According to a report by China Daily on March 25…
China’s artificial intelligence industry is entering a new stage, with leading companies not just competing globally, but reshaping how AI models are built, shared, and monetized.
Yang Zhilin, founder of Beijing-based Moonshot AI, revealed in an exclusive interview at the Zhongguancun Forum that his company is exploring a major overhaul of large model architecture, including a novel technique called “attention residuals”, praised by Elon Musk earlier this month.
“China’s willingness to openly share models and technical breakthroughs could accelerate global innovation while giving it an edge over more closed ecosystems,” Yang said.
From Algorithms to Infrastructure
Yang predicts that as AI models achieve similar performance levels, competitive advantage will shift from algorithms to infrastructure — especially the ability to process vast volumes of “tokens”, the basic units of AI computation.
“In the long run, the bottleneck may no longer be model capability, but how quickly you can build large-scale ‘token factories’,” he explained, noting energy costs and computing infrastructure as decisive factors.
He also suggested that AI-generated tokens could eventually become a proxy for economic activity, potentially reshaping traditional metrics such as GDP.
Open Systems and Economic Value
According to Yang, open-source AI systems may ultimately dominate by enabling a broader ecosystem of developers, applications, and distribution channels. While closed models retain market share, open platforms could generate greater total token output and therefore higher economic value.
The attention residuals innovation reconfigures how information flows through model layers, improving training efficiency and performance, though Yang did not disclose detailed benchmarks.
“Standards have formed, but within those standards, there is still a lot that can be overturned. I think there are still many possibilities,” he said.
China’s Talent Pipeline
Yang highlighted China’s decades-long investment in education, from primary to doctoral levels, as a key strength. The large pool of technically skilled talent enables companies to accelerate machine intelligence development efficiently and cost-effectively.
As China positions itself as a driver of structural change in the AI ecosystem, Yang believes the combination of open innovation, advanced infrastructure, and human expertise will define the next phase of global AI development.