Moore's Law is changing the rules of the game again. This time AMD and Intel could both be on the losing team.
AMD has retained JPMorgan Chase to explore options for the struggling microprocessor company, Reuters reports. These options range from an outright purchase of AMD to a spinoff of parts of the company, or a sale of some of the company's patent portfolio. No matter how it goes down, turning to the markets in this way illustrates just how tough business has been for AMD lately. AMD is currently denying Reuters' talk of an outright sale.
AMD is a perpetual underdog, and a scrappy one at that. It recently announced what it calls an ambidextrous approach to its core CPU market, developing both for the x86 instruction set as well as licensing ARM's technology. Sections of AMD's business are reported to be doing fairly well. Embedded chips for gaming and signage and the company's video coprocessors are doing well, analysts say. Its embrace of ARM is so new it's impossible to tell yet if it will bear fruit.
So why hire JPMorgan now?
One reason is AMD's stock price. It's under $2 per share right now, and that leaves AMD with little hope of making acquisitions or starting new large-scale products unless it gets a cash infusion from a patent sale or a spinout. The company hasn't seen its stock price this low since early 2009.
The cash situation is so dire that AMD has shut down research centers and laid off workers during the past few weeks. Announcing its new ARM strategy for servers in this climate will leave most watchers concerned about AMD's ability to execute.
Another reason to explore alternatives now is AMD's also-ran status in pretty much every category in which it competes. While the company has solid technology and can find buyers at the right price, stars like Nvidia outshine AMD in its market specialty. Even when AMD scores big -- like being the chosen CPU in the world's fastest supercomputer, Titan at the Oakridge National Lab -- AMD gets overshadowed by others. In this case, while AMD delivers the CPU, Nvidia provides the coprocessor horsepower with its Kepler compute architecture. Investors want AMD to dominate something.
Who are the likely buyers for AMD? IBM is always a good bet for patents; Intel could be too. You could speculate wildly that if Google thinks Motorola was a good deal, AMD should make at least as much sense.
Apple has made noises in recent weeks (or perhaps more accurately, others have made noise for Apple) about dissatisfaction with Intel. However, the popular thinking is that Apple may prefer to consolidate on its own ARM-based chips rather than continue to develop for two chipsets. A move like that is years away, but there's obvious appeal for Apple. With AMD setting out to gain expertise in ARM, buying AMD could be an attractive option. Hopefully Apple execs will consider the fortunes of others who've gone a similar route -- including HP and Compaq (which brought HP the Alpha chip) and Oracle with Sun (lipstick on a pig, anyone?).
Nvidia or a company like it is another possible buyer. At about five times AMD's size, Nvidia could look to broaden its portfolio and head directly after Intel and ARM. AMD does have an excellent team of engineers.
I don't think that's likely to happen. There's one other market force at work here -- one that is ironically also affecting Intel, whose stock price has also taken a beating during the past eight months. Moore's Law has a way of changing what's important.
For a couple decades it was clock speed -- until continuing that push meant creating chips that ran way too hot to be practical. Then the race was for core density and on-chip cache. After a decade or so, that race has given way to putting more system components on the chip, such as memory and I/O controllers. But even as these things make their way onto the chip, the single core as a unit of processing is often bigger than workloads demand. On the server side, so-called microserver architectures are increasingly interesting. If you can fit (and power) fifty microservers into a space that could have accommodated eight or sixteen cores on one or two chips, perhaps there really isn't a need for virtualization.
On the consumer side, unless you're a gamer, the need for performance has largely been met. CPUs now spend most of their time waiting for the rest of the computing architecture to give them something to do. Power consumption, form factor and unique capabilities on the chip are more interesting. And here's the part that ARM understands that I don't think AMD and Intel do.
Because the gate count on a chip is so high, and performance is so good, device manufacturers in the consumer arena will give up some performance in order to customize systems on a chip to their own liking.
If AMD and Intel want to compete in the consumer market, they'll have to give up control of the ultimate end design to the manufacturers. If they don't, ARM will be all too happy to eat their lunch.
Download the new issue of Must Reads, a compendium of our best recent coverage on IT-as-a-service. It includes articles on cloud computing myths, how to build an IT service catalog, security problems, and more. (Free registration required.)