Chinese AI chip manufacturers including Huawei and Cambricon have sharply increased prices on AI processors as a critical shortage of high-bandwidth memory drives up costs. This development underscores the challenges Beijing faces in developing home-grown alternatives to Nvidia amid ongoing US export restrictions.
- Huawei’s Ascend 950DT price increases over 20% amid HBM scarcity
- US export controls limit advanced memory access, inflating costs
- Chinese chipmakers rely on costly grey markets for memory supplies
What happened
Major Chinese AI chipmakers such as Huawei, Cambricon, MetaX, and Iluvatar CoreX have raised prices on their current and upcoming AI processors by 20% to 50% over the past two months. Huawei’s flagship Ascend 950DT accelerator card now exceeds 250,000 yuan ($37,255), making it significantly more expensive than earlier price points. The price hikes are echoed in older models as well, with increases of around 30%.
The root cause is a global shortage of high-bandwidth memory (HBM), a vital component for AI processing that allows rapid data access via vertically stacked memory chips. HBM production is dominated by South Korean and US companies, and US export curbs imposed in December 2024 have restricted Chinese chipmakers’ access to advanced HBM products. This has forced them to seek scarce and expensive supplies on grey markets, driving up component and finished product costs.
Why it matters
The price increases reveal critical supply chain pressures limiting China’s ability to scale competitive AI hardware domestically. Despite government efforts to develop home-grown alternatives to Nvidia’s AI processors, the high cost and limited availability of crucial HBM components present a bottleneck, raising the cost of AI infrastructure development substantially.
This situation complicates Beijing’s strategy to reduce dependence on US technology by stimulating its $50 billion AI chip market with local innovation. The shortage has also led chip companies to prioritize shipments and reshape production plans, affecting large customers including tech giants like ByteDance, which rely heavily on these domestic suppliers for AI computing power.
What to watch next
Key developments will include whether Chinese chipmakers can accelerate the rollout of proprietary HBM technologies such as Huawei’s HiBL 1.0 and HiZQ 2.0 to reduce dependency on restricted foreign sources. Monitoring pricing trends and supply chain adjustments in the coming quarters will provide insight into how effectively domestic substitutes gain traction.
Additionally, observing government policy responses and export control adjustments will be critical for forecasting the future pace and cost competitiveness of China’s AI chip industry. The evolving dynamics between international sanctions, grey-market supply routes, and internal innovation efforts will shape the region’s position in the global AI hardware landscape.