Sunday, October 19, 2008

AMD Breaks New Microprocessor Ground

AMD is feisty. Despite the brutal monopoly tactics used by rival Intel (illegal kickbacks to customers; using the cash accumulated by earlier monopoly pricing to try to drive AMD out of the market; etc.), AMD is not only still in the game, it presents a growing strategic threat to Intel and a growing opportunity for eveyone else in the computer industry.

In the third quarter of 2008 AMD made progress on four major fronts. It showed a degree of creativity and flexibility that we have not seen from Intel since the 1980s. While Intel, which "owns" about 80% of the global microprocessor market, continues to have enormous advantages, it is AMD that is leading. Even if you just bought a PC with an Intel processor, its main advances over processors from 5 years ago - combined 32 and 64 bit capability, multiple processors on a single piece of silicon, and lower energy use - were all invented at AMD and then reluctantly introduced by Intel.

You can read some of this history in articles listed at my main AMD page for investors. I'm going to focus this blog entry on recent stategic developments.

One place AMD has consistently lagged Intel is process technology, which is the technology of creating semiconductor chips with ever smaller surface features. Each new generation allows more gates and hence capacity to be fit on the same size of silicon (or you can put your old design on a smaller piece of silicon, which costs less). It is very, very costly to develop each new generation, both in terms of R&D expense and then in creating the equipment that the new fab (semiconductor factories are called fabs) needs to make the improved chips. Intel made so much money from its microprocessor technology that it was able to keep well ahead of everyone else in the industry, often having as much as a two year lead time. For makers of chips requiring less transistors, this was not a big problem, unless they competed directly with Intel for a particular product.

AMD met this challenge by doing more with less; their processor designs were significantly better than Intels from about 2000 until 2006. But if AMD could be at the same level of process technology at the same time as Intel, it would be in a better position to press its design advantage. During the last decade AMD, working with IBM, has created very sophisticated fabs in Germany, and it has narrowed Intel's lead time. But it is still 6 months to a year or more behind Intel, and Intel has the money to drive the research and investment to keep that lead or even extend it. AMD is now shipping chips using 45 nm technology, but Intel started shipping theirs at the beginning of 2008.

In the past decade many major semiconductor players have emerged that don't have their own fabs. These fabless companies, like NVIDIA and Marvell, have done very well. But the independent foundries that produce chips for them have not been able to keep up with AMD in process technology, much less Intel. You might think that independent equipment makers like Applied Materials and KLA Tencor, would focus on getting ahead of Intel, but most chip makers don't have to put but a fraction of number of gates on a chip as AMD or Intel do, so the economic incentives are not there.

AMD's new stategy is to start allowing other companies to use its foundries when they need the advantage of cutting-edge chip manufacturing. As this happens research and development costs can be spread out over a larger number chip makers. With IBM, AMD and others working together it is possible that the AMD coalition could catch up with the Intel axis in process technology over the next few years.

To accomplish this strategy AMD brought in new investors, basically the nation of Abu Dhabi. AMD is contributing its fabs and some of its intellectual property to the Foundry Company, but Abu Dhabi will be the majority owners, in return for a multi-billion dollar cash infusion. In addition to continuing to expand the AMD plants in Dresden, Germany, AMD will break ground on a new plant in New York State. At a time when most new foundries are being built in places like Malasia and China, you might wonder why New York State. Well, what some people believe is the most sophisticated chip factory in the world, one of IBM's fabs, is in New York. These factories are mainly run by robots. There is no need for a lot of low-paid laborers doing hand work.

Except for the cash infusion, we won't see the results of this strategic turn until 2010 and beyond. But there are the three other fronts I spoke of: graphics, servers, and profits.

AMD aquired NVIDIA rival ATI for about $5 billion back in 2006, just in time to see ATI lose much of its share of the graphics chip market (at about the same time re-designed Intel microprocessor chips caught up with AMD's and Intel lowered prices to make all microprocessor sales unprofitable). I want to repeat what I've said many times, I think the NVIDIA team may be the smartest guys in the tech world. Still, they are AMD rivals in graphics. Recently ATI (now an AMD division) stole a march on NVIDIA with a new line of chips and graphics cards that became very popular very quickly with game enthusiasts this summer. All NVIDIA could do was lower the prices on their own chips while they rushed back to the drawing board (okay, these guys never leave the drawing board, they sleep on the drawing board, but you know what I mean).

NVIDIA still owns the graphics card market, but in Q3 AMDs graphics division revenue jumped to $385 million from $285 million in Q2, a 55% increase in 3 months. More important, AMD/ATI has two advantages over NVIDIA that might come into play in the next 2 years and allow them to eat further into NVIDIA market dominance. AMD leads NVIDIA in process technology, precisely because NVIDIA does not have its own fabs. And AMD can integrate its processor (CPU) development with its graphics development, which NVIDIA can't. This is also a three way race because Intel has long made low-end graphics chips and is trying to upgrade that division in response to the AMD threat. All this will be very good for consumers, who are already seeing inexpensive video cards that are targetted at gamers but that can easily handle high definition video and that can be used as super computers those of us who like to crunch numbers.

In server chips, so important to the Internet world, AMD is shipping its first 45 nm Opterons. Depending on what they are used for, they can be seen as better or worse than Intel's current server chips. AMD chips use less power and are better at math and sending information to and from memory. Intel chips are better at heating your building (actually, they are quite good, having finally incorporate much of the Opteron circa 2002 design improvements). The same advantage you see in Opteron server chips will soon become available in Phenom chips for desktop computers. AMD already has great notebook computer chips, the Puma series.

And for the final strategic advance, we have: profits. While still working towards GAAP and even non-GAAP net income, in Q3 2008 AMD showed operating profits. Much of the industry was concerned that if AMD continued to lose money it would be forced into bankrupcy, and Intel would go back to its high-prices and low performance everyone came to expect in the 1999s. By lowering costs, including by making the Dresden fabs highly efficient, and raising sales, AMD would have been back in the game even without the Abu Dhabi deal.

I own AMD stock and two computers containing AMD processors. I also own stock in Marvell. In the past I owned NVIDIA stock, but do not at present.

See also

No comments:

Post a Comment