US capitalism mainly depends on a set of pillars that include ownership, secrecy and profits. When someone builds something in the west so they by virtue hold the sole ownership and maintain darkness on the process of how they did it. Finally, they use all means to make maximum profits out if it. So, they will set a minimum value, then triple, quadruple and multiply it by billions. Investors then dream big and bet on it and eventually make profits, which is the ultimate goal of capitalism.
Little-known Chinese startup DeepSeek released its latest AI model on the very day when US President Donald Trump was sworn in for the second time, spreading a shocking wave among their rivals in the US and elsewhere, as well as threatening the technology world order, as to how the Chinese firms can thrive even in the middle of outside restrictions and regulations. It reportedly took only two months to build an open-source large language model using reduced-capability chips from Nvidia called H800s.
DeepSeek claimed their AI models do it on a par or better than industry-leading models like OpenAI’s ChatGPT in the US at a fraction of the cost, according to Reuters. Along comes the question of pouring billions of dollars on the advancement of AI in the US and other parts of the world.
DeepSeek-V3 model was reportedly developed under a budget of $6 million, while OpenAI is said to pour up to $1 billion for their AI models. Other US tech giants including Google, Meta, Amazon and Microsoft are spending billions of dollars AI development, thus, DeepSeek tests the notion that only huge investments can results cutting-edge AI.
How they did it then so cost efficiently? Chetan Puttagunta, general partner at Benchmark venture capital firm based in San Francisco, put the entire process in a simple way. “They can take a really good, big model and use a process called distillation. Basically you use a very large model to help your small model get smart at the thing you want it to get smart at. That’s actually very cost-efficient,” he told CNBC.
Another thing to point out is to set a preference to research over profits. But Western observers missed the emergence of "a new generation of entrepreneurs (in China) who prioritise foundational research and long-term technological advancement over quick profits,” Marina Zhang, an associate professor at the University of Technology Sydney, told BBC. “Having grown up during China's rapid technological ascent, they are deeply motivated by a drive for self-reliance in innovation.”
Deepseek's founder Liang Wenfeng, 40, who studied AI at Zhejiang University, also is one such example. In an interview with Chinese media outlet Waves in July 2024 shortly after DeepSeek's V2 model rose to fame, Liang Wenfeng said, “We believe that both AI and API services should be affordable and accessible to everyone.” He also insisted on keeping DeepSeek as an open-source platform – a contrast to the west secrecy model over AI developments. DeepSeek’s R1 code is publicly available under the MIT License where its nemesis ChatGPT operates on a closed model. Hence, developers can freely use, modify, and distribute the code of DeepSeek, allowing for greater innovation and customisation.
Leaders in the technology sector, however, embraced the release of DeepSeek’s accomplishments. “To see the DeepSeek new model, it’s super impressive in terms of both how they have really effectively done an open-source model that does this inference-time compute, and is super-compute efficient,” Microsoft CEO Satya Nadella said at the World Economic Forum in Davos, Switzerland.
Jack Clark, co-founder of the AI startup Anthropic, wrote in his newsletter that DeepSeek’s R1 model “challenges the notion that Western AI companies hold a significant lead over Chinese ones.” Legendary venture capitalist Marc Andreessen called it “AI's Sputnik moment.” Cheng Lu, a research scientist at OpenAI, said DeepSeek’s chatbot performed impressive Chinese conversational skills. “It’s the first time I can feel the beauty of Chinese language created by a chatbot,” he said in an X post.
This is not the first time Chinese startups rivaled US competitors on cost efficiency. Kai-Fu Lee, the founder and head of 01.ai, said his company trained one of its advanced AI models using 2,000 GPUs with just $3 million. “The thing that shocks my friends in the Silicon Valley is not just our performance, but that we trained the model with only $3 million and GPT-4 was trained with $80 to $100 million. GPT-5 is rumored to be trained with about a billion dollars. […] We believe in scaling law, but when you do excellent detailed engineering, it is not the case. […] As a company in China, first, we have limited access to GPUs due to the US regulations [and a valuation disadvantage compared to American AI companies],” he said.
What are the lessons then?
The Chinese state maintains a ‘great wall,’ especially in terms of technological advancements, to the rest of the world, but some Chinese firms have started making expectations even though their vision are no out of questions. Yet, it is a good sign. Chinese firms often outbid European and US competitors by cutting prices and manufacturing in large scales. While “big US business is inherently anti-competitive by nature, is monopolistic by choice, and is, in most cases, very likely to be grossly inefficient as a consequence, with the costs being borne by the consumer,” according Richard Murphy, professor of Accounting Practice, Sheffield University, UK. But DeepSeek challenges the foundations of modern US capitalism in many ways. So, what are the ways out? As Satya Nadella puts, “We should take the developments out of China very, very seriously,” so both the US and China must embrace each other in technological advancements in future for the betterment of themselves as well as, for the rest of the world.