Chinese company DeepSeek launched its generative AI ‘r1 model’ in January. Run and primarily funded by Liang Wenfeng, a billionaire and former trader, DeepSeek focuses on novel AI research. DeepSeek has published scientific papers detailing the technological advances used in its ‘Large Language Model’ (LLM) r1 and released the weights of the model open source.
The model uses less energy to run and process the large amounts of human-created data required to ‘train’ it. For what was at that point a relatively unknown company, having to work around the embargo on the export of the chips needed to produce LLMs, r1 performed at a level comparable to the massive LLMs made by tech giants such as Google, OpenAI and others; and at a much cheaper cost.
Advances in science and technology do not exist in a vacuum or separate from wider society. The ‘AI arms race’ shows clearly that the development of more powerful AI models is driving competition between companies and, wider than that, nation states, in the pursuit of their own economic interests and profits. Restrictions on the importing of advanced computer chips needed to develop AI models, introduced by US president Donald Trump in his first term and strengthened by Joe Biden, are part of the raft of protectionist measures ostensibly to strengthen the position of US capitalism against the challenges of the Chinese economy.
US capitalists hoped to use their previously dominant position in AI to entrench an advantage against Chinese firms. They used both their economic might on a world scale to enforce the ban on chip exports, and utilised and expanded the vast data processing centres that already exist and are owned by the likes of Amazon and Microsoft. Data centres have been given vast tax breaks to encourage investment.
DeepSeek r1’s release undermined the idea that US-based firms have an unreachable lead in AI models over Chinese firms. The impact was immediate, puncturing somewhat the tech companies bubble. Nvidia, the US company which makes chips required for LLMs, lost $600 billion in market value in a single day – a fall of 17%! Now embroiled in the general downturn in stock prices as a result of the destabilising effects of the Trump presidency and announcement of tariffs, tech stocks have not recovered to the levels they were at the start of the year.
Despite the embargo on chips coming into China, the Chinese state took seriously a strategy to gain access to them and direct research and development across the whole economy to compete with US tech firms.
Because of the nature of the Chinese state – presiding over a unique form of state capitalism – it has been able to do so. There are many billionaires in China and private companies that make massive profits, but the Chinese Communist Party (CCP), led by Xi Jinping, has the power to direct economic production in ways that other capitalist states cannot. It is motivated by defending its own rule, not the interests of the mass of Chinese workers, but in a number of fields it has shown the advantage that state-directed economic planning can play in competing with the giant of US capitalism.
In 2017 the CCP enacted the New Generation Artificial Intelligence Development Plan. This set out the long-term strategic plan for China to become the world leader in AI by 2030. State and private investment was pushed for AI development and infrastructure. While there are no exact figures on the amounts invested, one example is AliBaba – the shopping giant, and one of the main firms in China developing LLMs – which plans to invest over $50 billion in AI and cloud computing over the next three years.
All sectors of the economy were included in plans to integrate AI into production. To fight the embargo, chips were stockpiled and computing power in data centres in different countries has even been rented by Chinese firms.
The World Economic Forum (WEF), a global capitalist think tank, published a report, Blueprint to Action: China’s Path to AI-Powered Industry Transformation, in January 2025 – before the release of DeepSeek r1. It looked at the AI industry in China, then valued at $70 billion, and the ways in which state-directed planning has taken advantage of technological advances. While admitting that “China’s model may not be universally applicable”, it includes case studies to influence capitalist governments to learn from China’s successes. However, many of the examples given show that the lessons are only applicable to the unique organisation of China’s society. A strong state with influence at every level and the ability to dictate to the capitalists how and what they should invest means that new technology is able to be implemented in more ways than in the anarchic production of the rest of the capitalist world.
One example which shows this is Guangzhou province’s ‘vehicle-road-city’ model, linked to the country’s ‘vehicle-road-cloud’ model. Urban planning and city infrastructure building is being done with the integration of new technology. Roads and streets are being designed with the collection of large amounts of data and with smart and autonomous vehicles and the necessary links to cloud computing in mind.
Similarly, transportation projects have been designed with the deployment of autonomous freight systems in mind. 5G internet connections, data connections, electrical power systems are being planned and integrated with projects centrally, as opposed to the responsibility of various competing privatised entities. The report includes examples of the ways in which AI has been implemented in increasing efficiency of supply chains and logistics, through the power of the state and regulations.
However, the direction AI is being utilised in China is in the interests of the CCP leadership, not the masses of working-class people. One example given in the WEF report, which other capitalist countries may well seek to emulate, is the push to integrate AI into retail. Data is used to serve up hyper-targeted ads and generative AI can create virtual ‘hosts’ of online shopping channels, popular on apps such as Douyin. These are aimed at developing the internal market in China, to make its economy less reliant on exports and vulnerable therefore to protectionist measures.
DeepSeek has been deployed in a few hospitals to take the place of doctors diagnosing patients, but this has been dismissed as a gimmick and won’t address the issue of medical debt, a huge burden on many workers in China.
DeepSeek has shown in a limited way that there are massive improvements in the capabilities and energy efficiency of the currently wasteful AI models. It has also given a glimpse of how the real potential of science and technology could be unlocked if capitalism were replaced by a system of democratic public ownership and planning of the economy at a national and international level. Instead of being used to run machines of war, or to increase the efficiency of the production of unnecessary goods or services, with the working class in power, and with international cooperation, AI and other scientific and technological developments could be set to use improving the living standards of all, including through the protection of the environment.
Mark Best