Headlines seized upon DeepSeek’s modest $6 million training cost, casting doubt on mega-projects like OpenAI’s $500 billion development venture known as Project Stargate. Frantic investors dumped shares of Nvidia and American tech stocks, driving a brief but astounding $1 trillion market drop. Meta and OpenAI would later capitalize on the panic, urging the White House to block state-level AI regulations and approve training on copyrighted material.
Wenfeng, who hails from a village in southern China, had already proven a knack for boldly undercutting his rivals: months before DeepSeek’s global ascent, his firm had sparked a domestic price war by releasing a language model so aggressively priced it forced Chinese tech giants like Baidu and Alibaba to slash rates by more than 95%.
But while R1 was impressive—especially for a company that spun out of Wenfeng’s trading firm a short two years earlier—the narrative surrounding it relied on selective information. The $6 million figure, for instance, ignored the roughly $1.6 billion DeepSeek spent to assemble its cluster of computing chips, according to research firm SemiAnalysis—not to mention the seven-figure salaries the company reportedly pays its top researchers. More tellingly, the frenzy conveniently overlooked that a Google DeepMind model released one month earlier offered similar performance for an even lower price—part of a wider trend that has seen the cost of AI fall precipitously. Nonetheless, Wenfeng’s formidable self-mythologizing—and ability to bring competitors’ backs against the wall—has proven to be a powerful market-mover.