China's AI chipmakers raise prices 20–50% amid HBM shortage
In brief
- Huawei's Ascend 950DT accelerator jumped 20–50% to over 250,000 yuan due to HBM constraints
- US export restrictions on advanced memory, effective December 2024, closed Chinese chipmakers' primary sourcing channels
- Cambricon, MetaX, and Iluvatar CoreX implemented matching 20–30% price increases
- Some Huawei models now ship no earlier than Q4 2026, forcing buyers to grey-market sources
Price hikes across the board
Huawei's Ascend 950DT accelerator card now costs more than 250,000 yuan, roughly $37,255—a 20–50% jump from quotes two months prior. The Ascend 950PR climbed approximately 30% to exceed 80,000 yuan since the start of 2026. The Ascend 910C rose about 22%, now priced above 110,000 yuan.
Huawei isn't alone. Cambricon, one of China's other prominent AI chip designers, has raised the price of its upcoming 690 chip by 20–30%. Smaller players like MetaX and Iluvatar CoreX implemented similar increases in the same range.
Delivery timelines have deteriorated sharply. Expected availability for some Huawei models has been pushed to Q4 2026 at the earliest, meaning companies placing orders today face nine-month waits.
Supply chain collapse and grey-market scramble
The root cause traces to US export restrictions on advanced HBM that took effect in December 2024, specifically targeting the memory modules that power cutting-edge AI accelerators. High-bandwidth memory stacks memory chips vertically and connects them with through-silicon vias, enabling the massive data throughput that large language models and AI workloads demand.
For Chinese chipmakers, this meant their primary channels for sourcing top-tier HBM from suppliers like SK Hynix and Samsung effectively closed. Major suppliers have allocated most of their output for 2026 already, leaving latecomers and restricted buyers competing for remaining supply at auction-like prices.
The result has been a migration toward grey-market sources, where HBM is available but at significantly higher prices and with less reliability. Domestic AI labs and cloud providers are caught between long waits (Q4 2026 or later) and inflated grey-market costs—both eroding competitiveness and delaying new model deployments.
