China’s AI race just took another sharp turn, and this time it’s coming from DeepSeek. The Hangzhou-based company has rolled out a preview of its much-anticipated V4 model, and while it’s still not the final version, the message is already loud — China isn’t just catching up in AI anymore, it’s trying to redefine the pace. What makes this launch even more interesting is the shift behind the scenes, especially its deeper collaboration with Huawei.
Unlike its earlier models that leaned heavily on chips from Nvidia, the new V4 is being positioned around Huawei’s ecosystem. That’s a big strategic move, not just technical. It reflects how Chinese companies are slowly stepping away from U.S. hardware dependencies, especially after restrictions tightened in recent years. Huawei confirmed that its Ascend AI chip lineup now fully supports the V4 series, hinting that this partnership isn’t just experimental — it’s long-term thinking.
From a performance point of view, DeepSeek is making bold claims. According to the company, the V4 Pro version is already outperforming most open-source AI models in world knowledge benchmarks. The only model it admits still sits ahead is Gemini-Pro-3.1, and that too belongs to a closed ecosystem. There’s also a lighter “flash” version of V4, clearly designed to push affordability — something that helped DeepSeek grab attention earlier this year in the first place.
But this launch isn’t happening in a vacuum. It lands right in the middle of rising geopolitical friction, especially between the U.S. and China over AI dominance. The White House recently accused China of systematically taking intellectual property from American AI labs, a claim that directly puts companies like DeepSeek under the spotlight. Firms such as OpenAI and Anthropic have also raised concerns about model “distillation,” suggesting their proprietary systems may have been indirectly replicated.
DeepSeek hasn’t fully addressed all of those allegations. It has acknowledged using Nvidia chips in the past but stayed vague on whether those chips were subject to U.S. export bans. The company maintains that its earlier V3 model relied on naturally sourced data through web crawling and not on synthetic outputs from competitors. Meanwhile, China’s official stance has been to reject the accusations outright, calling them baseless and reiterating its commitment to intellectual property protection.
The timing here matters. With a major U.S.-China leadership summit approaching, developments like this could add tension to already sensitive negotiations. AI is no longer just about innovation — it’s becoming a key piece in global power dynamics, and launches like V4 only intensify that narrative.
Back home in China, though, the impact was immediate. The release of V4 reportedly shook investor confidence in rival firms, with companies like Zhipu AI and MiniMax seeing their shares dip after the announcement. That reaction says a lot about how seriously the market is taking DeepSeek right now.
Behind all of this is an even bigger ambition. Backed by High-Flyer Capital Management, DeepSeek is reportedly aiming for a valuation above $20 billion. There are also talks of potential investments from giants like Alibaba and Tencent, which, if they materialize, could further strengthen its position in the AI race.
Right now, V4 is just a preview, but it’s already doing what it was supposed to do — spark conversation, shake competitors, and send a signal globally. Whether it truly lives up to the hype will depend on its final release, but one thing is clear: DeepSeek isn’t playing safe anymore. It’s playing big, and the world is watching closely.
