On Oct 20, DeepSeek, a Chinese artificial intelligence company, announced that it was open-sourcing DeepInfer, the inference framework that supports its large-scale AI services.

While much of the AI industry remains focused on developing increasingly powerful models, DeepSeek has chosen to share an important part of the technology used to operate them. This decision reveals an important difference between the approaches being pursued by leading Chinese and American AI companies. More fundamentally, it reflects a juxtaposition of values that could have profound implications for the world.

DeepInfer powers the large-scale inference service behind DeepSeek V4.1 Flash. According to the company, a single cluster serves approximately 155 trillion tokens per day, with revenue of around $38 per graphics processing unit (GPU) hour at off-peak prices. These figures demonstrate the enormous scale at which DeepSeek operates and the importance of efficiency in delivering AI services.

DeepSeek has now decided to share some of the technologies behind its inference operations. Released under the Apache 2.0 open-source license, DeepInfer allows other developers to benefit from its work on making AI models run more efficiently. Although DeepSeek has not released every part of its production system, the technologies being shared could help other companies provide AI services at lower costs.

A broader trend toward lower-cost and more open AI technologies is also emerging in China. Models such as DeepSeek, Qwen and Kimi have become highly competitive internationally. While they may not outperform the most advanced American models in every respect, they offer sufficiently good performance at significantly lower costs. Many are also open-weight, allowing developers to deploy and customize them.

Meanwhile, American companies such as OpenAI, Anthropic and Google have led much of the development of frontier AI models, supported by enormous investment in computing infrastructure, research and talent. Their most advanced models are generally proprietary and command premium prices.

For businesses, the difference is becoming increasingly important. A company using AI for customer service, document processing or routine business analysis does not necessarily need the world’s most powerful model. It needs a model that can perform the required tasks reliably and economically.

As AI becomes part of everyday business operations, cost and accessibility will become increasingly important. Lower costs can make previously uneconomical applications viable, allowing smaller enterprises to adopt technologies that were once mainly available to large companies.

The competition in AI is therefore no longer simply about who can develop the most intelligent models, but also about who can make these technologies more widely accessible and useful.

These different approaches are also shaped by the economic and social systems in which the companies operate. The American economic system has long emphasized private enterprises, intellectual property, individual entrepreneurship and financial returns. This system has produced many of the world’s leading technology companies. For companies investing enormous resources in frontier AI, protecting proprietary technologies and maintaining technological advantages are a natural part of their business strategies.

China’s development has followed a different path. Over the past several decades, Chinese companies have repeatedly demonstrated their ability to combine technological innovation with large-scale production and continuous cost reductions. This has been evident in consumer electronics, electric vehicles, batteries and renewable energy. Products that were once expensive and available to relatively few people have become increasingly affordable and more widely used. AI is now showing signs of following a similar path.

China’s development experience has also shaped a different way of thinking about innovation. Among some of its leading technology entrepreneurs, commercial success remains important, but so does the contribution their technologies can make to national development, societal betterment and humanity in general. This reflects a broader orientation in China’s political and social thinking, which places considerable emphasis on connecting economic and technological progress with wider societal considerations.

DeepSeek’s decision to open-source its technologies is consistent with this orientation. The company may also benefit commercially from a larger developer community and a stronger ecosystem around its technologies. At the same time, other companies and developers can build on DeepSeek’s work. The benefits of its innovation could therefore extend beyond the company itself.

These ideas are consistent with the direction articulated by China’s leadership. At the 2026 World Artificial Intelligence Conference in Shanghai in July, President Xi Jinping emphasized the importance of openness, collaboration and sharing in AI development. He said that AI should contribute to shared prosperity, help bridge the technological divide between countries and benefit humanity more broadly. He also reiterated China’s commitment to providing international public goods relating to AI.

China has begun translating these ideas into concrete initiatives, including AI training opportunities for developing countries and plans for international cooperation on AI applications. These efforts reflect China’s emphasis on making the benefits of AI development available to a broader range of countries and societies.

Of course, this does not mean that all Chinese companies follow the same approach. American companies have also made important contributions to open-source technologies. But as Chinese AI companies become increasingly competitive, their emphasis on affordability and openness is emerging as an important feature of the global AI landscape.

Chinese companies are no longer exporting only products, technologies and business models. Increasingly, they are also exporting values.

As Chinese AI technologies become more widely adopted, the emphasis on openness and sharing may also gain greater international acceptance. Developers and businesses that use and build upon these technologies could gradually influence how AI is developed, shared and applied around the world.

At the same time, geopolitical competition is pushing the world toward what has been described as “one world, two systems”. AI offers an interesting example of how greater technological openness and deeper geopolitical divisions can emerge simultaneously.

The United States retains major advantages in advanced chip design, frontier models and cloud infrastructure. China, facing restrictions on access to advanced American semiconductors, is accelerating the development of domestic computing technologies while finding new ways to improve efficiency through software and system-level innovation.

The two countries are developing increasingly distinct technological ecosystems, shaped by geopolitical competition and national security concerns. Some observers expect this divergence to eventually force other countries to choose between the two systems.

The reality will be more complicated. The American and Chinese systems will continue to compete, and some parts of the AI industry may become increasingly separated. This is particularly likely in advanced semiconductors, computing infrastructure and technologies considered critical to national security.

But in other areas, especially AI models and applications, the two systems may continue to coexist and become intertwined. For companies outside China and the US, there may be compelling reasons to work with both systems.

For example, a company in Southeast Asia might use an advanced American model for complex research while relying on a lower-cost Chinese model for routine business operations. It might even deploy Chinese open models on American cloud infrastructure. The decision would depend on the company’s business requirements, the performance of the models and the costs involved, rather than simply on which country developed the technology.

In fact, AI development is already highly interconnected. DeepInfer itself was originally built primarily on vLLM, an open-source inference engine, and drew inspiration from other international projects. DeepSeek has benefited from the work of developers around the world and is now making some of its own technologies available to the same international community.

This kind of international exchange has long been an important part of technological progress. Geopolitical restrictions will undoubtedly affect it, but the practical needs of businesses and developers will continue to encourage cooperation across different technological systems.

Chinese AI technologies are expected to gain international adoption at an accelerated pace. Their competitive performance, relatively low costs and the availability of open models make them particularly attractive to small and medium-sized enterprises and countries in the Global South.

Open-weight models also allow local developers to adapt AI to their own needs. They can tailor these models to local languages, industries and market conditions, developing applications for healthcare, education, agriculture and other areas that may not be priorities for the largest international AI providers.

As these applications expand, more developers and businesses could acquire the experience needed to create their own AI solutions. Countries that have traditionally relied on imported technologies may gradually develop stronger domestic capabilities. This could become particularly important for economies seeking new sources of productivity and growth.

Greater international adoption could, in turn, strengthen the ecosystems surrounding Chinese AI technologies. As more developers use these models and create applications, they can contribute improvements and generate new demand. A broader developer community could stimulate further innovation and help bring costs down, encouraging still more businesses to adopt AI.

As AI becomes an increasingly important capability across industries and societies, the question of who controls it will become more consequential. If the most advanced and useful AI capabilities remain concentrated among a small number of companies, those companies may also exercise considerable influence over access, pricing and the economic value created by AI.

If these capabilities become more widely accessible, the opportunities could be distributed much more broadly. Smaller businesses could compete in areas previously dominated by large corporations. Entrepreneurs in developing countries could create new services and businesses. More countries could participate in technological innovation and develop industries suited to their own economic and social needs.

Therefore, this is more than just a competition between technologies. Behind the different approaches being pursued by Chinese and American AI companies are different ideas about the relationship between innovation, commercial interests and the broader benefits to society. These ideas, rooted in the historical experiences and political and social beliefs of the two countries, are increasingly finding expression in how AI is developed and made available to the world.

These different values will play an important role in shaping the future of AI. As AI becomes more deeply embedded in economic and social life, they could influence who creates wealth, who benefits from technological progress and how opportunities are distributed. What begins as a difference in approaches to AI could ultimately contribute to fundamental changes in socioeconomic systems around the world.

The author is founder and CEO of Gao Feng Advisory Co, a strategy and management consulting firm with roots in China.

The views do not necessarily reflect those of China Daily.