There's no doubt ARM has done a good job of developing processor cores that score well in terms of performance per watt and per dollar. Its business model is as a kind of shared industry resource on processor architecture. However, as miniaturization continues the processor core becomes only a part of the chip architecture.
ARM has made one major broadening of its IP with the move into graphics processing initiated with the acquisition of Falanx Microsystems AS in June 2006. This has become the successful Mali GPU core range, second in the market place to Imagination's PowerVR line.
But as we move on to an era when system-chips could contain hundreds and thousands of processing elements new software and digital and analog hardware architectures will be required.
Segars needs to help his team consider the trend towards instruction set processors for more precise things beyond graphics rendering. This could be cognition processing close to the image sensor, sound processing, software-defined radio, and even more radical architectures such as research topics as imprecise processing for power saving.
In addition to logic there ARM needs to keep its eye on opportunities in non-volatile memory technology and RF.
At the same time ARM needs to go higher in abstraction – as it already is doing with work in the HSA Foundation – as software frameworks and intermediate languages will determine which architectures get selected for multicore heterogeneous computing.
Segars also needs to think about ARM can get paid for work in software which will become an increasing part of the company’s and its ecosystem's workload.
A very interesting article but I kept wondering about the pictures in the "slide show". How did most of these relate to the paragraph above them? Did I miss something or is my web viewer not working for some reason? Also, I was looking for 10 and not 9 "slides" but did not see that either.
I think you need to read the ARM 2012 Annual Report (see ARM's website) rather than guessing
"We have also seen the penetration of Mali, ARM's graphics processor, grow to more than 20% of Android smartphones and over 50% of Android tablets worldwide"
Mali is growing fast for them in terms of licences and shipment. In 2011 there were 50m units shipped, 2012 120m and the forecast for 2013 is 240m.
Given you are unlikely to know the R&D spend or the licensing and royalty revenues generated I suspect you have very little data on which to define it as a "commercial failure". And that would assume, wrongly, that R&D spend on something like Mali can be judged a commercial failure over such a short space of time. The R&D on Mali, just on the designs already announced, will be generating revenues for ARM for many, many years to come. It's how the business model works
"This has become the successful Mali GPU core range, second in the market place to Imagination's PowerVR line."
Successful...hmmm, they have certainly had some low end chinese design wins, but other than Samsung, they don't have a tier 1, and even samsung has diluted with them preferring imagination for their flashship Exynos octacore. Given that their shipped volume of graphics IP is much less than 1/2 of imaginations, and their published royalty rate is around 4.5c for graphics cores compared with of IMG's average of 25c, I would suggest that todate its graphics dept has been a complete commerical failure relative to the R&D spend.