Chinese AI chip suppliers, including Huawei, Cambricon, MetaX and Iluvatar CoreX, have sharply raised prices on both current and next-generation AI processors.
According to Reuters, the increases run as high as 30% across current and upcoming chip generations.
Why memory is the bottleneck
The price rises are driven primarily by a global shortage of high-bandwidth memory, known as HBM, a specialised type of memory chip essential to training and running large AI models.
That shortage is straining Chinese efforts to build competitive domestic AI accelerators, even as demand for such chips continues to grow.
What this reveals about China's position
Two implications follow from this development.
First, it is difficult to shift away from HBM entirely, since China would likely be leading such a shift already if effective alternatives were easy to implement, given how directly the shortage is constraining its own chipmakers.
Second, the broader growth of China's domestic memory industry could be constrained if vendors such as YMTC and CXMT prioritise HBM production specifically, a plausible outcome given the country's intense focus on AI development.
Constraints and exposure
The price increases suggest that despite years of investment aimed at building an independent domestic chip supply chain, Chinese AI hardware makers remain exposed to the same global memory constraints affecting manufacturers everywhere else.
That dependence is likely to keep HBM availability, rather than raw chip design capability alone, as one of the clearest constraints on how quickly China's AI chip industry can scale.
