Recently, multiple domestic AI chipmakers including Huawei, Cambricon, Muxi and Iluvatar have raised product prices, with price hikes of 20% to 50% for certain models. This wave of computing power price increases serves as a wake-up call for the industry: the extensive development model of artificial intelligence relying on brute-force scaling of computing power is no longer sustainable, and improving quality and efficiency has become the core direction of industrial transformation.
With the large-scale deployment of large language models, computing power and power consumption for training and inference have grown exponentially. Shifting the focus to "efficiency competition" is critical to strengthening the resilience of the AI industrial chain and fostering new-quality productive forces. China has explored three pathways for intensive development. First, lightweight large models: model compression and inference optimization deliver top-tier performance with fewer resources. Second, deep hardware-software collaboration: domestic large models are jointly tuned with homegrown chips. For instance, DeepSeek V4 made its debut on Ascend, substantially cutting inference and hardware costs. Third, leveraging China’s abundant industrial scenarios to drive model adoption through vertical applications, where scenario data continuously feeds back into model iteration.
The AI industry still faces shortcomings such as weak core technologies, inadequate computing power support, and insufficiently deep industry applications. Future efforts will focus on three fronts: advancing breakthroughs in core technologies including high-end chips and foundational software; coordinating the national computing power network, promoting the integration of computing power and green electricity, and refining mechanisms for data element circulation; and targeting vertical sectors such as intelligent manufacturing and assisted healthcare, with supporting policies including subsidies and first-set procurement incentives to resolve bottlenecks in large model implementation.
The rise in computing power costs represents both a challenge and an opportunity for transformation. Only by consistently boosting computing power utilization efficiency, and cultivating domestic hardware and software ecosystems together with industry applications, can China’s AI industry seize development initiative and achieve high-quality growth.
This article synthesizes information from public industry reports and is for informational reference only.