Electricity will decide the fate of the AI race: How China is turning energy into intelligence
The outcome of the next stage of the global race for supremacy in the field of artificial intelligence will be determined not by algorithms and semiconductors, but by electricity. And here China has a clear advantage. Western tech giants emphasize closed, capital-intensive models that require huge computing power. China has opted for open-source AI models and is actively building renewable and nuclear power capacity, positioning itself for mass adoption of powerful AI technologies, but without the high costs. Jeffrey Wu, Director of MindWorks Capital in a column for Project Syndicate writes that these differences are fundamental.
A matter of public importance
The US and its allies see AI as a private technology, while China sees it as a public infrastructure. It is building an open AI ecosystem with the same philosophy as industry: mass adoption, rapid iteration, and continuous cost reduction. China's open source models, including DeepSeek, Qwen, and Kimi, are not just scientific advances. They are strategic tools for collective participation, and they are changing the economics of AI.
The latest version of DeepSeek reportedly has the capabilities of advanced systems like those in the US, but its cost of computing is much lower. Prices for Qwen and Kimi application services have fallen by orders of magnitude. To put it in the language of economists, the marginal costs of "thinking" are coming down. The cost of operation -inference- of a number of Chinese models is about 10 times lower than GPT-4 from OpenAI.
But the cheaper AI becomes, the more the world consumes it: every token saved generates a thousand new tokens. A dynamic that was once a driving force in the coal age is still seen today - in the digital age. In China, this is a deliberate strategy: the low cost of inferencing and open access to the parameters of Chinese models are meant to incentivize experimentation in universities, startups, and the authorities on the ground. However, all this activity requires energy: the International Energy Agency expects that by 2030 the world's electricity consumption by data centers will double - relative to 2024. This will happen mainly because of AI. GPT-4 training alone probably required millions of kilowatt-hours - enough to power San Francisco for three days.
Competition in kilowatts
The old competition of algorithms is fast becoming a competition in kilowatts, and China is determined to win. In 2024, the country will bring 356 gigawatts of renewable energy capacity online - more than the United States, the European Union and India combined. Solar, wind and hydropower account for 91% of all new capacity. Energy storage capacity has tripled since 2021, and ultra-high voltage grids are transmitting clean energy thousands of kilometers - from deserts to data centers.
China is also actively investing in nuclear power. According to the Information Technology and Innovation Foundation (ITIF), Chinese spending on nuclear R&D is about five times that of the United States. Fourth-generation reactors and small modular plants are moving from the pilot project stage to the deployment stage in China. Nuclear power has quietly begun to provide baseload power that fickle renewables cannot.
The combination of open AI models, cheap renewable energy, and the stability of nuclear power are forming what can be called an "energy-computing" flywheel: more clean energy allows more computation, which in turn optimizes the grid. Machine learning systems are already predicting solar power generation, managing energy storage, and balancing load in China's giant power grid in real time. The result is a reorganization of the industry: the traditional boundaries between energy, semiconductors, and software are being blurred. Data centers are the new power plants, and GPUs are the new turbines. China is not just electrifying the industry, it is electrifying intelligence.
In addition to strengthening clean energy networks, China is exporting components of this new global energy system. In August, exports of the clean-tech sector -clean-tech including solar panels, grid batteries and electric vehicles - reached a record $20 billion, surpassing shipments of consumer electronics a decade earlier. The West may export chips and software, but it is China that produces the electricity that makes them work.
Meanwhile, energy constraints in the West - aging grids, slow licensing, high prices - are creating bottlenecks in the digital sector. In the U.S. and elsewhere, data center expansion is increasingly hampered by limited access to reliable sources of electricity. Some jurisdictions, such as Virginia and Dublin, have imposed moratoriums on new data center construction.
The new oil is clean energy plus AI
Industrial revolutions have always favored those societies that could convert energy into productivity in the most efficient way. In the nineteenth century, coal was the key to world domination. In the twentieth century, oil became the queen. And in the 21st century, that role will go to clean energy paired with computing power. Whoever has the cheapest energy will have the cheapest intelligence - and enjoy growing abundance on both fronts.
China is preparing to take this coveted position by building a system of investment and incentives. Democratic countries will not be able to replicate these steps quickly. But China is not alone in reaping the benefits of these successes. For developing countries that have long been denied access to high-performance computing because of high prices, open-web models and lower electricity costs can make AI affordable and even indispensable, much like electricity or broadband.
However, abundance does not guarantee stability. Without sufficient investment in clean energy generation and storage, a surge in energy demand due to AI could overload the grid and slow decarbonization. As in the Industrial Revolution, efficiency can lead to excesses and progress can be accompanied by increasing imbalances. Finding the balance between abundance and abstinence will determine whether AI becomes a tool to help empower people or a new driver of growing inequality.
Two hundred years ago, the steam engine turned heat into motion and changed the world economy. Today, AI is turning electricity into intelligence. Whoever masters both will be able to rewrite the rules of progress.
Copyright: Project Syndicate, 2025.
This article was AI-translated and verified by a human editor
