From maelstroms to black holes, whirlpools make compelling images. A powerful force draws all around it into a singularity. The swirling disk of motion masks the nature of its peril until it is too late, paradoxically increasing in force as it pulls in more victims.
Now the semiconductor industry confronts a catastrophic vortex of its own: the cell phone handset.
The force at the center of the handset business, the maw poised to suck in a good portion of the industry, is convergence the very phenomenon that has enabled incredible advances in system-on-chip integration, abetted by consumers' taste for all-in-one gadgets.
The handset is rapidly consuming every other aspect of mobile consumer electronics: PDAs, cameras, GPS receivers, MP3 players, DVD players and game consoles. In the process, the SoC companies and intellectual-property (IP) providers that had planned to make a living in each of those areas will be drawn in for the most part, to their doom.
"This trend absolutely appears to be happening," said Ludo Deferm, vice president of business development at the Interuniversity Microelectronics Center, or IMEC (Leuven, Belgium). "Already, in Europe, you can see that all the kids have to have cameras on their cell phones. And chip architects are planning for the addition of other functions as rapidly as possible."
Batman, without the belt
Convergence is being driven by a simple consumer want: "Don't make me carry a bagful of toys when one will do." Two electronic gizmos in a package are better than one, as long as the form factor doesn't get out of control or the user interface become inscrutable. This is what's happening with second- or third-generation PDA/cell phone combinations, which are rapidly spreading through the ranks of professional users.
Now the same force is drawing in less likely devices, the digital camera being the obvious example. Cameras have become a nearly ubiquitous handset feature in Japan and Europe; they're even catching on in North America. What started out as a novelty has turned into a serious competitor to standalone digital still cameras, with standalone-quality resolution and operating features coming by the Christmas shopping season.
A similar pattern will occur in other areas. The driving force behind demand is not the added functionality of having the digital device connected to the cellular network, but the lures of novelty and convenience. It's about having your iPod, your Nikon, your Gameboy, your Palm and your movie previews all clipped to your belt, without your looking like Batman.
If convergence is driven by consumer convenience, it is enabled by the similarity in architectures of the devices that are converging. In every case, the standalone device drawn into the black hole of convergence is, at heart, a system-on-chip with a control CPU, often a specialized DSP core, keyboard input and both graphic and audio output.
The complexity of the various blocks changes from application to application for example, a game console has a complex graphics-rendering engine, a camera a complex image-compression processor. But the architectures are similar.
This similarity, on the surface, should be great news for the advocates of reconfigurable-computing SoCs and programmable processor arrays. But it is not.
In fact, the suppliers of reconfigurable arrays have watched in dismay as the similarities they could exploit have done nothing to slow the development of fixed-function, application-specific SoCs for each of the aforementioned applications. A considerable investment has been channeled into those developments, both by the venture community and by established IC vendors. The situation is similar to the network processor glut of 2000 or the 802.11 glut of last year. This time, however, the limiting factor for supplier opportunity is not a saturated market, but a maelstrom.
One salient characteristic of the black hole until Stephen Hawking recently revised his views on the matter is that nothing comes out again, not even radiation. The notion holds up better when applied to the handset-chip business: Nothing has ever come out, not even profits. There are exceptions, but for most IC vendors, the handset business is a perfect black hole.
The exceptions involve either a legally imposed monopoly or a genuine shortage of capacity that holds up prices. The former applies to ARM Ltd., with its iron grip on the ARM instruction-set architecture, and Qualcomm Corp., with its similarly fierce hold on the principles of CDMA. Those vendors have orbited the handset hole like giant satellites, drawing increasing kinetic energy from the accretion disk without actually being drawn in.
The third notable exception is flash memory. As more features go into a handset, it requires disproportionately more flash. And global capacity for the types of flash chips used to store programs hasn't kept up with the combined increase in unit demand. So there has been some good money made in the flash business.
Yet it is too early to tell whether the flash vendors will remain free of the black hole or be gradually pulled in, as DRAM makers were drawn into the PC market years ago.
The vast majority of SoC vendors that target mobile consumer electronics and the library vendors, low-power design tool vendors and foundry operators that support them will find that they have no such ambiguous position. They will be drawn by the irresistibly attractive reality of the cellular industry: Handsets are free.
This is the way it works. Every new handset generation comes with a compelling new set of features. Each is subsidized by service providers to get it on the market. But each feature set quickly triggers a market share war among service providers, causing them to offer the handset for bubble-pack pricing or to simply bundle it with a service contract and give it away. The only money for the service provider is in services not in hardware.
The death spiral
This exerts incredible pricing pressure on handset makers, both to innovate and to ruthlessly eliminate their own margins.
Handset OEMs, in turn, negotiate the obliteration of profit margins from any of their suppliers that face competition and among SoC vendors and IP providers, that would be everyone. So the companies that developed the IP come up with nothing.
"Perhaps the only people who stand to make money at the end of the day may be the service providers," IMEC's Deferm said.
At the same time, the very cellular handsets that are bringing them nothing are destroying the SoC vendors' original markets. Free handsets with 2- to 3-megapixel cameras, good MP3 players, decent organizers and good videogames will decimate each of those standalone markets. The only survivors among mobile consumer devices will be high-end niches temporarily beyond the reach of the handset's electronics professional digital cameras, for example.
The erosion of standalone digital devices will not be complete or instantaneous, said Jordan Selburn, principal analyst at iSuppli Corp. "If you ask whether it is in the numbers today, the answer is clearly no," Selburn said. "Take digital cameras as an example: The cameras in today's handsets resolve a few hundred kpixels that's not going to make anyone throw away their digital camera.
"But we are seeing introductions this year of multimegapixel cameras. That may well make someone think twice about buying a low-end digital camera separately."
Selburn sees a similar case for MP3 players. "The disk-based players like the iPod are not initially vulnerable," he said. "But the flash-based players would certainly be in question if there was a significant MP3 capability in the cell phone. And if you add network-based storage, so that users could download tunes or listen to them in real-time, that raises questions about the disk-based systems as well."
Selburn cautioned that battery life would be an issue with all of these possibilities. But he also pointed out that the numbers involved are huge.
"You have to think of the cell phone market on the scale of the human population of Earth," he said. "Maybe one-tenth of the people in the world could get a new handset this year." That puts annual volumes in the hundreds of millions of units.
Tracing that back to the meaning for the semiconductor industry, if 10 percent of those handset sales in 2005 displaced one mobile consumer device, each, that would be perhaps 50 million units of low-end cameras, flash MP3 players and midrange PDAs that wouldn't get sold. Estimating the silicon content of any one of those devices at between $15 and $25, that would suggest a loss to the semiconductor industry of around $1 billion per year in expected revenue.
Just like a star drawn into the accretion disk of a black hole, the SoC provider that had in its sights a rapidly growing consumer device market has instead found itself torn from its intended market, rapidly stripped of its semiconductor product line, reduced to IP vendor status and then driven quickly to zero or negative margins.
There's no escape, and there's no return. And unlike the myriad consumer devices that are being drawn into the black hole, there's no button to push when you don't like the picture and want to start over.