Huawei announces Kirin 970 smartphone chipset bound for the upcoming Mate 10, with a dedicated neural processing unit
Huawei's Kirin 970 smartphone chipset features mobile artificial intelligence with a dedicated neural processing unit, an 8-core CPU, and a 12-core GPU, along with 5.5 million transistors.
Context & Ripple Effects
Huawei is announcing the Kirin 970 weeks before the Mate 10 launch, and the headline feature is structural rather than incremental: a dedicated neural processing unit sitting alongside the 8-core CPU and 12-core GPU, moving AI workloads off the general-purpose cores. It is the first time Huawei has carved out silicon specifically for on-device inference.
That choice turned out to be the through-line of Huawei's entire chip decade. The NPU carried into the 7nm Kirin 980 the following year, and after US restrictions cut Huawei off from advanced chips, the same Kirin line re-emerged built on domestic SMIC process tech in the Kirin 9030 — making this announcement the origin point of both the mobile-NPU race and Huawei's self-reliant silicon strategy.
First-order effects
- Mate 10 buyers get AI tasks handled by dedicated hardware rather than shared CPU/GPU cycles, giving Huawei a concrete performance-and-battery talking point for its next flagship against Apple and Samsung.
- Huawei's silicon roadmap now leads its phone marketing: the chipset announcement precedes the device, signaling that SoC differentiation — not camera or screen specs alone — is the pitch.
Second-order effects
- Rival handset makers face pressure to match the dedicated-AI-silicon framing, since a phone marketed around an NPU forces competitors to explain what their own chips do with machine-learning workloads.
- Chip suppliers and IP vendors see demand shift toward neural-engine blocks as standard SoC components, changing what a flagship mobile processor is expected to contain.
Third-order effects
- If the pattern holds, every premium smartphone SoC ships with a dedicated AI accelerator as table stakes — a structure Huawei itself validated by carrying the NPU forward through the Kirin 980 to the domestically produced Kirin 9030.
- The announcement also foreshadows the strategic split the US restrictions later forced: Huawei's early investment in owning its own AI-capable silicon gave it a design lineage to rebuild on when cut off from advanced US chips, though whether logic-stacking and SMIC nodes can sustain parity remains genuinely unresolved.
The trend: Smartphone SoCs are absorbing dedicated neural processing units as standard architecture, and Huawei's NPU-first Kirin roadmap became the template that later anchored its push toward chip self-sufficiency.