On the day of Donald Trump’s presidential inauguration, a small Chinese tech company founded in only 2023, DeepSeek, shocked the AI industry, rattled US financial markets, and created a truly global stir. Notwithstanding controversies about cost and intellectual property surrounding what is still a genuine and significant accomplishment, it is fair to say that DeepSeek’s commoditisation of AI is exactly how transformational technology happens. Innovation and competition drive down costs, leading to the technology becoming embedded in the whole economy, not just the narrower, innovative parts.

These AI industry implications are centre-stage for now. It has also been argued that DeepSeek’s model represents not so much a ‘Sputnik’ moment as a case of hyperbole, and that the consequences of the AI arms race will throw up many surprises. Not the least of these is China’s persistent dependency on semiconductor technology developed by the US and its close allies, heightened by tariffs and export controls, which Trump’s administration looks set to tighten. American tech companies will want to exploit this in developing artificial general intelligence themselves.

Wider economic and geopolitical considerations are equally important. DeepSeek is, after all, part of China’s dichotomous economic state. On the one hand, China has many dynamic, modern sectors. Several firms and industries are in the vanguard of scientific and technological development in clean energy, electric vehicles, and batteries, as well as other so-called strategic emerging industries in which China seeks to dominate and become self-reliant. These include industrial machinery, semiconductors and computing, artificial intelligence and robotics, life-science industries, biotechnology, and pharmaceuticals. These industries, which, in Xi Jinping’s Marxist framing are called ‘new productive forces’, accounted for about 13 per cent of GDP in 2022. Beijing’s heavy emphasis on industrial policy and exports, and its uniquely large state-assistance programmes aim to push this proportion up.

At the same time, over 80 per cent of China’s economy is in the doldrums. The lion’s share of this part of the economy, roughly two fifths, comprises real estate and infrastructure, both of which feature over-expansion, excess supply, shrinking demand as the working age population falls, and severe financial problems among heavily indebted local and provincial governments. In this part of China’s economy, the principal features are slowing economic growth, stagnating productivity, and the misallocation and inefficient use of capital, as well as weak household income, consumer demand and employment.

The dichotomy is important because the vibrant part of the economy, typically, is relatively small. The key to economic success and rising productivity over time rests not only on innovation but also on flexible and inclusive political institutions and incentives, and laws and regulations that help to diffuse and embed innovation and efficiency in the rest of the economy. The repressive effects of China’s ideological and political party-state rule and its institutions are likely to be amplified rather than resolved by technology.

Although DeepSeek, for example, is a private firm, funded principally by a hedge fund that is also owned by its owner, Liang Wenfeng, it is certainly not free of political influence or interference. Its AI model propagates Chinese government narratives and arguments on sensitive topics like President Xi himself, Taiwan, Tibet and Tiananmen; it is fully aligned with the Chinese government’s data and information restrictions and policies, and it is expected to respect Xi’s dictum that ‘government, military, civilian, and academic, east, west, south, north, and centre, the party leads everything’.

While the previously stern political rhetoric about the political duties of private firms has softened with growing economic discomfort, private enterprises are still constrained by national security and anti-espionage regulations. These often conflate the collection, use, and transfer of information by businesses with misconduct and, sometimes, criminality.

DeepSeek is nonetheless, like other modern firms, an innovator, following a quite familiar script in which China embraces technological advances, often pioneered in the West, but scales production up to exceptional levels and at significantly lower cost. This has happened in various sectors from steel and shipbuilding, to solar, wind and other climate change mitigation equipment, as well as electrolysers (used in hydrogen technology), batteries and electric vehicles, and now AI.

Yet, there is much in the word ‘innovation’ that is glossed over. China’s strengths lie more in product than process innovation, and in efficiency solutions than the pushing out of technological frontiers. China’s mixed innovation performance is also highlighted by the World Intellectual Property Organisation’s Global Innovation Index. China currently ranks eighth out of 132 nations in so-called innovation outputs, such as intellectual property, patents, software spending, high-tech exports, and online creativity, but only 25th in innovation inputs. These inputs include, for example, the quality of regulatory, legal, and business institutions, educational attainment, research and development, and business and market sophistication.

There is no question, though, that DeepSeek’s catching up with its US rivals means the race for AI dominance has been joined, adding to tensions already created by China’s industrial policies, and posing important geopolitical challenges. Trade tensions and conflicts are more likely to increase than not. China’s attempts to expand its already high 33 per cent share of global manufacturing can only be realised by forcing a contraction in manufacturing and trade in other countries such as the US, Europe, and also important parts of the so-called Global South, such as India and Brazil. Pushback is both happening and inevitable.

AI is now widely seen as the new general-purpose technology in which there are vigorous economic incentives for competition and leadership. China would seem to have the advantage on price and pitching functional AI on affordable hardware to poorer countries, America on pushing the technological frontier and building artificial general intelligence models. Export controls and restraints as well as sanctions are likely to remain key tools of policy for the US.

Questions have been posed as to whether the Biden administration’s AI export controls, introduced in October 2023, were effective. There seems little doubt that they shaped DeepSeek’s strategy to rely on less powerful chips and seek greater efficiencies. Nonetheless, DeepSeek still relied on Nvidia chips and imported lithography machines. As Liang Wenfeng said recently, noting the still major gap between US and Chinese computing power, ‘our problem has never been funding: it’s the embargo on high end chips’. That looks likely to stay, and if anything, get tighter, placing global supply chains under further duress.

While all of this plays out on the global stage, we should keep a close eye on how China manages its economic challenges domestically because product innovation and cost suppression can only get you so far: to be economically transformative, they require political and institutional flexibility and reform. These do not figure on Xi’s agenda, and Japan offers a useful point of reference as to why they matter. A leading US academic, writing in 1987, said that Japan’s 

strategy will result in spectacular advances and growing supremacy in a variety of fields such as industrial ceramics, lasers, semiconductors, biotechnology, solar energy, robotics, superconductors and possibly in space exploration. These advances, in turn, will be largely used in consumer products and will lead to increasing exports, rising techno-nationalism, and deepening fears among Americans that we can no longer compete.

This passage could just as easily have been written today about China. Japan remains a rich country, but its share of world GDP has declined significantly over the last 30 years, from a position more or less similar to the one that China occupies today. It failed to live up to the hyperbole, despite its world-beating brands and industries, because of the inability to address deep-seated systemic economic flaws, misguided policies, and key institutional weaknesses. Many of these factors are present in China today. 

The lesson is that industrial world-beating firms and innovative flair cannot compensate for systemic economic imbalances, asset bubbles, political contradictions, and institutional rigidities. Technological islands of excellence – of which DeepSeek is a prime example – are wonderful to have. Ultimately, however, robust macroeconomic governance and well-institutionalised technology ecosystems that diffuse benefits throughout the economy matter just as much, if not more.