The global tech sector has been shocked by the release of DeepSeek’s R1, China’s potent new open-source AI model. It was offered for free and without any royalty fees, disrupting financial markets, and reversing Silicon Valley’s tightly regulated business model, which has long been criticized for its dominance of artificial intelligence.
DeepSeek’s open-source release is widely regarded as a key factor in the US’s trillion-dollar tech sell-off, which raises significant investor concern about China’s expanding competitiveness and the commodification of AI. R1 has irked investors and altered global AI geopolitics, earning itself the title “China’s answer” to OpenAI’s GPT4.
Reports show that using Nvidia’s H800 chips, R1’s compute costs were less than $6 million. This indicates a significantly more cost-effective model than proprietary counterparts, even though full development costs are still unnamed. It suggests that R1 might have been constructed for as little as OpenAI’s GPT4 costs, which are rumored to have been in the hundreds of millions. Due to its cost effectiveness and open access, DeepSeek’s business model is distinctly disruptive.
Chinese companies like Alibaba are releasing the Qwen3 Embedding series freely, and Mistral AI from France (with LLM as the first reason) is a step up. If the US doesn’t adopt open-source strategies, it could lose ground. After all, the early internet giants, like Google and Facebook, used free, user-centric services (like Gmail and Maps) to promote adoption before making money.
Giving away time-saving tools seems counterintuitive in a field where secrecy is common and models frequently locked up tightly. Even so, OpenAI, who was a pioneer with GPT4, now appears cautious. The $500 billion Stargate Project, which was established to avert AI leadership, has received strong support from CEO Sam Altman. With the launch of a new shopping feature, the practical expansion beyond ChatGPT has been slow. Innovation has not yet been pushed by US rivals (Google Gemini, Meta Llama, and Anthropic Claude).
Initial US dominance was fueled by incremental gains, which were aided by export restrictions on Nvidia chips and other technology that slowed China’s advance. Jensen Huang of Nvidia warned that these restrictions might have a negative impact, causing China’s chip industry to be slammed and ultimately weakened US control.
China’s strategic workaround is legal, scalable, and collaborative across all at once. It resembles how external developers create Android’s success. Similar to the Google Play model, Chinese companies now use open-source ecosystems to refine and scale models without having to pay any upfront costs.
Yann LeCun, Meta’s lead AI scientist, described DeepSeek’s rise as an open-source triumph, not just China’s ascendancy over the US. The geopolitical stakes are still present: Free access undermines the monetisation of proprietary models. Commercial models lose market share if open-source achieves parity.
Both speed and scale are China’s industrial strengths. By saturated the market with affordable, capable models, it exerts pressure on rivals until only the most popular, widely used model remains valuable, which is then monetized via data, advertising, or premium add-ons, a strategy well-tested by Google and Facebook.
US investors are aware of this for a reason. Systemic concern is at play in the reported $1 trillion decline following DeepSeek’s release. Open-sourcing is another component of a national industrial strategy for China: it uses “AI for good” to subvert, dominate, and claim benevolent intent.
If US technology is freely available, global rivals, including Chinese companies, can repurpose and surpass it. It might also be the opposite.
China also has its limitations. How open-source models trained in that environment can adapt to the demands of global content posed by its stringent internet censorship regime? RedNote (Xiaohongshu), a Chinese social media app that recently attracted a lot of Americans fleeing a potential TikTok ban, has already seen this. Although the cross-cultural exchange has been largely positive, tensions have grown, particularly in Taiwan and Xinjiang, where content is being moderated and censored for sensitive topics.
Chinese AI models may suffer from these limitations as they attempt to win over foreign customers for legitimacy and credibility.
China can compete in open-source AI without having access to cutting-edge US chips, changing the way the world uses AI. Leaders in the US have begun to acknowledge that long-term AI dominance depends not just on proprietary control but also on adoption, accessibility, and innovation at scale, from Elon Musk’s Grok-1 to OpenAI’s evolving stance.
In the end, China’s strategy of reshaping the global playing field requires using the principles of openness and decentralization to reshape the landscape. This may not be the case with guarding models behind closed doors.
The greatest irony is that China’s so-called “socialist AI” strategy may lead to the next era of US tech dominance.
