Set a favorite micro-grid, DIGITIMES Research released a report estimated that by 2018 global smartphone application processor (Application Processor; AP) shipments will grow 5.1%, of which Qualcomm is still in the leading position, followed by MediaTek; However, Apple, Samsung, Huawei's HiSilicon and other home-made chips infiltrated the overall smart phone penetration is expected next year, homemade chips accounted for the overall smart phone AP shipments will more than 30% .DIGITIMESResearch pointed out that in addition to a small amount of export Samsung Chip, Apple, Hisilicon chips are not self-external supply, however, the three manufacturers for the global smartphone shipments of the top three companies, their products equipped with self-made chips, the sale of smart phones on the AP market has squeezed with the industry concentration Increase and continue to expand, forecast high-pass shipments next year, fear of decline of 1.2%.
MediaTek's target market and self-developed chip market less overlap, and the introduction of competitive mainstream products by downstream manufacturers to adopt, DIGITIMES Research forecast next year's shipments increased slightly by 3.6%, but also by the smart phone shipments From the architecture perspective, DIGITIMES Research believes that the Cortex-A53 is still the mainstream, followed by the independent architecture of the chip; from the process of observation, 28 nm is still the mainstream, to maintain a certain standard , The flagship smart phone AP will be 7 nanometer, 10 nanometer, 12 nanometer upgrade, 14 nanometer and 16 nanometer proportion drops; from the base band (Baseband) technology observation, 2G and 3G smartphone AP has its rigid demand into niche market DIGITIMES Research said that with the advent of artificial intelligence (AI), edge computing has become the new direction of choice for smart phone AP manufacturers through deep learning in smartphones, with the combination of graphics processors (GPUs), digital signal processors DSP) or heterogeneous computing smartphone AP on a neural network processor is becoming a trend. Next year, 20.3% of the world's smart phone APs will use heterogeneous computing Li deep learning inference (Inference) even learning task.
