In last month's edition of InsideDSP, Jeff Bier's editorial discussed the advantages of (and potential issues with) designing an embedded system around an application processor originally developed for smartphones, tablets, and other high volume devices. To wit, at the beginning of his writeup, Bier mentioned that Texas Instruments had recently stated its intentions to de-emphasize application processors for smartphones and tablets, instead refocusing its OMAP ARM-based SoCs on embedded applications.
In striving to better understand the company's motivations and plans, BDTI recently interviewed TI's Greg Delagi, senior vice president and general manager of the embedded processing business unit, where he leads the company's microcontroller, digital signal processing, OMAP and wireless connectivity businesses. Delagi is a 30-year Texas Instruments veteran who has been actively involved in numerous strategy shifts like this one over his career. Subsequent to BDTI's conversation with Delagi, Texas Instruments announced the pending elimination of approximately 1,700 jobs worldwide, commensurate with the OMAP strategy shift.
At the start of our call, Delagi expressed no shortage of surprise at the extensive amount of press coverage that his commentary at a late September investor briefing had generated. Those comments, he felt, were only the latest in a series of public disclosures on the matter, the first occurring at the GSM Mobile World Congress show in March, followed by a subsequent briefing in May at another investor conference in New York City. Delagi also made a point of noting that this was a long-term product line discussion; the company's ARM Cortex-A9-based OMAP4 products are still selling in large volumes to approximately 150 customers, and the newer ARM Cortex-A15-based OMAP5430 and OMAP5432 will similarly ship in high volumes in traditional smartphone and tablet systems.
However, the increasing dominance of Apple and Samsung in these applications, coupled with both companies’ predominant reliance on internally developed SoCs, is at the core of Texas Instruments’ strategic recalibration. IDC's Q2 2012 worldwide smartphone market share forecast, for example, gave the combination of Samsung (30%) and Apple (17%) nearly half of the market. And more recent IDC estimates give Apple 50% of the tablet computer market, with Samsung taking an additional 18% slice. Delagi bluntly noted that TI invests approximately $250M USD per year on the OMAP product line and cannot continue to do so in the face of diminishing returns on that investment over time.
Somewhat surprisingly, Delagi claimed that the company's several-year-back disengagement from the cellular baseband business played "absolutely zero" role in its more recent decision to defocus from consumer electronics design wins. He rightly pointed out that Apple and Samsung, for example, are both strong advocates of the benefit of a split (i.e. separate application and cellular baseband processor) system architecture, versus leveraging a single-chip SoC from a company like Qualcomm or (in the near future, by virtue of last year's Icera acquisition) NVIDIA. “You double your R&D investment, he noted, "and it's commoditized. There are too many people chasing that business to zero. There are no switching costs for customers. And there is no sustainable differentiation."
In addition to the formidable expense of new chip designs, fresh iterations of which smartphone and tablet customers expect TI to roll out every six months, application-tailored application processors also demand other costly capabilities, such as exotic packaging and specialized memories. In resetting the company's product line to target the broader embedded market, for example, Delagi made a point of noting that current OMAP products require comparatively expensive mobile DDR SDRAM, whereas many embedded designs would prefer more cost-effective conventional DDR memory.
Delagi mentioned that several OMAP5 derivative products currently in development were more embedded-compatible from a feature set standpoint, although he declined to provide specification specifics or availability timeframes. He also declined to address whether or not the company's embedded embrace would extend backwards to the OMAP4 product line, noting instead that many embedded design opportunities were adequately serviced today by the company's ARM Cortex-A8-based, embedded-focused Sitara SoCs.
CE customer engagements are also resource-intensive, in each case demanding a team of Texas Instruments personnel to support them. Delagi commented that this situation had historically and fundamentally limited the company's ability to scale OMAP across numerous applications and a diversity of customers. However, he also admitted that the requisite culture shift within the OMAP business unit was probably the largest challenge facing the company in implementing the CE-to-embedded strategy change. With other groups such as Sitara as case studies, TI has clearly exemplified the ability to address embedded market customer needs at the corporate level. But the OMAP group will need to appropriately retool and realign itself to match.
Examples of unique embedded customer support needs mentioned in Bier's editorial from last month, and discussed in more detail with Delagi during the briefing, include a robustly trained and staffed worldwide distributor network, the ability to provide small product sample volumes, extended life cycle (multi-year, versus multi-month) production availability, and extended temperature operation. Delagi forecasts, in response, that you'll see a slower roll-out (and phase-out) of new OMAP products from TI than in the past, and that in exchange the company will invest more in documentation, common software platforms and other horizontal market support needs.
Despite the difficulties, Delagi is bullish on the embedded opportunity, which he feels is fueled by the emergence of (and customer embrace of) robust high-level operating systems such as Android. "I want a dual-core processor that runs Ice Cream Sandwich [Android 4.0]" is, according to Delagi, an increasingly common embedded customer request, although any particular customer may not use every SoC feature that a smartphone or tablet might use. To that point, however, Bier noted that TI will need to enable user-friendly access to these SoC features, whether through APIs to TI-developed software libraries or by enabling lower-level access through the toolsets.
Taking OMAP to the broad embedded market certainly has challenges. But Delagi is confident that TI is slowly but surely building the competencies to be able to address existing OMAP limitations. If this were a company with no embedded market experience to date, at least some skepticism of the company's chances might be warranted. Given TI's longstanding presence (and success) in the broad horizontal embedded market with Sitara and other product lines, though, the company's OMAP repositioning bet is one that you shouldn't automatically bet against. But the exact form that this repositioning will take is yet to be determined; will TI literally take OMAP to the broad market? Or given OMAP's complexity, will the company instead end up re-using various pieces of OMAP technology to enhance Sitara and other existing embedded products?