Just one week after United States President Donald Trump boasted in his inaugural address of US leadership in the Artificial Intelligence (AI) revolution, stock market shares in US tech giant Nvidia plunged 17%, from a share valuation of US$3.5 trillion to US$2.9 trillion.
This was the biggest single-day fall in value for a company in history and followed the meteoric rise of a Chinese open-source AI tool, DeepSeek, which matches the performance of Open AI’s iconic ChatGPT.
DeepSeek claims to have achieved this benchmark with fewer and less powerful chips and a smaller drain on energy.
This has disrupted the Trump administration’s AI development plan, which saw Nvidia reach its stock market high after Trump announced a US$500 billion AI infrastructure project called Stargate on January 21.
According to US journalist and data scientist Karen Hao, Trump’s AI plan is based on “scaling up” through building more giant data centres, at a huge environmental cost.
In a long thread posted on X, Hao wrote, “the biggest lesson to be drawn from DeepSeek is the huge cracks it illustrates with the current dominant paradigm of AI development”.
She argues that the US model is driven by an argument “based [more] in business than in science”.
The fossil fuel industry has a major interest in pushing this model.
It is no secret that apart from the tech and media oligarchs, other big billionaire beneficiaries of the Trump administration include the 15 biggest fossil fuel billionaires in the US, whose combined wealth has already risen by US$17 billion since the beginning of the year, according to the Climate Accountability Research Project.
Construction of the first of 10 giant data centres in Texas began during Joe Biden’s administration to keep the US ahead of China on AI.
Biden’s administration also used sanctions since 2022 to stop China from acquiring advanced computer chips and the technology to manufacture them. However, this seems to have backfired, according to Reihana Mohideen, an electrical engineer in energy and power systems and a Just Energy Transition expert from the University of Melbourne.
“Chinese AI companies have been forced to innovate due to the restrictions imposed on them by the US,” Mohideen told Green Left. “They have done this by prioritising efficiency, resource-pooling and collaboration. Rather than weakening China’s AI capabilities, US sanctions have spurred innovation in China’s AI sector, with brilliant startups like DeepSeek emerging.
“China is a world leader in AI research. It has world-class academic institutions, in particular Tsinghua University, which has produced the majority of its top AI start-ups. The founder of DeepSeek, Liang Wenfeng, with a background in information and electronic engineering, is an alumnus of Zhejiang University — one of the top universities in China for AI.
“US researchers are marvelling at the DeepSeek-R1 (DS-R1) model’s engineering simplicity, focusing on accuracy rather than detailing every logical step, and hence also saving on computing time and increasing efficiency. According to Liang, the goal is to continuously close the efficiency gaps, in computing power and data.
“The imperative to do more, with fewer resources, is driving collaboration and cooperation in the sector, with moves underway to combine research teams and labs.”
Mohideen said that AI innovation in China is fostered by significant government financial support. AI start-ups supported by prestigious universities are not simply replicating existing technologies, but pushing the boundaries of AI innovation, strengthening China’s ability to produce competitive AI models.
“DS-R1 demonstrates that sometimes these models are even superior, driven by high-level innovation.
“China describes the system that enables all this as ‘socialism with Chinese characteristics’. It’s interesting that they don’t backtrack on this, despite the level of capitalist penetration and the existence of a capitalist class in the country. They consider their system to be different from capitalism, which would have an impact on the young generation of scientists.”
However, while China is challenging the US in AI research and even applications, the level of investment in AI in the US is much higher — in terms of the number and value of investments, Mohideen said.
Apart from prioritising efficiency, a young generation of Chinese researchers are committed to an open-source culture, as they themselves have benefitted from it, Mohideen said.
“The DS-R1 code is open source. This means it can be further developed by those who haven’t been able to work with really large models, and its claims can be tested and verified.
“Because it requires less computational power, the cost of running DS-R1 is much less than the cost of similar competitors. DeepSeek is charging people using its interface around one-thirtieth of what OpenAI’s [Chat GPT] o1 costs to run, according to Nature magazine. The firm has also created mini ‘distilled’ versions of DS-R1 to allow researchers with limited computing power to play with the model.
“Using less computational power means better energy efficiency, or achieving the same results with a fraction of the energy use. This has important implications for reducing environmental impacts in an industry that is raising major concerns about its energy consumption, computational power and the water needed for cooling.
“It is AI with a smaller carbon footprint.
“The DeepSeek model has shown that the perceived US lead in AI has dwindled significantly. It’s wonderfully disruptive and a blow to Trump’s ‘drill baby drill’, AI-driven, MAGA [Make America Great Again] agenda.
“It shows that innovation and collaboration can be more important than how much money AI companies raise, the measure of tech companies working in AI. This is an important and inspiring lesson for the Global South.”