DeepSeek V4 model will run on Huawei chips as China accelerates AI independence

China’s AI race just took a turn that could reshape how the industry thinks about chips, control, and independence. DeepSeek’s upcoming V4 model is expected to run on Huawei-designed processors, marking a clear break from reliance on U.S. hardware.

The shift lands at a moment when access to advanced chips has become one of the biggest pressure points in global AI development. If V4 performs at a high level on domestic silicon, it challenges a long-held assumption: that frontier AI still depends on Nvidia.

The release, expected in the coming weeks, is being framed as a milestone in China’s push to build its own semiconductor stack. According to The Information, the model will run on Huawei’s latest chips, and demand is already building across the country’s biggest tech players.

“And in preparation for V4’s launch, Chinese tech giants, including Alibaba Group, ByteDance and Tencent Holdings, have placed bulk orders for Huawei’s upcoming chip totaling hundreds of thousands of units, according to five people with direct knowledge of the purchase,” The Information reported.

That level of demand sends a strong signal. Major players across China’s tech sector appear ready to back a homegrown stack at scale, not as an experiment but as a production-ready path forward.

China’s DeepSeek V4 ditches Nvidia, bets on Huawei chips in major AI shift

For months, DeepSeek has been working closely with Huawei and Cambricon to adapt its model to run on Chinese hardware. That process involved rewriting core pieces of the system and testing performance across new chip architectures. The effort reflects a broader shift behind the scenes, as software teams adjust to constraints that would have slowed progress just a few years ago.

The move stands in contrast to earlier claims that DeepSeek’s success relied on access to Nvidia chips. With V4, the company is signaling its ability to train, optimize, and deploy advanced models within China’s own ecosystem.

The implications stretch beyond one model. If domestic chips can support high-performance AI workloads, restrictions from Washington start to look less like a bottleneck and more like a forcing function that accelerates local innovation.

DeepSeek isn’t stopping at a single release. The company is said to be developing multiple V4 variants, each tuned for different use cases and built to run on Chinese hardware. That kind of vertical alignment, from chip to model, hints at a more self-contained AI stack taking shape.

Earlier this year, Reuters reported that DeepSeek chose not to provide U.S. chipmakers with early access to its upcoming model, breaking from common industry practice. Instead, the company worked closely with domestic partners, giving them a front-row seat during development.

The stakes are high. DeepSeek’s previous models, V3 and R1, rattled markets and raised fresh questions about the massive spending by U.S. AI firms on compute infrastructure. V4 arrives with even more attention, as investors and competitors watch closely for signs that the balance of power in AI hardware could shift.

Huawei and DeepSeek have not responded to requests for comment. The launch is expected in the coming weeks.

DeepSeek