- 0-20%: Unlikely - Lacks credible sources
- 21-40%: Questionable - Some concerns remain
- 41-60%: Plausible - Reasonable evidence
- 61-80%: Probable - Strong evidence
- 81-100%: Highly Likely - Multiple reliable sources
55%
Plausible
Current-generation smartphones are equipped with a small Neural Processing Unit (NPU) that occupies a small portion of the SoC and is required to share the CPU’s and GPU’s power budget. Naturally, this approach constrains the computing capabilities of the NPU, which is why an analyst states that Qualcomm and Chinese firms CXMT and GigaDevice have collaborated to develop a standalone solution that leverages 3D DRAM to handle more demanding AI tasks while operating efficiently.
The advanced NPUs will have a higher memory bandwidth than the LPDDR5X standard, thanks to TSV and Hybrid Bonding
As TF International Securities’ analyst Ming-Chi Kuo points out on Twitter, Qualcomm’s partnership with CXMT was not just to develop custom DRAM, but a dedicated NPU with sufficient memory and AI horsepower to push through those long-running tasks such as real-time video translation or background image generation. The customized 3D DRAM will have 4GB of memory and will boast a bandwidth higher than the LPDDR5X standard, thanks to leveraging advanced packaging technologies such as TSV (Through-Silicon Via) and Hybrid Bonding.
The resulting solution can deliver 40 TOPS (trillion operations per second). Keep in mind that the Snapdragon 8 Elite Gen 5 is marketed by Qualcomm to deliver up to 100 TOPS, but this figure can only be achieved in highly favorable circumstances. A dedicated NPU operating at a consistent 40 TOPS will effectively double the AI computing capabilities of a flagship. This AI accelerator is expected to be found in devices in late 2026 or early 2027 and will be aimed at Chinese smartphones priced between RMB 4,000 - 4,500 ($585 - $660 USD).
While a dedicated NPU sounds truly revolutionary, rising memory costs may discourage various Chinese brands from adopting these chips. Companies like Xiaomi, OPPO, Vivo, Huawei, and others are already brainstorming on ways to reduce DRAM and flash memory expenses to boost their margins, and the addition of a 4GB 3D DRAM NPU will push the bill higher.
Also, Kuo points out that despite the hardware being ready, there isn’t any program out that can properly take advantage of a smartphone’s on-device AI prowess. More importantly, even if this solution arrives to the market, consumers aren’t convinced that they need to fork over a premium for a dedicated NPU, seeing as how it might not be a life-changing feature as one might assume.
News Source: Ming-Chi Kuo
About the author: Omar Sohail is a reporter and analyst for Wccftech's mobile section, specializing in the technology and business of the mobile industry. His expertise lies in the intricate hardware supply chain, covering developments in semiconductor manufacturing, chip lithography, and camera sensor technology.
Follow Wccftech on Google to get more of our news coverage in your feeds.
