When discussing the outlook for next year, Dr. Liu holds a reserved stance on the beta of the computing power supply chain.

This year, market logic has been extremely "pure," largely attributed to the pure-play NVIDIA supply chain. However, the landscape is quietly shifting. First, Google has begun challenging the reigning leader. Its newly launched next-gen AI model, Gemini 3, is considered the most advanced multimodal comprehension system currently available, significantly outperforming competitors in benchmark tests. Second, and more critically, the validity of Scaling Law is facing increasing skepticism.

Scaling Law posits that model performance improves continuously with scale—a core assumption that initially drove rapid AI advancements. Yet, as models grow larger, performance gains have plateaued, casting doubt on this principle. These two marginal shifts could impact NVIDIA's valuation logic and, on a broader scale, potentially influence whether the Philadelphia Semiconductor Index's boom cycle transitions into a mid-term correction.

Another factor that cannot be ignored for the domestic computing power supply chain's beta is the potential "H200 to the East"—reports suggest the Trump administration is considering allowing NVIDIA to sell its H200 chips to China. NVIDIA's dominant CUDA ecosystem poses short-term sentiment risks for domestic computing power players.

However, there is more optimism regarding next year's beta for AI edge-side and applications. Narratively, developing AI edge-side aligns perfectly with China's strategic AI positioning.

The AI edge-side represents a vast and complex industrial ecosystem—integrating large models into consumer electronics peripherals, application scenarios, data mining, digital assets, and credit expansion cycles. Currently, Chinese models like Qwen and Zhipu are gaining global recognition among developers. Meanwhile, China's manufacturing prowess—spanning consumer electronics, EVs, drones, and robotics—powers the "Made in China" wave. The core significance of AI edge-side lies in monetizing the massive capital expenditures of AI infrastructure while forming a closed-loop industrial ecosystem.

While the West pushes Google's "full-stack self-reliance" model—aiming to integrate chips, software, models, and applications into a competitive advantage—China should forge its own AI strategy. China's strength lies in manufacturing power, with robotics, drones, smart peripherals, and EVs serving as key battlegrounds. Software is merely a tool; open-source platforms invite global users into the AI ecosystem, which is then embedded into cost-effective hardware (AI Agents) and sold worldwide.

The application scenarios and data resources spawned by this AI ecosystem will create a powerful flywheel effect, ultimately anchoring the vast wealth generated by AI within China.