Going to AMD's Q3 Earnings Report

AMD's Management promised a radical transformation of the company in Q1's CC. Did it happen? -- No.

Management also talked and talked about 'Asset Light'. Is any of that talk materialized in policies or strategies? -- No.

According to the numbers in Google Finance, AMD gained in equally spaced periods of 13 weeks, 89 millions in Q2 '06 (net income), when Athlon X2 was king and Intel didn't have but pitiful Netbursts and 32-bit Yonahs to compete. The inertia helped yet another quarter, 136 millions in Q3. By that time, ATI was already acquired, and then it followed losses of 574 millions in Q4 ($1.08 per share), 611 millions ($1.11 per share) in Q1 '07, and $600 millions ($1.09 p/share) in Q2. According to AMD's investor relations numbers, the losses in Graphics and Consumer Electronics accelerated in the same period.

Just like it was predicted in this blog, Barcelona was launched and disappointed in many ways: Late to launch, small numbers, slow clocks; even having the single die/Processor to Processor interconnect advantages it is barely at par in instructions per clock to Multi Chip Modules/Front Side Bus-choked Core processors. These products are so mediocre that in fact the top of the line dual core Opterons have far higher absolute and relative prices, according to AMD's own price lists. Furthermore, there is no high speed 65nm processor model in AMD's entire lineup, the consumer quadcore products are being pushed to 2008, and AMD dared to market triple-cores. That is, there is a mountain of indications that AMDs 65nm process sucks big time. And there doesn't seem to be any enthusiasm for Barcelonas among OEMs, so much for the often repeated talk of eagerness of customers. I invite the dedicated researcher to read the transcript of Q2's Conference Call, there you would see the 'alternative reality' of Mr. Meyer, President, and Dr. Ruiz, Chairman and CEO, who reiterated how wonderful the yields at 65nm are, and how extraordinarily good the 65nm process is that it allows the company to close Fab30 (that is, to cut the necessary expenses to convert Fab30 into Fab38).

On the other side, Intel has leaped forward in semiconductors with the double-punch advancements of High-K (dielectric constant) gate insulators (that allow thicker gate insulators than 5 molecules of Silicon dioxide while allowing the same electric field effect [insulator thickness is important because at these quantum physics scales the electrons are probabilistically simultaneously at both sides of the insulator, so there are leakage currents that grow very fast the thinner the insulator is ]) and metal gates, that increase the density of electrons in the gate and favors faster switching [ see IEEE Spectrum article ]. Also, Intel already announced products that will eliminate the disadvantages of the Front Side Bus. Since AMD's Direct Connect Architecture never was taken full advantage of, I think it doesn't take a very sophisticated P2P/Integrated Memory Controller architecture for a performance jump, that is, just a reasonable implementation of Processor to Processor interconnects will be enough to significantly improve the competitiveness of Intel's processors.

In the graphics department, AMD has lost more where it hurts most, in the mobile segment, and elsewhere, giving nVidia a feast of profits and market share gains.

AMD won't bring competitive products in the foreseeable future, and this is not speculation, just plain official roadmaps. Intel is zooming, with 45nm, high-k/metal gate, processor to processor interconnects, integrated memory controller.

Thus, AMD is conceding territory at everything high-end and going back to its unprofitable existence of value/volume player, the only difference being that now it is a supplier of all the major OEMs, and thus its market is less flexible and the losses are larger. The only way to hold on to market share is by selling cheap, but since the processor is turning less and less a cost component of a computer and as much a determinant of performance, at the low end of the performance spectrum it is less and less attractive to save a few dollars in a processor, that is, the low end of the market is becoming more price inelastic, and that means that price cuts to stimulate demand hurt more total revenues.

I am very eager to see how the losses of AMD have decelerated in this quarter, but what we are seeing is an inviable company: It can not produce more because it just doesn't have the products to drive market share expansion, it can not produce less because then it worsens its competitive position being weaker at silicon process development and manufacturing because less production heightens the economies of scale disadvantage in R&D; the company has rotated not just once, but TWICE! the debt it took to acquire ATI: 2.2 giga dollars in April and 1.5 giga dollars in August, that is, about half its present market capitalization value. It is important to notice that in this milenium, AMD has never earned more than $0.37 per share in a year, and lost ten times that, $3 per share, this year. I treated these themes in many posts, including 'Long year', January, 'Disaster', 'Catastrophe', and 'The ATI acquisition' back in April, and especially 'Does AMD know what it is doing?' in June, the historical perspective is particularly hilarious -- I am getting back on track, making predictions that actually turn true!.

Fortunately, the market for computers is strong, so, AMD may have experienced a break this quarter. How good would that have been? I can not say what is going to happen with the stock price from tomorrow on, not even if I knew exactly how much AMD is losing; what I know is that with patience, increasing bearish positions on AMD, the reality that this company is not viable long term will impose over any temporary stock price strength. And this may be the only way, patient and bearish, because Wall Street has every chance to know intimate financial details of AMD given that AMD will have to go cap in hand to Wall Street to borrow money to finance its inviable existence, and that will confer lots of opportunities to Wall Street to manipulate this stock price with obscenity and impunity; and on the other hand Intel will hold the keys for AMD's annihilation in the form of pricing and roadmaps; thus, there is no way the non-informed of privileged insider data could successfully day trade this stock.

Although not financial and not technical, there are very positive news coming from AMD, though. AMD decided to release specifications to enable Open Source drivers for its video cards. I celebrate this decision, but beyond my personal preferences, I think this is the correct strategy: Ally with entities with stakes in innovation, and not with status-quo maintainers like Microsoft and DRM, but that would require another post.