The business of modern handheld chipsetsĀ 

The economics affecting most companies behind the chipsets in modern handheld devices are pretty much identical; they're nearly all fabless semiconductor companies and have mostly the same manufacturing partners. There are two exceptions to the rule, however, which are Texas Instruments and Samsung. TI has its own fabs, but also outsources some work to TSMC, while Samsung is part of the IBM-Chartered-Infineon-Samsung process technology alliance, and does all chip manufacturing in-house.

So, besides for these two companies, everyone has access to the same manufacturing processes at pretty much the same time, and with pretty much the same prices. Thus, excluding factors such as company politics, the relative competitiveness is nearly exclusively based on three factors: how efficient their architecture is, whether their products and/or platforms are ready in the right timeframes, and how balanced their feature set is (rich enough to have some obvious marketing and customer appeal, yet simple enough not to cost too much for their target market).

Arguably, these factors are much the same for most players in the semiconductor industry, excluding those focusing on intellectual property (IP). But each sector of the industry has some potentially substantially different dynamics, regardless. In the PC CPU industry, for example, there is very little possible differentiation in terms of feature sets. So things are mostly based around architecture efficiency in terms of performance, and power consumption.

If you look at the mainboard chipset market instead, performance is only a differentiating factor for the enthusiast community, and feature-set (-per-dollar) is what tends to matter most. The economics are quite different in both cases, too; for example, for x86 CPUs, both AMD and Intel have their own fabs and different schedules for process adoption (VIA, which is a much smaller player, outsources manufacturing to TSMC). With mainboard chipsets, some companies such as Intel have their own fabs, but most don't. This can often be an advantage, but it sometimes it's not - for example when Intel's chipset division got capacity constrained in early 2006, with no easy way to fix the problem.

Now, coming back to the handheld industry, what's relatively interesting there is that nearly all possible and imaginable factors contribute to a company's overall competitiveness within their markets. Architecture efficiency in terms of both performance and power matter a lot, but so does the overall feature-set, the level of integration and having the product ready soon enough to hit your potential partners' design windows. In some parts of the industry, such as for the iPod, things are very cyclical too; there's a new revision coming out every year or so, and you've got to compete all over again every time.

But there is at least another quite unique factor to this market: software platforms. Obviously, nearly every single silicon chip needs to be able to interface with other parts of the device that will ultimately be sold to the actual users. But to take the example of the graphics card market, the software there corresponds to the drivers that interface with the computer's operating system. NVIDIA and AMD's graphics division don't decide what APIs they need to support; Microsoft does. And they don't have to deliver more than that, either.

In the handheld space, the software is a much bigger part of the complete solution you need to sell to companies like Apple, Motorola or Nokia. And you also need to support a variety of operating systems with different APIs. This will hopefully soon be standardized (via the Khronos Group) for media-centric interfaces, but isn't so just yet. Thus, no matter how great your hardware might be, they aren't going to be very interested if you tell them your Windows CE platform only includes, say, WMV and MPEG-4 decoding, and they'll have to write the codecs for other video formats themselves. But then, if your software is equally good as that of your competitors (and that is a very big if), how do you decide what your hardware specs need to be?