How China’s DeepSeek AI chatbot made Meta’s CEO Mark Zuckerberg ‘rethink’ Meta’s AI strategy, and led to a talent war

Chinese AI startup DeepSeek has emerged as an unexpected challenger in the race toward artificial general intelligence (AGI), drawing significant attention not only across Asia but also within Silicon Valley. Its model, DeepSeek-V2, reportedly prompted Meta CEO Mark Zuckerberg to reassess aspects of the company’s AI roadmap, according to The Information.

Meta executives were reportedly surprised when DeepSeek-V2, an open-weight language model developed in China, began outperforming several U.S.-backed models in key benchmarks. After being briefed on the model’s efficiency and performance, Zuckerberg is said to have asked Meta’s AI team to reevaluate how its foundation models are trained and deployed.

Citing sources familiar with the matter, the report notes that Zuckerberg emphasized the need for cost-effective scaling and localization—areas where Chinese players like DeepSeek are showing growing strength.

Intensifying the AI talent war
In response to the competitive threat, Meta has reportedly increased hiring for its AGI-focused team, recruiting top talent from rivals like Google DeepMind and Microsoft. This recruitment drive gained urgency after DeepSeek’s open-weight model drew international praise for balancing high performance with compute efficiency.