China’s DeepSeek AI Faces Scrutiny Over Bold Claims
Chinese AI startup DeepSeek has made headlines by unveiling its AI model, R1, which it claims was built for just $5.6 million using 2,000 Nvidia H800 GPUs. This announcement challenges the dominance of US tech giants like OpenAI and Google, which have spent hundreds of millions on similar AI models. DeepSeek’s claim also raises doubts about the effectiveness of US export restrictions on advanced AI chips.
Skeptics, however, question the feasibility of such a low-budget AI breakthrough. Some argue that DeepSeek may have had access to more powerful chips and hidden funding. OpenAI’s GPT-4, for example, reportedly cost over $100 million to train, using an estimated 25,000 high-end Nvidia H100 GPUs. Critics, including Oculus VR founder Palmer Luckey and billionaire Elon Musk, suggest that DeepSeek’s claims might be exaggerated or even part of a strategy to undermine US tech confidence.
Despite the skepticism, AI experts acknowledge that advancements in efficiency could lower AI training costs over time. Researchers point out that DeepSeek may have fine-tuned an existing model rather than building R1 entirely from scratch. While some see the achievement as groundbreaking, others remain cautious, waiting for further verification.
Meanwhile, DeepSeek’s rise has had financial impacts. Nvidia’s stock initially plunged 17%, wiping out nearly $593 billion in market value, though it later recovered. The controversy highlights the growing AI competition between China and the US, raising questions about the future of AI development costs and the real impact of export bans.
Whether DeepSeek’s claims are entirely true or not, one thing is certain—AI innovation is evolving fast, and the race to dominate this field is far from over.