In January 2025, Chinese AI company DeepSeek released its low-cost, high-performance reasoning model and briefly shook global markets by wiping out hundreds of billions from US technology stocks. It also massively challenged assumptions about the development costs of AI. DeepSeek’s open-weight model offered similar performance to leading AI tools but at a fraction of the usual cost, raising fears that traditional, hardware-intensive approaches could soon be obsolete.
The initial shock on Wall Street has largely subsided as leading US technology companies have doubled down on their massive investments in computing power. Nevertheless, DeepSeek has reshaped Chinese AI, advanced the spread of open source, and highlighted the different AI strategies of the US and China.