TI reminds me a lot of Wind River in that it's trying to push a general-purpose platform for others to use, and that it's primary competitors aren't other companies pushing their own general-purpose platform (i.e. Lucent Microelectronics and Motorola), but rather companies that develop their own DSPs in-house for proprietary solutions. Qualcomm, with its QDSP core, is a perfect example of this. Broadcom, which uses proprietary DSPs for its cable modem, DSL, and VoIP solutions, is another such example.
Here TI's problem is the same as Wind River's: as the progresss of Moore's Law allows embedded products to have greater and greater amounts of functionality, companies often find working with off-the-shelf solutions unacceptable for their needs, and chose instead to develop proprietary platforms aimed at specialized situations. Consider the following excerpt from a Techweb article regarding Element-14, a DSL chipset startup that Broadcom recently bought out:
techweb.com
"We have gone further towards the DSP engine than
some of our competitors because we want Firepath to be
generic in the communications space for us. But really
the technology is a mix of hardware and software."
But Knowles makes the point that Element-14 won't be
selling standalone Firepath or licensing the core,
which can have a major impact on implementation
concerns.
"The thing with Texas Instruments is that it has built
processors which will be programmed by others, so they
mustn't change the ground rules."
In TI's defense, like Wind River, the company does possess a competitive advantage by means of the Windows-like network effect it's generated for its developer base. The company's "investments" in university semiconductor teaching programs, so as to train engineering majors to work with TI's DSPs, have especially been brilliant moves on their part; and in the case of the handset manufacturers, while there's technically nothing there to stop them from switching DSP vendors, if they were to chose to do so, they'd have to train their engineers to work with not merely an entirely new processor core, but an entirely new System-on-Chip (SoC) reference design containing memory blocks, CPU cores (ARM is generally preferred), and other such products. With this kept in mind, the importance of the support currently being given for TI's OMAP platform for 3G handsets can't be underscored.
Of course, a handset manufacturer could circumvent this dependence on TI by means of purchasing a full-fledged, end-to-end chipset solution from a third party, a solution that includes the necessary alogorithms and protocol stacks related to a given technology (i.e. GSM, CDMA, etc.). In other words, something like what Qualcomm offers. If this is done, then the lion's share of the programming/development required on the part of the handset manufacturer is done away with, as it's now in a position very akin to that of a horizontally integrated operation.
With this said, one might not be surprised by TI's decision to buy out Dot Wireless and enter the W-CDMA chipset solution market so as to cover this tangent as well. However, I'm not sure that this move's a smart one on their part. For starters, TI's core expertise has always been in general-purpose semiconductor design related to the creation of DSPs and SoC solutions. Working with wireless protocols has been something that they've let their customers handle on their own. Coming up to speed on technologies like CDMA and GSM might not prove easy. Furthermore, this move of TI's is bound to create a great deal of channel conflict with third party 3G chipset
developers who might be interested in licensing DSP cores or SOC reference designs from the company.
Lastly, there's the risk involved with regards to lowering the switching costs related to moving away from TI. For example, if Nokia were to chose to move away from doing in-house 3G chipset work via TI's OMAP platform, and opted instead to buy full-fledged chipset solutions from the company, in the future, having now lessened or abandoned its prior in-house development work, they might be far more willing to buy an end-to-end chipset solution from another third party such as Qualcomm. Thus by gaining some additional short-term revenue as a result of selling full-fledged solution, TI would also potentially be paving the way for the ties related to its customer relationships to be weakened.
With all of this having been said, given the support that's been shown for the OMAP platform by Nokia, Ericsson, and Sony, in terms of providing the building blocks needed to develop 3G chipsets, TI currently appears to be in a fairly good position in the 3G chipset market for the time being. It'll be interesting to see to what extent handset manufacturers will continue to use TI's "building blocks" to create proprietary chipset solutions, and to what extent, given the pace at which handset design cycles are shortening, and the rate at which design/engineering demands for data-oriented handsets are increasing, they'll chose to go for off-the-shelf, end-to-end solutions from third parties, TI or otherwise.
Eric
PS - A few good articles on TI and DSPs in general.
DSP efforts by European vendors:
techweb.com
DSPs and wireless streaming video:
techweb.com
Issues related to SoC reference designs:
techweb.com
Competing 3G processor designs:
techweb.com
Competing DSP and SoC platforms from TI and Lucent:
techweb.com |
