2025 marked a transformative year for Chinese AI, highlighted by the rise of DeepSeek and its impact on the industry. DeepSeek's launch of the DeepSeek-R1 model in January disrupted expectations, particularly in Silicon Valley, where many began reconsidering the dominance of American models. With its Mixture-of-Experts architecture and backing from High-Flyer, a quantitative trading firm, DeepSeek showed that effective AI development could emerge from Chinese talent and capital. This laid the groundwork for a surge of open-source models from various Chinese companies, with nearly every notable release being open source. As restrictions on chip access eased, DeepSeek is well-positioned to continue its momentum into 2026. In contrast, the startup Manus initially generated excitement for its general-purpose AI capabilities but soon shifted its operations to Singapore. Following scrutiny from the U.S. Treasury over its funding, Manus opted to distance itself from its Chinese origins to access global markets and venture capital. This strategy reflects a broader trend among Chinese AI startups seeking international expansion, spurred by DeepSeek's success. The landscape is now crowded, with major players like OpenAI and ByteDance also entering the agent market. The ongoing U.S.-China chip war added another layer of complexity. Throughout 2025, both nations grappled with export controls and regulatory decisions regarding AI chips. The year saw a back-and-forth regarding the availability of advanced AI chips, with significant moments like Nvidia securing permissions to resume sales to China, only for the Cyberspace Administration of China to push for reduced orders. This tug-of-war highlights the delicate balance both countries must maintain as they navigate the competitive landscape of AI development.