'Significantly lower unit sales'
So far, our interpretation of the data has been fairly straightforward. Obviously, if price cuts happen, margins will go down. No, really! There are very rare exceptions to this rule of course, at least when ignoring short-term factors, but this is certainly not one of them. However, if AMD has been as aggressive as Intel in terms of price cuts - if not more so - shouldn't they have maintained their market share, or even gain some more? Instead, they suffered from significantly lower unit sales, especially so in the channel. Unless they lost 100% of their channel sales in just a few weeks (err... no!), however, you would presume that OEM unit sales are also a fair bit lower than seasonality alone can explain.
Obviously, you would expect Intel's Core 2 Duo to gain momentum over time, and have its design wins progressively capture more and more of the market. But there is another important point from our point of view, and it's linked to the information we presented on the previous page. AMD was still quite strong in the single-core market last quarter. But as the prices of dual-core processors dropped, their addressable market increased and that of single-cores decreased. At the same time, Intel's low-end product line is disadvantaged by its use of the Netburst architecture, and that means worse performance/$ and worse performance/watt compared to AMD's single-core offerings. One good way to visualize that is TG Daily's Price-Performance curves.
So, it seems normal to us that AMD would lose share if they had to address a larger percentage of the market with chips that aren't as competitive as Intel's. Nearly every price cut plays in AMD's disadvantage here, ironically enough. At this point in time, it's probably worth noting that this piece does not take into consideration the upcoming release of AMD's Barcelona core, as it is simply impossible to evaluate its impact on profitability without more precise and reliable performance data. Furthermore, the performance dynamics in the server and client space are quite different, with per-core performance being much more important for desktops and laptops. Finally, it should be noted that Barcelona won't really change things before Q4, if not later. The design wins in the server space won't begin ramping up until Q4 or even next year according to AMD, while the desktop variants will likely lag Barcelona by about a quarter. So it's completely irrelevant for Q2 and Q3.
Back to the present. Another factor that was highlighted in AMD's statement is that the channel was the weakest part of their mix. Many analyst reports in recent months have been pointing out that inventory has been building up in the channel in the last months, but that doesn't seem very coherent with what AMD said during their Q406 conference call:
Henri P. RichardIf I may add, just on the question of inventory, Mark, as you know, we were lower than we would have liked in terms of channel inventory in the third and fourth quarter. We were able to restore that at the very end of the fourth quarter in order to position ourselves in a much better velocity for the first quarter. But at the same time, we also took a prudent approach to the channel inventory in the graphics business, where we think that there’s model improvement that we intend to execute in the first and second quarter of ’07.
So, who's right? The likely explanation is that AMD didn't properly feed the channel for part of 2006, but that during Q107, it definitely had excess inventory which explains why AMD couldn't sell much more to it. So, how come there was excess inventory? Well, part of it is that their fab capacity increased (65nm mass production started) while they might also have lost loyalty in the channel due to not servicing it properly in Q3 and Q4. Another factor is that Intel has obviously been gaining tremendous momentum there. Enthusiasts are flocking to Core 2 Duos and Intel's marketing campaign for the chip is quite aggressive.
Overall, it seems fairly obvious that AMD's original forecasts were significantly too reliant on the channel accepting pretty much all the inventory they could send its way, which was not realistic. Their expectations in terms of margins also make it seem like they weren't expecting the February 12 price cuts to be necessary. Anyway, this transcript excerpt highlights the problem quite nicely:
Robert J. Rivet No, I think you have that backwards. Even though revenue we’re guiding to be lower than fourth quarter, with the improvement of 65-nanometer, more of that output continuing to come on board, servicing the channel as Henri just talked about a minute ago, we believe gross margins will actually improve quarter on quarter. In particular, the microprocessor business. We also believe so in the former ATI businesses too, as we continue to execute the game plans there.
Certainly, increasing the ratio of 65nm CPUs compared to 90nm CPUs would theoretically improve margins compared to Q406, but that answer makes it look like they were expecting the channel to buy those parts rather than OEMs. So if their sales did collapse there during Q1, it might not actually have improved anything because many of those 65nm parts would just be inventory. Remember that OEMs contracts tend to be for specific models, so you can't just drop-in a 65nm chip when the contract originally was for a 90nm one. That's why it made sense for AMD to focus on the channel for 65nm parts, but it looks like that strategy might not have served them well. If that is the case, however, that is at least one thing that they could easily improve upon in Q207.
