China's AI Advancements Intensify Global Competition

Chinese AI startup DeepSeek has released a preview of its V4 large language model, expanding its open-source AI lineup. The model is available in 'pro' and 'flash' versions and is claimed to achieve strong performance against domestic competitors. DeepSeek's earlier R1 model disrupted markets with its strong performance at a lower cost, intensifying pressure on global AI spending strategies. The release of V4 has implications for the technological war between the U.S. and China, with DeepSeek reportedly not sharing its upcoming AI model with American engineers. The model's ability to run on local chips, including Huawei's Ascend AI processors, could help Beijing achieve more AI sovereignty and reduce reliance on Nvidia.

The Trump administration is planning to crack down on Chinese firms accused of exploiting U.S. artificial intelligence (AI) models. A memo from Michael Kratsios, the president's chief science and technology adviser, alleges that Chinese entities are engaging in industrial-scale campaigns to extract capabilities from leading U.S. AI systems. The administration plans to work with American AI companies to identify such activities, build defenses, and punish offenders. This move comes as China is narrowing the gap with the U.S. in the AI race, with some Chinese AI models able to compete with U.S. giants at a fraction of the cost. The U.S. House Foreign Affairs Committee has also shown bipartisan support for a bill to identify and punish foreign actors that extract key technical features of U.S.-owned AI models.

DeepSeek, a Hangzhou-based AI start-up, has released its next-generation foundational artificial intelligence model, V4, in two versions: V4-pro with 1.6 trillion parameters and V4-flash with 284 billion parameters. Both models have a context window of 1 million tokens, achieved with 'world-leading' cost efficiency. Huawei has announced 'full support' for DeepSeek's V4 models with its Ascend chips and supernode systems. The release is expected to improve the capabilities of domestic graphics cards and their adoption. While V4-pro is too large to run on consumer-grade hardware, the technical report and V4-flash model are likely to benefit global AI developers.