolved.

CS: The truth is that this is a very difficult argument to make. Not only was Northwood in 2002 a much more polished design than Williamette in 2001 but it successfully used the 130nm process. Some of AMD's revenue drop in 2002 should be related to its slower transition to 130nm. Also, it isn't clear that AMD's numbers represented genuine gains in marketshare. There is good evidence that AMD's share bounced up and down during 2001 and that the high in Q4 2001 was just temporary. There were also changes in purchasing patterns caused by the preparation for Y2K and reactions to the attacks on the World Trade Center. After you separate out temporary and external factors you are probably looking at a potential damage award of between $200 and $500 Million. Intel could pay this out of available cash without batting an eye.

But look at all the companies that used to be Intel exclusive and are now selling AMD!

CS: Yes but these companies go where the money is. For example, although Gateway and Dell now sell AMD, their product offerings are still dominated by Intel. HP has 50/50 Intel and AMD but HP has always sold AMD. In contrast, Tiger Direct who sold mostly AMD from 2002-2005 now offers mostly Intel. Likewise, Sun is now going to offer some Intel products. There is no evidence of any major backlash against Intel, merely some flexibility in vendors' product offerings.

Yes, but it is so liberating to know that SOI is less susceptible to overheating than bulk silicon like Intel uses.

CS: True but Intel countered this by putting in thermal throttling. There has been no evidence of heat related failure of Intel chips under normal conditions so this isn't much of an advantage for AMD.

Yes, but Intel's chipset monopoly is now broken.

CS: It is true that AMD's ability to offer factory branded chipsets and to target neglected areas will be more advantageous over time. However, it isn't clear how much time this will take. Nor is Intel likely to suddenly forget how to make its own chipsets. Intel will adapt to changing market conditions but is still likely to hold onto the biggest share of the integrated graphics market.

Intel was stupid to keep pushing its overheating hyperpipelined netburst chips.

CS: The Northwood Pentium 4's were the undisputed performance leaders during all of 2002 and most of 2003 with no problems of overheating. The P4 design however did not prove readily adaptable to 64 bit extensions or to higher IPC's. It seems likely that Intel could have avoided the overheating problems of Prescott if it had been able to leave out the 64 bit extensions. Remember that K8 was designed around a 64 bit architecture while Prescott had them poorly grafted onto the Northwood design. Finally, it should be noted that it took Intel about the same amount of time to replace Prescott as it took AMD to replace K5.

Intel has alienated a lot of customers.

CS: This is true however after seeing a big revenue drop in 2006 Intel has become much nicer to both potential and existing customers. This is no longer likely to be a big factor.

K10 will destroy Intel!

CS: If AMD experienced the same growth that it saw during 2004 and 2005 when K8 was leading it would end up around 30% marketshare. It is both foolish and impractical to suggest that AMD's total manufacturing capacity could make up for Intel's share. A transfer of 7% to AMD is possible but a transfer of, say, 15% is not.

But, Intel's expenses are much higher than AMD's; Intel couldn't survive a significant drop in revenue.

CS: Intel employs nearly 100,000 people, runs multiple design teams, and owns several world class 300mm FABs. If Intel actually did lose significant volume share there would be nothing that would prevent them from selling off one FAB and reducing the employee count by the same amount to cut expenses. It is simply not conceivable that any amount of volume share that AMD could take even with a third FAB could cause Intel to go bankrupt or drop below 50% volume share. One therefore has to conclude that Intel's position is secure through at least 2012. However, with this much lead time Intel could probably almost painlessly just cut back on future FAB upgrades to 32nm and 22nm and phase out older facilities without having to do anything so drastic as selling off a FAB.

AMD already has SOI and will move to immersion scanning before Intel. AMD's FAB was rated #1 in the world for the past five years and AMD uses the much more sophisticated APM production control software. Clearly Intel is trailing AMD.

CS: AMD's SOI has not yet created any real gap in performance between AMD and Intel. And, although AMD is moving to immersion first it is doubtful that this will slow Intel much when it does decide to move to immersion for 32nm production. And, while it is true that FAB 30 was rated #1 it is likely that Intel's FABs were also among the world's best. Again, AMD has not yet been able to show any real advantage in terms of more sophisticated production, higher yields, or significantly faster product cycles. There is also reason to believe that if APM does become an advantage for AMD that Intel will create its own version. It is wrong to view either AMD or Intel as somehow technologically backward; both companies represent what is likely the most leading edge and sophisticated chip production in the world. It could also be argued that it is far easier to have the top rated FAB when all of your efforts go into a single FAB versus many production FABs as Intel has. Finally, although AMD and Intel do approach manufacturing differently there has yet to be demonstrated any significant difference in results.

Intel will take a big loss on its inventory which is piling up.

CS: Intel's inventory has stayed pretty steady for the last four quarters. Unfortunately we don't know what Intel's inventory mix is or how much it should be. If we base inventories on cpu revenues alone then it is a bit higher than AMD's however if we base inventory on total revenue then it is actually lower than AMD's. And, it is likely during the last four quarters that older inventory has been replaced with newer.

Intel has fallen way behind in HPC.

CS: While it is flattering to have your chips powering the world's fastest supercomputer these sales are relatively small volume. It is far more important in terms of revenue to see HPC wins translate into server sales. It remains to be seen if this will happen for AMD during 2007.

I prefer to support the underdog.

CS: Personal preferences aside, AMD has to produce competitive products to keep making money and expanding share. Very few buyers argue for spending more money and getting less. Being the underdog is not likely to be enough.


Axel said...

The first somewhat balanced entry I've seen yet on this blog.

Another one would be:

Intel has not managed to sell off its excess Netburst inventory and will eventually have to write it off and dump it all into a landfill!

CS: Perhaps a large writeoff is inevitable. But unless AMD manages to re-capture and dominate the high end, Intel's C2D and Penryn lineup will keep ASPs up and they can absorb the writeoff.

Scientia from AMDZone said...

I have to laugh at that. My arguments in this post are very similar to the last one.

I don't think a writeoff will be necessary. Intel's inventories are the same as Q4 but some of this is probably conversion of older inventory to newer.

Also, your statement about ASP is incorrect. AMD's quad core will take the top spot. However, ask yourself if you would rather have the top quad ranking or the second ranking with 5X the volume. In other words, AMD will recapture the the high end but won't have the volume to dominate it. AMD is looking at 1 Million quad chips before year's end.

Penryn is not really a factor in 2007 as it will be far less than 1% of Intel's volume. I think we can guess that in the 4th quarter it will hit servers (where the markup is high) and not much else. On the other hand Intel should have a lot of Clovertown volume which will be able to run quad socket with Caneland. This isn't exactly bad lineup.