China’s AI Rise: Innovation Overcomes Chipmaking and Investment Gaps

China is not idle. In early 2025, researchers completed a functional prototype of an EUV lithography machine, developed by former ASML engineers using reverse-engineered technology and parts from older equipment. This prototype is part of a national initiative to achieve self-sufficiency in chip production by 2028 to 2030. The Shanghai Institute of Optics and Fine Mechanics has also made progress in EUV light source technology, developing laser-produced plasma EUV sources driven by solid-state lasers as an alternative to ASML’s CO2 laser systems.

However, the prototype has not yet produced working chips. Challenges in photoresists, chemicals, and yield rates persist, forcing China’s semiconductor sector to rely on older deep-ultraviolet (DUV) lithography machines as a stopgap. Yao Shunyu, Tencent’s chief AI scientist and former OpenAI researcher, emphasised this point at a Beijing conference: “The main bottlenecks are production capacity, including lithography machines, and the software ecosystem.”

Algorithm-Hardware Co-Design: Doing More with Less

With limited access to advanced hardware, Chinese AI firms are adopting algorithmic efficiency strategies to maximize performance. This approach integrates hardware development with AI algorithm optimisation, allowing large models to run efficiently on less powerful, more affordable hardware. It is a pragmatic response to the constraints imposed by the lack of cutting-edge chipmaking tools.

Baidu’s Kunlun AI chips are just one example. By incorporating specialised accelerators tailored for transformer operations in large language models, Kunlun chips achieve an 80% reduction in total cost through combined hardware and algorithm optimisations. Similarly, Huawei’s MindSpore AI framework works in tandem with its Ascend chips to identify and address performance bottlenecks, creating a feedback loop that accelerates innovation.

Startups like Listen AI take this strategy further by embedding algorithm-derived architecture from the outset. Their system-on-chip technology supports end-to-end AI scenarios, such as voice interaction and real-time speech processing, on smaller, lower-power hardware. This method reduces costs and ensures AI models can operate efficiently in resource-constrained environments.

A Culture of Risk-Taking Emerges

Younger Chinese entrepreneurs are increasingly embracing high-risk ventures, a trait once associated exclusively with Silicon Valley. This change is driven by government policies designed to foster innovation and reduce financial risks for startups. Tang Jie, founder of Zhipu AI, highlighted this shift at the AGI-Next Frontier Summit in Beijing: “If we can improve this environment, allowing more time for these risk-taking, intelligent individuals to engage in innovative endeavours, this is something our government and the country can help improve.”

The Chinese government has introduced a range of initiatives to support this culture. The AI+ Initiative, launched in 2024, mobilises resources across government and industry to integrate AI into the economy. Public-private Government Guidance Funds provide capital to deep-tech AI startups, while local governments offer subsidies, tax incentives, and AI industrial parks. The National AI Industry Investment Fund, valued at over USD 8 billion, further fuels innovation by providing substantial backing for startups.

The Investment Gap: A Persistent Challenge

Despite these advances, China’s AI sector remains behind the US in infrastructure investment and computing power. The US private sector invested over USD 100 billion in AI infrastructure in 2024-2025, nearly 12 times China’s USD 9.3 billion. US hyperscalers like Nvidia, AMD, and Intel dominate the market, with the US maintaining a 6-12x advantage in computing infrastructure.

China is making progress, with the government pledging over USD 150 billion by 2030 and spending USD 50-70 billion annually on AI chips and data centres. Initiatives like “Eastern Data, Western Computing” aim to build a national computing infrastructure to support AI development. However, China’s reliance on government-led investment, rather than private sector funding, remains a key difference and a potential vulnerability.

Lin Junyang, technical lead for Alibaba’s Qwen large language model, acknowledged this gap during a panel discussion at Tsinghua University: “The US computer infrastructure is likely one to two orders of magnitude larger than ours. But I see that whether it’s OpenAI or other platforms, they’re investing heavily in next-generation research. We, on the other hand, are relatively strapped for cash; delivery alone likely consumes the majority of our computer infrastructure.”

Closing the Gap: China’s AI Ambitions

Experts remain divided on whether China can overtake the US in AI leadership. Chinese researchers, including Yao Shunyu, believe there is a high likelihood of a Chinese firm becoming the world’s leading AI company within the next three to five years. However, this optimism is tempered by the reality of technological and financial constraints.

While China has narrowed the AI model performance gap from three years in 2023 to 6-12 months in 2025, the US still holds a decisive advantage in infrastructure and private investment. Richard Clode, a portfolio manager at Janus Henderson, argued that the US is likely to pull ahead in the AI race due to its superior infrastructure: “The AI infrastructure that US tech companies are able to access will be so much more advanced than the infrastructure that Chinese tech companies have”.

Yet, China’s strengths—government backing, a vast talent pool, and a culture of innovation under constraints—remain significant. The country’s ability to deploy AI at scale, particularly in AI-powered manufacturing, logistics, and urban infrastructure, could prove decisive in the long run. As AI systems become embedded in real-world applications, China’s focus on mass adoption and deployment may accelerate its progress.

A New Era for Global AI

China’s AI sector is redefining what is possible under constraints. By leveraging algorithm-hardware co-design, government support, and a culture of risk-taking, it is closing the gap with the US. The recent IPOs of MiniMax and Zhipu AI are early indicators of what could become a transformative era for China’s technology landscape.

The global implications of this race are profound. If China succeeds in narrowing the gap with the US, it could reshape the technology landscape, challenging Western dominance in AI and semiconductors. For now, the world watches as China demonstrates that innovation is not just about having the best tools. It is about making the most of what you have.