News
Huawei Raises Prices on Flagship AI Chip by 60 Percent in Three Months

Huawei has notified customers of a roughly 60 percent price increase on its most advanced AI accelerator, the Ascend 950DT, over three months. The cause is a shortage of HBM memory triggered by US export controls.
Contents
Huawei has told customers about a sharp price increase for its most advanced artificial intelligence accelerator. The suggested price of the Ascend 950DT card has risen by roughly 60 percent over the past three months, to more than 250,000 yuan per unit, or about $37,000. The cause is a shortage of high-bandwidth memory (HBM), which is hitting the entire Chinese AI chip industry.
The news was first reported by Reuters, citing sources familiar with the matter and pricing documents sent to customers. The Ascend 950DT is currently the most advanced AI accelerator in Huawei's portfolio, with its full market launch planned for the fourth quarter of 2026. The company is pricing it close to Nvidia's B200 chip, which in itself shows how hard Chinese manufacturers are trying to keep pace with American competitors despite the sanctions.
Where the Shortage Comes From
The root of the problem lies in HBM, high-speed memory stacked directly next to the processor core that is critical to AI accelerator performance. Washington tightened export controls on advanced HBM chips to China in December 2024, cutting Chinese manufacturers off from the main legal supply channels.
As a result, companies such as Huawei, Cambricon and MetaX are increasingly turning to gray-market supplies, where HBM prices run several times higher than what buyers outside China pay. Since memory accounts for a large share of the cost of producing an AI accelerator card, the rise in its price feeds directly into the price of the finished product reaching customers.
Demand Outpaces Supply
The price hikes are not just a reaction to costs, but also a reflection of a genuine demand boom. China's AI chip market, estimated at $50 billion, is growing because US export restrictions on Nvidia have pushed domestic tech companies toward local suppliers. Chinese data centers need ever more computing power, and manufacturers can't keep up with production.
The pressure is strong enough that suppliers have started shifting resources between customers. Iluvatar CoreX doubled its annual GPU deliveries to ByteDance to 100,000 units, redirecting chips originally earmarked for other contracts to that customer. This shows that even the biggest players in China's tech market are now competing for access to finished chips.
Not Just Huawei
Cambricon, China's second-largest AI chipmaker, raised prices on its upcoming 690 model by 20-30 percent compared with levels signaled to customers just two months earlier, though the company has not disclosed specific figures. MetaX and Iluvatar CoreX introduced similar increases, though without publicly known pricing details.
This is a sign that the HBM problem is systemic, not isolated. China's entire AI chip design ecosystem, which was meant to be an answer to sanctions on Nvidia, has itself fallen into a trap of dependency on components controlled by Western supply chains.
What This Means for the Global Market
For companies and investors tracking the global race for AI infrastructure, Huawei's chip price increases are further evidence that the HBM memory shortage is hitting not just the West, where Samsung and SK Hynix have reported record-low inventories for months, but also China, despite its push for technological self-sufficiency.
For Polish and European companies buying hardware or computing services based on Chinese accelerators, this means higher deployment costs and longer delivery times. Rising prices for AI components on both sides of the US-China rivalry suggest that price pressure across the entire global AI infrastructure market will persist for a while longer, regardless of which supplier a customer chooses.

