Artificial intelligence will play a critical role in shaping China's fortunes as a great power. Yet Beijing's attempt to translate wins by today's innovators like DeepSeek and others into wider gains for the $18 trillion economy will be challenged not just by the United States but also by the Chinese Communist Party's own desire to maintain control.

Technological revolutions have underpinned history's great power transitions. Just as the United States supplanted the industrial dominance of the British empire in the 20th century through widespread adoption of electricity, machinery and motor vehicles, AI offers a similar opportunity for China to shift the balance of power in its own favour.

Whether Chinese leader Xi Jinping aims to displace America in technology or economic or military terms is an ongoing subject of debate in Washington.

Mike Waltz, national security adviser to the US President Donald Trump, reckons America and China are in a cold war, though Beijing's 2017 "New Generation Artificial Intelligence Plan" simply calls for the People's Republic to establish itself as a world AI leader and innovator by 2030.

If there is a battlefield, Hangzhou-based DeepSeek is opening a new front. Its AI training models are groundbreaking and perform as well as those of Western rivals like OpenAI at a fraction of the cost. It's the latest sign, along with similar offerings from Alibaba, Tencent  and peers, that China can lead on innovation to close the gap with the $27 trillion US economy.

To that end, Xi is mobilising his corporate troops. In a rare and highly choreographed meeting last week he urged the chiefs of homegrown innovators such as BYD the world's top electric vehicle maker, and telecoms conglomerate Huawei, which is leading China's efforts to develop high-end chips to rival California-based Nvidia, to "show their talent". The successes of DeepSeek and Xi's show of support for private sector has pushed Hong Kong's Hang Seng Tech Index up by more than a quarter since the start of the year, outperforming the Nasdaq.

Yet sustaining its AI advances will require a big effort. Like the United States and others, China needs to invest huge sums to build out physical infrastructure – including for power generation. It will also need to keep overcoming Washington's latest measures blocking its access to cutting-edge chips and chipmaking equipment.

The next hurdle may prove to be the hardest to clear, though. For gains to be broad-based, China will need the technology to be adopted and integrated across industries. Here, China may run into a "diffusion deficit", a term used by Jeffrey Ding, a US-based political science professor.

In his recent book, "Technology and the Rise of Great Powers: How Diffusion Shapes Economic Competition", Ding argues that countries with weak "skill infrastructure" - institutions that help widen the base of new skills - tend to struggle to implement and disseminate innovations across an array of sectors.

He concludes that policymakers in Washington are too focused on innovation, rather than diffusion, and may be overestimating China's rise as science and technology superpower.

The People's Republic, for example, boasts one of the largest pools of engineers and scientists in the world but fourth-year computer science majors in the United States "substantially" outperform peers in top universities in China. Ding's analysis of Japan during the 1980s and 1990s provides a cautionary tale on how this could morph into a big disadvantage.

Japanese firms once controlled global computer and chip production, but the country suffered from a talent gap with the United States; annual inflows of skilled workers into the American information, communication and technology labour pool outpaced those in Japan by 68 percent in 1995; by 2001, this gap had widened to 300 percent. This, among other things, contributed to low growth and poor productivity in many sectors.

Area chart showing China's patent applications have surpassed the US by far

Indeed, Chinese businesses are laggards when it comes to adopting technologies. According to a domestic survey last year, more than 60 percent of 500 small and medium sized enterprises polled are only in the "early" stages of digitisation, using basic data management and IT applications. Less than a fifth of Chinese companies that responded to a separate survey said they had integrated generative AI into their business processes, trailing the United States by five percentage points.

Some of the diffusion problem can be explained by the Chinese Communist Party's desire to protect itself.

China has access to vast amounts of data required to train AI models but the People's Republic requires these models to "adhere to core socialist values" and exclude content that "endangers national security". Strict censorship may have already slowed the country's progress: In 2017, Tencent shut down an AI-powered chatbot installed on its messaging app after it uttered a political faux pas.

Meanwhile, China's top-down policy of identifying specific industries to back has produced some winners but it has also resulted in many inefficient enterprises that often have little incentive to take risks or try out new technologies; industrial planning has also led to a massive misallocation of capital. Between 2009 to 2018, growth in China's total factor productivity - an indicator of an economy's efficiency and competitiveness - significantly slowed to 0.7 percent, from the 2.8 percent in the decade before the global financial crisis, a sharper deceleration than world productivity growth.

Analysts at Goldman Sachs estimate AI could lift productivity growth in China by 8 percent over the next decade. That lags the United States, where the bank estimates AI will unlock a 15 percent boost or roughly $4.5 trillion of annual GDP gains. The difference comes down to diffusion: half of China's jobs are in manufacturing, construction and other sectors less likely to use AI, per Goldman.

Of course, it is still early days in the field and Beijing's big push to develop other parts of its economy means China could benefit much more from AI than is currently envisioned. To go beyond the innovation phase, though, China probably needs to give its schools and private companies a freer hand.