MediaTek announced a launch event on September 15, expected to unveil the Dimensity 9600 series, including standard and Pro versions. The Pro version uses TSMC's 2nm process, offering significant performance gains, and will be featured in vivo X500 and OPPO Find X10 series.
CXMT announced that its self-developed LPDDR6 chip has entered mass production and commercial use, debuting in the Xiaomi 18 Fold foldable flagship phone, marking the first global commercial deployment of LPDDR6 in a flagship phone. The chip supports a maximum transfer rate of 12800Mbps, offering 60% higher bandwidth and 20% lower power consumption compared to LPDDR5X. CXMT stated that LPDDR6 will be used in phones, PCs, smart cars, edge servers, and more, representing a major milestone in its high-end memory development.
The article analyzes the supply shortage in China's domestic AI chip market, comparing domestic vs imported chip shipments (4 million vs 3 million), and notes that under export controls, domestic computing power has become a seller's market, tilting pricing power toward suppliers.
At its autumn flagship launch event, Xiaomi unveiled the Xuanjie O100 chip, a 1.22TB/s high-bandwidth AI accelerator featuring the world's first 6nm 3D wafer-level stacked advanced packaging and hybrid bonding process. The chip offers 16x the memory bandwidth of traditional phones, achieving 330TPS ultra-high-end on-device inference speed.
At the HarmonyOS 7 launch event, Huawei officially unveiled the Kirin 9050 Pro processor, its first new Kirin chip in six years. The chip uses logic-folding technology, stacking logic units to improve performance. The Mate XT 2, the first phone to feature it, sees a 42% performance boost.
Multiple research firms report that OmniVision surpassed Samsung in global CMOS image sensor (CIS) market share for the first time in 2025, becoming the world's second-largest supplier behind Sony. According to Techno Systems Research, the global CIS market was about $23 billion in 2025, with OmniVision's revenue around $3 billion, a 13% share. Nikkei BP also shows OmniVision second with 13.4%, Sony at 53.6%, and Samsung at 13.2%. OmniVision's growth is driven by automotive CIS, with revenue of 7.471 billion yuan, up 26.52% year-over-year, and over 30% global share in automotive CIS, ranking first.
DeepSeek is reportedly planning to purchase 160,000 Huawei chips to boost its AI computing power. This move may signal a deepening reliance on domestic chips by Chinese AI companies and could reshape the domestic AI computing landscape.
Xinsijie has released a 400Gbps back-illuminated PIN PD chip, aiming to accelerate the deployment of global AI compute optical interconnects. The chip may be used in high-speed optical communication modules for data centers and AI computing infrastructure.
Rayson has launched a 96GB, 9600MT/s LPDDR5X LPCAMM2 memory module for high-performance AI PCs. Compared to traditional DDR5 SO-DIMM, it reduces board area by about 60%, power consumption by about 50%, and improves energy efficiency by about 70%, and has completed adaptation validation with leading international PC manufacturers.
GenStorAIGE unveiled the PCIe Gen6 retimer GE6216 at the 2026 Open Data Center Conference, supporting 64GT/s for AI servers and CXL link expansion. The chip compensates high-loss channels for signal integrity and is pin-compatible with mainstream solutions.
An AI research institute points out that Huawei's AI computing power still lags far behind Nvidia's. The report may be based on comparisons of chip performance and ecosystem, but no specific figures are provided.
This article from Xueqiu discusses China's shortage of computing power and emphasizes the importance of self-developed AI chips. The author believes that Huawei's chance of catching up with Nvidia by 2030 is almost nil, reflecting China's challenges in the AI chip sector.
BYD is making consecutive strides in the chip sector, first launching a 4nm computing chip and then announcing mass production of a 4D radar chip. These chips are mainly for intelligent driving and automotive electronics, showcasing BYD's vertical integration capabilities in semiconductors.
On September 5, Horizon Robotics founder and CEO Yu Kai stated at the 2026 Yabuli Entrepreneurs Forum Summer Annual Meeting that the company's goal next year is to surpass Nvidia in China's high-level autonomous driving chip market. Yu Kai revealed that Horizon launched a 560 TOPS high-compute chip in 2025 and currently ranks second in this field, behind Nvidia. Earlier, Horizon announced that its Journey series autonomous driving chips have surpassed 15 million units in mass production, with partnerships with over 40 brands.
Horizon Robotics founder Yu Kai stated that the company aims to surpass Nvidia in market share for high-performance autonomous driving chips in China next year. This reflects Chinese chip companies' attempt to challenge Nvidia's dominance in specific segments.
Huajing Motors apologized for delivery delays of its Huajing S model due to a temporary shortage of cockpit chips for the HarmonyOS cockpit, caused by industry-wide chip supply chain fluctuations. The company has initiated special supply assurance efforts, including expanding chip supply channels and promoting diversified supply solutions. The Huajing S is the first model of SAIC-GM-Wuling's premium intelligent brand, with August deliveries of 7,306 units and cumulative deliveries exceeding 20,000 units.
On September 3, SiEn Integrated and Li Auto signed a new round of strategic cooperation agreement and jointly witnessed the successful roll-off of 8-inch silicon carbide wafers. Their cooperation has expanded from vehicle electric drive to on-board thermal management systems, following the mass production of 6-inch SiC wafers in February 2025 for Li Auto's i-series models. The 8-inch products are about to enter SOP mass production, marking a key milestone in their partnership.
Qujing Technology and Moore Threads announced a strategic cooperation to produce high-quality computing power using domestic chips. This reflects China's trend of promoting domestic substitution in AI computing power.
Huawei engineer He Tingbo disclosed measured data for the Kirin 2026 chip in a preprint paper, which uses logic folding technology to increase transistor density from 155 million to 238 million per square millimeter, a 55% increase. At equivalent performance, NPU power consumption drops 66%, GPU 58%, and CPU big core 41%, with lower temperatures. The technology shortens signal routing, greatly reducing interconnect capacitance and power.
IT Home reports that a blogger leaked a domestic chip scoring 2084 in GB6 single-core at 3.1GHz, a 12.4% improvement over the previous generation, with significant energy efficiency gains. Netizens speculate it is the Huawei Kirin 9050 series, possibly for a new foldable phone.