Samsung is likely to introduce new artificial intelligence chips for Android smartphones in the second half of 2018, The Korea Herald's Investor reported Thursday, citing industry sources close to the Seoul-based tech giant. While the company has already developed several iterations of AI chips, i.e. neural processing units meant to be embedded into traditional SoCs powering contemporary handsets, it's currently unclear whether it intends to aggressively push for their commercialization in the coming months or if the current creations are meant to be a proof of concept for its higher-ups, as well as a basis for truly commercial applications. Samsung's AI efforts are said to be much clearer on the server front, with its NPUs designed for servers now reportedly being close to commercialization.
The original equipment manufacturer had to ramp up its AI chip endeavors last fall after both Apple and Huawei debuted major advancements with their mobile NPUs integrated into the chips powering their latest flagships. The upcoming Galaxy S9 lineup is expected to feature such a solution as well, or at least the variants powered by Qualcomm's Snapdragon 845 are given how that silicon comes with its very own NPU. The Exynos 9810 expected to power the international models of the Galaxy S9 and Galaxy S9 Plus is yet to be detailed in its entirety but no NPU has been mention in regards to the latest chip from Samsung's foundry business. The Galaxy Note 9 that's set to be announced next summer should boast an NPU regardless of its SoC of choice, some insiders believe.
The main advantage of having an NPU on a mobile system-on-chip is allowing it to store and process large amounts of data in an efficient manner without relying on a cloud service. Such solutions are sometimes described as brain-like seeing how their architecture is unlike the one used by traditional CPUs and instead operates in a manner similar to a human brain, boasting tens of millions of artificial neurons. Despite its Bixby software push in 2017, Samsung hasn't been as dedicated to AI hardware like NPUs until Huawei and Apple started implementing them into their commercial offerings but is understood to have caught up to them in a matter of months as far as raw processing power is concerned.