The United States is pouring unprecedented sums of money into artificial intelligence, with private investment reaching $285.9 billion in 2025. The Economist’s analysis examines how China manages to keep pace on a fraction of the budget. In stark contrast, China’s private AI investment during the same period was just $12.4 billion, a 23-fold difference. Yet, despite this massive funding gap, Chinese AI models are rapidly closing the performance gap with their American counterparts, raising questions about the efficiency of US capital deployment.
The 23-Times Gap in Context
The disparity in spending is most evident in the capital expenditures of the major tech giants. American companies like Google, Microsoft, Amazon, and Meta are expected to spend over $740 billion on data centers and AI infrastructure this year alone. In China, the spending is far more constrained. Alibaba, for example, pledged to invest more than $53 billion in AI infrastructure between 2025 and 2028—a significant sum, but one that pales in comparison to the budgets of its US rivals, according to The Economist.
Three Reasons China Punches Above Its Weight
How is China managing to keep pace on a fraction of the budget? The answer lies in a combination of strategic focus, engineering efficiency, and a heavy reliance on open-source foundations.
First, Chinese AI developers have proven highly adept at building upon existing open-source architectures, such as Meta’s LLaMA models. Rather than spending hundreds of millions of dollars training frontier models from scratch, Chinese startups often use open-source models as a starting point, focusing their resources on fine-tuning, optimization, and specialized applications. This approach significantly reduces the compute and capital required to bring a competitive model to market.
Second, Chinese firms are highly focused on inference efficiency—the cost of running a model once it has been trained. Companies like DeepSeek and Moonshot AI have engineered their models to operate at a fraction of the cost of US proprietary models. By optimizing for efficiency, these companies can offer highly capable AI services at prices that undercut American competitors, making them attractive to cost-conscious enterprises.
Third, the constraints imposed by US export controls have forced Chinese companies to innovate in hardware utilization. Denied access to Nvidia’s most advanced AI accelerators, Chinese engineers have developed sophisticated software techniques to extract maximum performance from older, less capable chips. This hardware-software co-design approach has allowed Chinese firms to train large models using fragmented or less efficient compute clusters, mitigating the impact of the chip bans.
The result is a highly competitive AI ecosystem that punches far above its financial weight. Models like Zhipu AI’s GLM-5.2, Moonshot’s Kimi K3, and Alibaba’s Qwen series are consistently ranking near the top of global leaderboards, challenging the dominance of OpenAI and Anthropic.
The US strategy of overwhelming financial force has undoubtedly produced the world’s most advanced frontier models. However, the Chinese approach demonstrates that raw capital is not the only determinant of success in the AI race. By prioritizing efficiency, open-source collaboration, and cost-effective deployment, China has built a resilient AI industry capable of competing globally, even as the US outspends it by an order of magnitude.
What the Efficiency Gap Means for Investors
The implications for global AI competition are significant. If Chinese firms can continue to deliver frontier-adjacent performance at a fraction of the cost, they will increasingly dominate the markets where price sensitivity is highest, particularly in developing economies across Southeast Asia, Africa, and Latin America. These markets represent the next frontier of AI adoption, and the company or country that captures them will have a significant advantage in shaping global AI norms and standards.
For investors, the efficiency gap raises important questions about the sustainability of US AI valuations. If Chinese models can match or approach the performance of US frontier systems at dramatically lower cost, the premium that markets are currently placing on companies like OpenAI and Anthropic may be difficult to justify over the long term. The Economist’s analysis suggests that the US AI sector is pricing in a level of market dominance that China’s frugal competitors are actively working to undermine.
The 23-times spending gap also has implications for the long-term sustainability of the US AI buildout. EastFrontier has previously covered Tencent pledging higher AI investment in 2026 after chip curbs hit its capex plans. The current wave of hyperscaler capex is predicated on the assumption that AI will generate returns commensurate with the investment. If Chinese firms can deliver comparable capabilities at a fraction of the cost, the addressable market for premium US AI services may be smaller than current projections suggest.
That is not a problem that export controls or trade restrictions can solve. It requires a fundamental rethinking of how the US AI industry creates and captures value. The Economist’s analysis suggests the answer may lie in applications and distribution rather than model performance, areas where US companies still hold significant advantages in global markets but where Chinese competitors are moving fast.