DeepSeek R1 and the Nvidia Crash

2025

Date: 27 January 2025 · Place: Hangzhou-based DeepSeek, then the Nasdaq · Figures: Liang Wenfeng, DeepSeek and High-Flyer Capital founder · Jensen Huang, Nvidia CEO · Sam Altman, OpenAI CEO · short sellers everywhere · Tags: AI, China, $600B Wipe 01. Open Weights From Hangzhou On 20 January 2025 the Hangzhou hedge-fund spinout DeepSeek releases R1, an open-weights reasoning model that matches OpenAI's o1 on multiple benchmarks. The reported training cost is roughly 5.6 million dollars. By the morning of 27 January the DeepSeek app overtakes ChatGPT atop the US App Store. Nvidia shares fall 17 percent that day, wiping 589 billion dollars in market cap in a single session, the largest one-day loss for any US stock in history. The thesis that frontier AI requires unlimited GPU spend wobbles publicly for the first time. 02. The Meme Aftermath The cycle generates a week of memes about American AI labs explaining to investors how a Chinese hedge fund matched their two-billion-dollar training runs with a side project. Sam Altman concedes DeepSeek is impressive. The whale logo, the lower-case branding, and Liang Wenfeng's quiet Chinese-press interviews become the canonical iconography of the moment. "DeepSeek's r1 is an impressive model." — Sam Altman on X, 28 January 2025 Sources • Reuters, 'Nvidia sheds $589 billion in record one-day loss' (27 January 2025) • The Wall Street Journal, 'DeepSeek Shocks Silicon Valley' (27 January 2025) • DeepSeek, R1 release announcement (20 January 2025)