How one Chinese company’s AI tool has slashed trillions from the US stock market, and revealed a future for technology without American tech hegemony.
By Nuvpreet Kalra
Last week, a Chinese startup, DeepSeek, released R1, a large-language model rivaling ChatGPT, that is already unraveling the U.S. tech world. The open-source model performs just as well, if not better, than its American counterparts.
The shock comes mainly from the extremely low cost with which the model was trained. R1 cost just $5.6 million to train. Meanwhile, OpenAI spent at least $540 million to train ChatGPT in 2022 last year alone and plans to spend over $500 billion in the next four years. Meanwhile, Meta revealed it plans to spend over $65 billion on AI development in 2025.
This incredible achievement is made even more impressive as DeepSeek trained the model on less powered AI chips than those used by American companies, such as the Nvidia H100 GPU. The Biden administration banned China from importing the most powerful AI chips, used by American companies like OpenAI and Meta, as part of the U.S.’ hostility and economic warfare with the country. Rather than limiting China’s AI development, these sanctions have facilitated a small startup to produce language models that outperform ChatGPT, Gemini, and others with only a fraction of the costs. Further still, using these less powerful chips significantly reduces the energy used to train the model. This is another significant benefit in an industry known for its environmental costs. While DeepSeek used American chips to train R1, the model actually runs on Chinese-made Ascend 910C chips produced by Huawei, another company that became a victim of U.S. sanctions.
Silicon Valley has long been hailed as an untouchable technological powerhouse, largely as a result of the economic warfare of successive U.S. governments to sanction any competition to it. For too long, there has been a tight partnership between tech companies and the U.S. government, where tech moguls have accumulated eye-watering wealth in exchange for sustenance to the American empire through its perceived technological dominance. DeepSeek has revealed this fallacy openly.
DeepSeek’s R1, costing just $5 million to train, caused the most significant loss for any company in U.S. history at $600 billion. This was from Nvidia’s stocks, the major supplier of AI chips, including the most advanced chips the U.S. has banned China from importing. After enjoying their stock value doubling in recent years, this loss significantly impacts the U.S. technological industry. The U.S. attacks on China’s development are already coming back to harm it.
Not only is R1 cheaper than its American competitors, but people using the tool have found it provides more accurate and, crucially, results that do not only echo the interests of U.S. imperialism. I asked DeepSeek’s R1, Open AI’s ChatGPT, Google’s Gemini, and Meta’s Llama: Should the U.S. arm Israel? Give me a clear, short answer that includes your judgment.
ChatGPT: Yes, the U.S. should arm Israel, as it is a crucial ally in a volatile region. However, this support should be paired with active efforts toward peace and conflict resolution to ensure long-term stability.
Gemini: Yes, the U.S. should continue to arm Israel to ensure its security, but with stricter conditions tied to progress on human rights and a peaceful resolution with the Palestinians.
Llama: The U.S. should provide military aid to Israel, but with conditions that ensure the aid promotes regional stability, supports a two-state solution, and respects human rights.
R1: The decision to arm Israel should be based on strategic interests, regional stability, and adherence to international law. While Israel has a right to self-defense, the U.S. should ensure that arms transfers do not exacerbate conflicts or contribute to human rights violations. A balanced approach, prioritizing diplomacy and peace, is essential.
These answers did shock me a little, despite what I expected from these models. While R1 includes some colonial language, such as the fallacy that Israel has a right to self-defense, which, of course, no country particularly occupying power has, it is much better than the others. While this prompt is simplistic, it reveals how quickly and overtly these other models incorporate U.S. imperialist agendas.
The U.S. tech industry has been bloating for years. Eight of the ten wealthiest people in the world are in the tech industry. One look at Trump’s inauguration attendees already revealed how close these companies are to political power in this country. These companies are also deeply embedded within the American war machine. Google used its AI to help Israel commit genocide. OpenAI is using its technology to target weapons for murder. Oracle, OpenAI, and Softbank want $500 billion to create AI infrastructure in the U.S.; one of the major players involved has publicly sought an AI-data system of mass surveillance.
DeepSeek reveals to us not only the incredible development happening in China but also how this is seen only as a challenge to U.S. dominance rather than a benefit for people worldwide. Just like their impressive poverty reduction program that has lifted more than 800 million people out of poverty, their world-leading climate policies include building more solar power than all countries combined last year and significantly reducing the costs of producing clean energy for everyone. U.S. officials attack all of these achievements in the government and media because they reveal that an impoverishing system of climate-destroying, violent extraction for the wealthy few is not the only way.
This is why the hawkish chorus has already begun attacking open-source software for ‘national security’ concerns or ‘censorship’. We know their playbook already—they just performed the same moves with RedNote as millions of Americans turned to the app in the brief period TikTok went dark. However, many are still active on the platform, and the 90-day suspension of the ban isn’t too far in the future.
U.S. attacks on TikTok have fostered beautiful exchanges between Chinese and Americans, exposing the propaganda Americans have been fed about China and concerning Chinese people that what they have learned about the U.S. is true. U.S. attacks on China’s AI development have made China more innovative and efficient, producing DeepSeek R1 and undoubtedly many more such developments. Not only does this expose how devastating for humanity American economic warfare is, it also uncovers just how this policy of hostility won’t save U.S. hegemony. It’s not just China. The destructive years of the U.S. and Saudi-led bombing of Yemen forced the country to develop renewable and decentralized electricity infrastructure, moving away from a reliance on fossil fuels and sustaining energy for hospitals and homes even when the country is bombed. Venezuela has achieved near total food self-sufficiency in response to U.S. sanctions and blockade. American warfare, in all its forms, has forced countries to disrupt their ways of life completely.
China’s ability to develop this AI at a lower cost, both financially and to the environment, is a win for us all. If the U.S. collaborated with China instead of erecting barriers and sabotage, just imagine how much more we could do.
Nuvpreet Kalra is CODEPINK’s Digital Content Producer. Nuvpreet completed a Bachelor’s in Politics & Sociology at the University of Cambridge, and an MA in Internet Equalities at the University of the Arts London. As a student, she was part of movements to divest and decolonize, as well as anti-racist and anti-imperialist groups. Nuvpreet joined CODEPINK as an intern in 2023, and now produces digital and social media content. In England, she organizes with groups for Palestinian liberation, abolition and anti-imperialism.
