DeepSeek Shuns US Chips for AI Model Launch

DeepSeek, a prominent Chinese artificial intelligence lab, has deviated from industry norms by withholding early access to its forthcoming flagship model, V4, from U.S. chipmakers. Traditionally, AI developers provide pre-release versions to companies like Nvidia and AMD for performance optimization. However, DeepSeek has granted this crucial early access exclusively to domestic suppliers, notably Huawei Technologies, allowing them several weeks to optimize their processors.

This strategic decision is perceived by some analysts as part of a larger initiative by the Chinese government to create disadvantages for U.S. hardware and AI models within China. It also occurs as U.S. officials express concerns that DeepSeek's latest model may have been trained on restricted U.S. technology, potentially violating export controls. DeepSeek's models have seen significant adoption on open-source platforms, fueling competition with U.S. AI labs and intensifying debates in Washington regarding chip exports to China.