A concept picture of AI city File photo: VCG

With Google unveiling Gemini 3 and Alibaba launching its new Qwen and Lingguang apps, the global race in artificial intelligence (AI) innovation seems to be intensifying. High-tech enterprises in China and the US are stepping up their investment as AI iteration cycle quickens to reach unprecedented speed, accelerating an important contest over the future of AI leadership, analysts said.

Both countries are making massive capital bets on AI. The four major US tech giants, Microsoft, Google, Meta, and Amazon are expected to  spend over $200 billion on capital expenditure in 2024, with much of the capital going toward AI data centers as well as developing advanced AI chips, Bloomberg reported.

Meanwhile, Chinese high-tech companies, to some extent constrained by insufficient hardware supplies, have gone all-in on capital investment in many AI-related fields in their hope their AI innovations won't fall behind. 

A report released by SAP on October 15 shows that Chinese companies—both large firms and fast-growing upstarts — are projected to spend an average of $42.8 million on AI research and development in 2025, covering hardware, software, talent teams and consulting services. This level of investment ranks first globally, ahead of the US ($37 million) and Germany ($34 million).

China's rapid AI ascendance has set off "alarm" bells in California's Silicon Valley and Washington DC. To stay ahead, the US administration unveiled an AI Action Plan in July 2025, calling for cutting red tape to build more data centers and generate enough electricity to power them. The US will do "whatever it takes to lead the world in artificial intelligence," US President Donald Trump said in a speech then, Bloomberg reported.

According to industry analysts, China and the US are no longer racing on the same AI track, but are moving along distinct technological paths and ecosystem models shaped by their own resources and market dynamics.

Path divergence

The US companies have focused on making advanced computing chips and large language models that underpin today's generative AI chatbots. The companies such as OpenAI and Google have concentrated on producing AI systems that mimic the human process of reasoning and generate videos, images, and audio content. They have also released so-called AI agents — designed to field increasingly complex tasks on users' behalf, Bloomberg reported.

However, most leading US frontier models, led by OpenAI's flagship systems, remain closed to the public. Even though a few open-source options exist, the mainstream American AI ecosystem is still built around restricted access to seek high profit margins. 

"Their multimodal progress is impressive, but fundamentally those models are anchored in a closed-door approach," Liu Gang, chief economist at the Chinese Institute of New Generation AI Development Strategies, told the Global Times on Thursday.

However, Chinese enterprises have moved in the opposite direction. Key developers such as DeepSeek, Qwen and Kimi have advanced rapidly by embracing open-source frameworks that allow widespread use, broad adaptation and community-driven improvement and upgrading, Liu noted.

According to a 2024 report by ITIF, a US tech think tank, China's large language  models are rapidly narrowing the performance gap with their US counterparts, with several Chinese chatbots even outperforming leading American models on bilingual benchmarks.

Unlike US companies, which are fixated on pushing parameter counts ever higher, Chinese companies have taken a highly pragmatic, engineering-driven path. Backed by supportive policies, they have rapidly rolled out new capabilities in mini-app generation, on-device inference and integrated "AI + hardware" solutions, analysts said.

China has already built more than 35,000 high-quality AI datasets, and Chinese content now accounts for over 60 percent of the training data used by mainstream Chinese AI models, providing a solid foundation for serving AI users, according to a report by the People's Daily.

A recent industry report shows that as of June this year, China has 515 million generative AI users — an increase of 266 million people from December 2024, according to the People's Daily's report.

China's "application-first" strategy, combined with its vast user base, has accelerated AI adoption. By introducing AI courses in primary and secondary schools and embedding AI tools into everyday services—from tourism and shopping to healthcare —China's domestic models are rolling out in vertical applications at a pace that surpasses the US', analysts said.

Distinct ecosystem 

China now has 14 out of the global top 20 AI models when ranked on tasks like reasoning, knowledge, math, and coding skills, according to OpenCompass LLM Leaderboard on October 18, 2025. Although US players still hold the top positions, nine Chinese rivals are open-source.

In the US, top artificial general intelligence (AGI) players are produced by trillion-dollar tech giants, where investment flows are driven largely by market valuations. This has created an ecosystem built on chip monopolies, big-tech capital and Silicon Valley finance, which exposes some structural bottlenecks ¬— limited high-quality data and an energy system that cannot keep up with surging computation demand, Tian Feng, president of the Fast Think Institute and former dean of Chinese AI software giant SenseTime's Intelligence Industry Research Institute, told the Global Times.

China places greater emphasis on turning AI capability into real productivity. Its abundant power capacity and extensive grid infrastructure give it a solid foundation for scaling computation, Tian said.

China's expanding open-source AI ecosystem is rapidly becoming a major lever for its large models to grow their international user base, according to Wang Peng, an associate researcher at the Beijing Academy of Social Sciences. For example, DeepSeek and other Chinese models are now widely used in engineering-related training programs in Africa, helping boost local development efficiency while reducing costs.

"Chinese large models match their US counterparts in performance but at a far lower cost, and they run on extremely inexpensive hardware," Toyin, an entrepreneur from Abuja, Nigeria, told the Global Times. For him, where computing resources are both scarce and expensive, the ability to deploy AI on cheaper, more energy-efficient platforms is essential for African tech startups to serve their users.

Bi Qi, chief scientist at the China Telecom Research Institute, told the Global Times that US companies often pursue "ultimate performance" in developing large AI models, which sharply drives up costs. By contrast, Chinese models — with their strong price-performance ratio, simpler architectures and lower power consumption — are better suited to engineering and commercial needs.

And, Liang Huaixin, a researcher at the Institute of National Security and Governance at the University of International Business and Economics in Beijing, noted that these attributes are making Chinese AI models widely popular not only across the Global South markets but also among a growing number of companies in developed countries. 

Despite their divergent paths, the two remain far from decoupled. Their competition shows that the global AI industry is not converging toward forming a single standard, but rather carving out distinct lanes. For global investors, recognizing this differentiation — US-led breakthroughs in basic research versus China-driven scale-up of real-world AI applications — will be key to assessing where technological dividends will flow in the coming years, analysts said.