Cheap, fast and everywhere: Why China leads AI adoption

The US-China competition over generative AI is often framed as a race to build more powerful models. From DeepSeek and ChatGPT to Seedance and OpenAI’s Sora, model-to-model comparisons have reinforced this narrative.

But model comparisons tell only part of the story. An equally important question is who can put generative AI to work faster, cheaper and at greater scale. Seen from this perspective, China’s comparative advantage lies less in producing the world’s best foundation models than in converting generative AI into usable products, services and workflows at speed and scale.

That capacity rests on a combination of conditions: large-scale unmet demand, mature platform ecosystems, open-source and cost-effective models, manufacturing depth, and regulation that sets guardrails for rapid adoption.

Scale and unmet needs drive adoption

By the end of 2025, China had about 1.125 billion internet users, while the number of generative AI users had reached 602 million, with a national adoption rate of 42.8%. These figures suggest that generative AI in China is no longer confined to early adopters but is becoming part of everyday digital life.

China’s scale matters not simply because it gives AI companies a large market, but because it amplifies the benefits of efficiency. Across healthcare, education and small businesses, many needs remain unmet or unevenly served. This creates practical demand for tools that can lower costs, extend expertise and help service providers do more with limited resources. Even modest improvements can be multiplied across hundreds of millions of users and service providers.

For example, China has long faced pressure to improve access to quality medical services beyond top urban hospitals, which explains why Chinese authorities have called for broader AI application in the health sector. The key implication is not that AI will replace human doctors, but that China has strong incentives to adopt AI wherever skilled human resources are scarce.

Falling costs have also lowered the threshold for adoption. With cheaper, smaller and more open models available, Chinese firms do not always need frontier-level systems. In many commercial and industrial settings, “good enough” models that are affordable, customisable and easy to integrate may matter more than benchmark-leading performance.

From digital tools to physical systems

AI video shows how China can turn generative AI from model capability into commercial use. China already has a mature ecosystem of short video, livestreaming e-commerce, digital advertising, microdramas and online entertainment, all of which depend on fast and low-cost visual production. A model that can produce product clips, virtual presenters, storyboards or advertising materials is therefore not just a technological novelty, but a tool that can enter the production process itself.

Kuaishou’s Kling AI illustrates this shift. Company figures show that Kling AI reached an annualised revenue run rate of US$240 million in December 2025, served more than 60 million creators worldwide and had generated more than 600 million videos. These numbers show how quickly AI tools can spread through China’s platform economy when they are useful, fast and cheap enough to fit into daily content production.

China’s capacity to apply generative AI is not limited to screens. It also extends into hardware, logistics, factories and robotics, where AI can be embedded in physical systems rather than used only as a digital service.

According to the World Robotics 2025 report, China was by far the world’s largest industrial robot market in 2024, accounting for 54% of global deployments, with 295,000 units installed and an operational robot stock exceeding 2 million. This matters because factories, logistics networks and production lines provide real-world environments where AI-enabled systems can be tested and integrated outside purely digital settings.

Humanoid robots show both the appeal and the risk of this shift. Although the sector remains vulnerable to hype and premature commercialisation, China’s dense industrial base gives it many sites where AI systems can be tested, adjusted and scaled in real-world settings.

Optimism and guardrails speed adoption

Public optimism about AI is another enabling condition for adoption. According to the 2025 Stanford AI Index Report, 83% of respondents in China said AI products and services were more beneficial than harmful, compared with 39% in the United States, 40% in Canada and 36% in the Netherlands.

This gap should not be explained simply as a matter of cultural difference. In China, AI is often introduced less as a distant frontier technology than as a practical tool embedded in familiar platforms and services. When AI appears in shopping, healthcare or the workplace, users are more likely to judge it by whether it solves immediate problems, saves time or lowers costs.

That does not mean privacy concerns are absent, but they are often weighed against convenience and perceived usefulness. Compared with societies where AI debates are more strongly framed around individual privacy, liability and rights, Chinese users may be more willing to try generative AI tools when the benefits are visible and immediate.

Legal and regulatory guardrails also help shape adoption. China’s 2023 interim measures cover public-facing generative AI services, while its 2025 labelling rules require explicit and implicit identification of AI-generated synthetic content. These rules do not remove risks, but they give firms and users clearer expectations about how AI can be commercialised and used within red lines around content security, data governance and platform responsibility.

Application speed brings both advantage and risk

None of this means that China will dominate every layer of generative AI. The US retains major strengths in frontier models, advanced chips, research talent and private capital. China’s strength lies in the speed and scale with which it can turn AI from model to product, from product to workflow, and from workflow into everyday use. By contrast, the US path from model capability to widespread application is more uneven, constrained by fragmented AI regulation, greater litigation risks around data privacy and copyright, and the firm-by-firm work of embedding AI into existing workflows.

Yet the same speed that helps China scale AI applications can also amplify risks. Rapid deployment can bring copyright disputes and content-authenticity problems to the surface sooner. It can also accelerate labour displacement, heighten privacy concerns and fuel speculative investment. China’s strength in application does not remove these problems, but it may make them visible earlier and at a larger scale.

The future of generative AI will not be decided only by who builds the most powerful model, but also by who can put it to work quickly, cheaply and at scale. On that terrain, China is moving fast.

Related: Winning the AI race without an OpenAI | How AI is rewiring white-collar work in China