Agreement will take a long time to come to fruition
However, it is clear from the article that this is long-term agreement that will take considerable time to bring financial benefits to Samsung and involves potential risks and delays given the technological advancements involved in Samsung's new 14nm FinFET process. A research report quoted in the article notes that Samsung's 14nm process for external customers will only start to contribute in the second half of 2015, illustrating that mass production for customers such as Apple is still some way off.
Given the high amount of investment required to bring its advanced 14nm FinFET process to high volume production, it is obvious that Samsung needs clear commitments from third parties that will utilize the process with significant volume. In the article, apart from Apple, Qualcomm and NVIDIA are also mentioned as potential customers.
TSMC continues to benefit from its existing leadership position
Meanwhile, TSMC, which is manufacturing Apple's current generation A8 and A8X and may also produce additional next-generation Apple chips before the Samsung agreement bears fruit, will continue to to benefit from supplying the vast majority of Apple's chips for some time, starting from its current 20nm production. Even though TSMC's 16nm FinFET process technology (evolving from 16FF to 16FF+) is less advanced in terms of feature size than Samsung's 14nm process, its more evolutionary nature most likely will allow TSMC to ramp to stable production much earlier and more smoothly (see also my earlier article), with Apple interested in using it, just as most other major SoC players. In fact Apple was reported to have a signed a multi-year deal with TSMC last year including not only TSMC's 20nm process but also its 16nm FinFET technology.
Capacity constraints at TSMC may disappear
There is also potential for the current shortage of wafer capacity at TSMC to become much less acute in the future, which would diminish one of the main motivations for Apple and other companies to invest in Samsung's production capacity. DigiTimes has already reported that chip orders for Apple's current latest generation devices have reached their peak and are already declining, which may result in a less constrained capacity environment at TSMC in the near term.
Transition to more cost-effective SoCs for performance-oriented segment may reduce wafer requirements
Another potential reason for less wafer capacity requirements at TSMC in the near future is the transition in the performance-oriented segment of the smartphone SoC market from large die size, relatively costly chip platforms such as Qualcomm's Snapdragon 800 series (more or less comparable to Apple's SoCs in terms of cost) to more economical Cortex-A53-based SoCs such as Snapdragon 610, Snapdragon 615 and MediaTek's next generation of chips.
Currently, Qualcomm supplies the Snapdragon 800 series to a large part of the market (even for segments for which its performance is greater than required) because it is less affected by the cost of the chip because of the large margins it achieves through its patent royalty schemes, as well as a performance gap in its product lines of SoCs with integrated modem. However, even for Qualcomm transitioning to less costly chips has the potential to increase profits, and in the past Qualcomm has already benefited from a somewhat similar transition to ARM's economical Cortex-A7 processor core for cost-sensitive platforms.
The new Cortex-A53-based SoCs have the potential to provide performance that is adequate for a large part of the performance segment, while being power efficient and due to the smaller die size may significantly reduce the total number of wafers required for performance-oriented smartphone SoCs.
Sources: Korea Times article, Blog article from 3 October
Updated December 5, 2014.
No comments:
Post a Comment