Huawei Targets 12 Billion Dollars in AI Chip Revenue as China Builds Its Own Stack
Huawei expects its AI chip business to generate 12 billion dollars in 2026, a 60 percent jump driven by mass production of the Ascend 950PR and growing domestic demand after Nvidia's restricted access to China.

Key Takeaways
- Huawei expects AI chip revenue to hit 12 billion dollars in 2026, up 60 percent from last year
- The Ascend 950PR processor entered mass production in March with 750,000 units targeted this year
- ByteDance committed over 5.6 billion dollars in Ascend chip purchases, funding much of the production ramp
- DeepSeek V4 was optimized for Huawei's chips, reducing China's reliance on Nvidia hardware
Huawei is projecting its artificial intelligence chip business will generate roughly 12 billion dollars in revenue this year, marking a jump of more than 60 percent from the 7.5 billion dollars it earned in 2025. The dramatic surge is being driven by mass production of the company's latest Ascend 950PR processor and a wave of domestic orders rushing to fill the gap left by Nvidia's increasingly restricted access to the Chinese market.
Ascend 950PR Enters Mass Production
The Ascend 950PR is Huawei's most powerful AI chip to date and entered mass production in March 2026. The company is targeting production of roughly 750,000 units this year, with full-scale shipments expected to ramp up significantly in the second half. ByteDance, the parent company of TikTok, has committed more than 5.6 billion dollars in Ascend chip purchases alone, a massive increase from nearly zero the previous year. That single deal effectively co-finances a large portion of Huawei's entire production ramp. An upgraded version of the chip called the Ascend 950DT is planned for release in the fourth quarter of this year, signaling that Huawei intends to keep iterating on its hardware at a rapid pace.
DeepSeek V4 Fuels Demand for Domestic Chips
A major catalyst behind the revenue surge is DeepSeek V4, a large language model released in April 2026 that was specifically optimized to run on Huawei's Ascend hardware. A large language model, often called an LLM, is a type of AI system trained on massive amounts of text data so it can understand and generate human language. By tailoring DeepSeek V4 to work efficiently on Ascend chips, Chinese AI developers no longer need to depend on Nvidia's graphics processing units for training and running their most advanced models. Industry analysts now project that Huawei could capture around 60 percent of China's AI chip market by the end of the year, according to the Financial Times.
The rapid growth underscores how US export restrictions have accelerated rather than slowed China's push to build a fully self-sufficient AI supply chain. With domestic software now optimized for domestic hardware and major tech companies pouring billions into local alternatives, Huawei is firmly positioned as the anchor of an emerging ecosystem that no longer depends on American chips.
Stay Informed
Weekly AI marketing insights
Join 5,000+ marketers. Unsubscribe anytime.
