China's AI powerhouse DeepSeek is preparing to launch its most ambitious model yet and it will run on chips made by Huawei, not Nvidia. According to a report by The Information, cited by Reuters, DeepSeek's upcoming V4 model has been optimized specifically for Huawei's latest Ascend processors, marking a decisive break from the Western chip ecosystem that has powered global AI development for over a decade.
The move signals more than a product launch. It is a geopolitical statement one that could reshape the global AI hardware landscape and accelerate the split between American and Chinese technology stacks.
Alibaba, ByteDance, Tencent Line Up
In preparation for V4's release, some of China's biggest tech companies have already placed massive orders for Huawei's upcoming chips. Alibaba Group, ByteDance, and Tencent Holdings have collectively ordered hundreds of thousands of units, according to five people with direct knowledge of the purchases cited in The Information's report.
The scale of these orders suggests strong industry confidence in both DeepSeek's new model and Huawei's ability to deliver competitive AI silicon a development that would have seemed unlikely just two years ago when Chinese chips were widely considered a generation behind Nvidia's offerings.
Months of Collaboration Behind the Scenes
This was not a last-minute pivot. DeepSeek has reportedly spent months working directly with Huawei and another Chinese chip designer, Cambricon Technologies, to rewrite portions of V4's underlying code for compatibility with domestic processors. Engineers optimized the model's architecture to run efficiently on Huawei's Ascend chips, ensuring that performance would not be sacrificed in the move away from Western hardware.
The collaboration represents the deepest integration yet between a leading Chinese AI lab and the country's domestic chip ecosystem. Earlier this year, Reuters reported that DeepSeek had broken from standard industry practice by refusing to share V4 with American chipmakers like Nvidia and AMD for pre-release optimization instead granting exclusive early access to Huawei and Cambricon.
What Is DeepSeek V4?
V4 is expected to be DeepSeek's most powerful model to date. Based on available technical details, it uses a Mixture-of-Experts architecture with approximately one trillion total parameters, though only around 37 billion activate per token keeping inference costs low while enabling frontier-level performance.
The model is reportedly native multimodal, meaning it can process and generate text, images, and video within a single system rather than bolting on separate vision modules. It also features a one million token context window, putting it in the same class as Google's Gemini models. Internal benchmarks suggest V4 could rival or exceed top-performing models from OpenAI and Anthropic on coding and long-context tasks.
The model is expected to launch within weeks, with reports indicating that a lighter variant, "V4 Lite," briefly appeared on DeepSeek's website in early March.
The Geopolitical Dimension
The decision to build V4 around Huawei chips carries enormous strategic weight. U.S. export controls have severely restricted China's access to advanced Nvidia GPUs, with Nvidia even halting China-bound H200 production in early March 2026. Washington has been tightening restrictions to slow China's AI progress.
But DeepSeek's V4 suggests those controls may be producing the opposite of their intended effect. Rather than crippling Chinese AI, the restrictions appear to be accelerating the development of a fully independent domestic AI stack from chips to models to cloud infrastructure.
DeepSeek's previous models, V3 and R1, already demonstrated that competitive AI could be built at a fraction of U.S. costs. The release of those models triggered a global tech stock selloff last year, wiping hundreds of billions off Nvidia's market capitalization and forcing investors to question whether Western AI companies really needed to spend as much as they were.
V4, running entirely on Chinese silicon, would take that argument even further.
What Comes Next
Neither Huawei nor DeepSeek responded to Reuters' requests for comment. But the direction is clear. China is building a parallel AI ecosystem that does not depend on American hardware, American software, or American permission.
For the global AI industry, DeepSeek V4 is not just another model launch. It is a proof of concept for technological self-sufficiency and a signal that the AI race now has two fully independent tracks.
