Gasgoo Munich- AXERA announced the establishment of a wholly-owned subsidiary, Aixin Computing. The AI chip maker—which just surpassed 1 billion yuan in revenue last year—is creating a separate entity to navigate the increasingly fierce competition in the computing power market.
According to public information, Aixin Computing inherits the parent company's NPU technology roadmap, focusing on on-device and edge-side inference compute. Its product lineup includes compute cards and inference servers, targeting scenarios such as embodied AI, smart offices, retail, and industrial sectors in the physical AI field. At the recently concluded WAIC 2026, AXERA displayed its Yuanxi series inference cards, rated at over 1,000 TOPS. These products will serve as the debut offering for the new company.
Image source: AXERA
AXERA founder Qiu Xiaoshen defines the spin-off as a "strategic decision," yet the market is more focused on the commercial logic. The AI compute sector is currently enduring a brutal price war, with international giants like Nvidia and AMD applying pressure, while domestic rivals such as Cambricon and Horizon Robotics accelerate their push in inference chips. AXERA's 2025 financial report shows R&D investment accounted for more than 40% of revenue; independent operation may help the new company gain greater flexibility in decision-making efficiency and cost control.
It is worth noting that Aixin Computing's customer structure includes a significant proportion of automotive electronics and smart hardware manufacturers. These clients are highly sensitive to cost-performance ratios and supply chain security. Whether the new company can control costs while maintaining technological iteration will directly impact its market performance. Industry insiders point out that while the AI inference market has broad prospects, the risk of product homogenization is rising, and relying solely on hardware specs is difficult to establish long-term barriers.
Aixin Computing emphasizes a "high intelligence-to-price ratio" positioning in its official statement, but whether this can be delivered in mass production remains to be tested by the market. For this seven-year-old AI chip company, the real test lies in product execution and customer retention rates following the spin-off.
