DeepSeek, the Chinese AI startup that wiped nearly $600 billion off Nvidia’s market value in a single day with launch of its R1 model, is reportedly ready for the unveiling of its next AI model. According to a report by The Financial Times, the company is preparing to launch its next major model, V4, as early as next week. The report says that the release marks a high-stakes moment for China’s tech industry as it tries to prove yet again that it can keep pace with US giants despite strict American limits on high-end computer chips.
What is DeepSeek V4 and it may be ‘problem’ for US companies
Unlike previous versions, V4 is expected to be “multimodal” – which means that it won’t just process text; it will be able to understand and generate images and videos as well – like Google Gemini 3. 0. This is the first major move from DeepSeek since January 2025, when its “R1” model shocked the world. The reason was DeepSeek’s claim that it had built an AI as smart as top Silicon Valley models but used only a tiny fraction of the power and cost.
This claim caused investors to panic, fearing that Nvidia’s expensive chips might no longer be the only way to build powerful AI – sending Nvidia’s stocks into a ‘freefall’. DeepSeek has reportedly optimised V4 to run on Chinese-made hardware from companies like Huawei and Cambricon even as American giants rely on Nvidia chips – a somewhat similar scenario.
By using local chips, DeepSeek is bypassing US export controls that prevent China from buying the latest Nvidia and AMD processors – a move that is expected to accelerate China’s transition away from Western technology for “inference” – the process where an AI actually generates a response for a user.
DeepSeek controversy and competition
DeepSeek’s rise hasn’t been without drama. Last week, the US AI company Anthropic accused DeepSeek of “distillation attacks” – a practice where a company uses the answers from a more advanced AI (like Anthropic’s Claude) to “teach” its own smaller model how to behave. The report also noted that while DeepSeek was previously praised for being “open” and sharing its secret engineering techniques, it is expected to be more guarded this time, releasing only a short technical note alongside the V4 launch.
