Huawei introduced the Kirin 9050 Pro in the Mate XT 2 as its first commercial smartphone processor using LogicFolding. Huawei says the chip contains no US components, but the company has not disclosed the processor’s node or foundry. That leaves a sharp distinction between what the chip does and how completely its supply chain can be described.
What Huawei introduced — and when
Huawei announced the commercial use of LogicFolding in the Mate XT 2 on September 9, 2026. The Kirin 9050 Pro is the phone’s processor, and the device runs HarmonyOS 7.
The chip’s headline claim is not simply that it is fast. Huawei presents it as a way to improve density and efficiency through circuit layout, using vertical integration rather than relying solely on a smaller conventional process node. Richard Yu, chairman of Huawei’s consumer business group, described it at launch as the company’s most powerful Kirin chip.
That claim has a concrete architectural counterpart: LogicFolding distributes selected logic across vertically connected silicon layers and uses hybrid bonding to shorten signal paths. It does not mean that an entire conventional chip die was physically folded in half. The name is catchy; the engineering is more specific.
What LogicFolding changes
In a conventional planar design, the relevant logic sits across a single active layer. Huawei’s approach connects selected logic vertically, bringing parts of the circuit closer together. Shorter signal paths can reduce delay and help pack more transistor capacity into a given area.
Huawei reports a density of 238 million transistors per square millimeter for the Kirin 9050 Pro, compared with approximately 155 million in the planar predecessor used as its baseline. Huawei describes that difference as a 55% increase.
The company also reports lower power consumption at matched performance: 66% for the NPU, 58% for the GPU and 41% for the CPU performance core. Those are component-level, matched-performance figures—not a promise that the Mate XT 2’s battery will last 66% longer in ordinary use.
| Metric | Kirin 9050 Pro figure | Comparison or baseline | Condition |
| Transistor density | 238 million/mm² | Approximately 155 million/mm² planar predecessor | Huawei-reported density comparison |
| NPU power use | 66% lower | Stated predecessor at matched performance | Huawei-reported figure |
| GPU power use | 58% lower | Stated predecessor at matched performance | Huawei-reported figure |
| CPU performance-core power use | 41% lower | Stated predecessor at matched performance | Huawei-reported figure |
| NPU performance | 70 TOPS | High-performance mode | Huawei-reported figure |
| GPU performance | 42% higher | Huawei’s stated comparison | High-performance mode; Huawei-reported figure |
| CPU benchmark result | 18% higher | Huawei’s stated comparison | High-performance mode; Huawei-reported figure |
What the independent testing adds
An independent benchmark review tested the Kirin 9050 Pro with SPEC 2026, 3DMark Steel Nomad Light and NPU throughput workloads. It also examined CPU and GPU behavior under demanding loads. One visible NPU result reached 67.7 TFLOPS INT8 in the test material.
That testing gives the performance discussion something more solid than a launch slide alone. It covers measured behavior under specific workloads, while the supply-chain question belongs to a different category entirely: benchmarks can measure throughput, efficiency and thermal behavior, but they do not audit a chip’s component origin or manufacturing equipment.
The processor is described as a nine-core LinxiCore design reaching up to 3.1 GHz, with a Maleoon GPU that supports hardware ray tracing. Huawei’s Mate XT 2 product material also advertises real-time hardware ray-tracing acceleration for 50-million-line scenes. The Da Vinci NPU is reported to support a 30-billion-parameter mixture-of-experts model on the device.
What “no US components” means here
Huawei says the Kirin 9050 Pro contains no US components or American technology. That is Huawei’s claim about the processor. The phrase itself does not specify whether it refers only to packaged components, or also to design software, intellectual property, fabrication equipment, servicing and the wider supplier chain.
Those distinctions matter in semiconductor manufacturing. A chip can be designed, fabricated, packaged and tested through a long chain of companies and tools. The identity of the processor’s branded components alone would not describe every stage involved in producing it.
Huawei has not disclosed the Kirin 9050 Pro’s process node, foundry, manufacturing volume or yield. It has also not published a complete design-tool and equipment chain for this processor. Earlier analysis of a different Kirin generation associated that predecessor with SMIC, but that information does not identify the foundry or process used for the Kirin 9050 Pro.
The performance story is clearer than the manufacturing story
The Kirin 9050 Pro is a commercial Huawei chip, it powers the Mate XT 2, and LogicFolding is a real architecture Huawei has brought into a smartphone. Its reported density and matched-performance power figures are specific manufacturer measurements, while independent testing covers selected performance and efficiency workloads.
The broader “US-free” description remains a statement from Huawei rather than a published, auditable map of every design and manufacturing input. That is why the chip’s architecture and benchmark behavior can be discussed with considerably more precision than its complete supply-chain origin.
