On October 31, the 2026 AI Plus Conference opened in Beijing. Zheng Weimin, an academician of the Chinese Academy of Engineering, professor at Tsinghua University, and chairman of the Jiuyuan Intelligent Computing System Ecosystem Consortium, delivered a keynote on the domestic intelligent computing ecosystem at the main forum. He stated plainly: the quantity and performance of domestic chips keep improving and pricing is gradually turning competitive, yet real-world deployment still suffers from an insufficient software ecosystem — for domestic computing power to achieve true scale deployment, the key lies not only in chip performance but in the software ecosystem and actual usage costs.
Zheng further broke down the composition of usage cost: the purchase price of a chip does not fully determine its cost of use; the resources invested in software adaptation, development, and optimization equally affect the practical application efficiency of domestic computing power. Addressing this, he presented the Jiuyuan consortium’s practices — from chip adaptation to system-level innovation — reducing adaptation costs across different chips through a unified software architecture so that upper-layer applications can be developed once and run on many chips. On the same day, Qijing Technology, together with Zhipu AI and Moore Threads, released an inference service solution based on domestic heterogeneous computing. Multiple threads point to the same proposition: moving domestic computing power from available-to-buy to good-to-use and affordable-to-run.
From an industry-landscape perspective, Zheng’s judgment pinpoints the watershed of domestic computing substitution: the hardware-parameter race has delivered visible results, and the next phase of competition will revolve around software stacks, toolchains, and migration costs. It also means that the maturity assessment of domestic computing power is shifting from peak computing power to full-stack total cost of ownership — whoever enables developers to switch at low cost and applications to migrate painlessly will capture real share in intelligent computing centers and enterprise-side deployment. The main battlefield of the computing-power race is quietly shifting from silicon success to ecosystem success.
This judgment strongly resonates with Junyi Microelectronics’ engineering practice: the reason the Rockchip-based J-3568/J-3576/J-3588 core boards have become a stable choice for many equipment makers is not merely the price-performance of the chips themselves, but the long-term accumulation of invisible ecosystem work such as board-level BSPs, drivers, and system adaptation. Junyi’s K-series edge AI motherboards and the AI-BOX-3588 edge computing host ship with complete SDKs and adaptation support, minimizing customers’ onboarding cost on domestic platforms — treating the software ecosystem as an integral part of the product is the most straightforward and effective path to scaling domestic computing power.
Source: http://finance.sina.com.cn/jjxw/2026-10-31/doc-inixcnhn4764946.shtml (October 31, 2026)