Huawei and Cambricon sharply increased prices for their AI accelerator cards amid a global HBM shortage, underscoring supply‑chain pressures in the AI chip market.
China’s leading AI chipmakers have announced steep price hikes for their accelerator cards, as a worldwide shortage of high‑bandwidth memory (HBM) tightens supply chains and drives up costs.
Price spikes hit flagship products
Huawei’s Ascend series and Cambricon’s MLU‑3000 saw price increases of up to 30% compared with last quarter, according to statements from the companies. The hikes affect both enterprise‑grade cards and developer kits, raising concerns for data‑center operators and AI startups that rely on Chinese hardware.
Why HBM scarcity matters
HBM provides the ultra‑fast data throughput required for large‑scale neural‑network training. Global demand for the memory has surged after major cloud providers announced new AI services, while production capacity at key fabs remains constrained.
- Limited fab slots for HBM stacking at major foundries
- Long lead times for raw silicon wafers
- Competing demand from graphics cards and high‑performance computing
Impact on the Chinese AI ecosystem
The price rise could slow the rollout of AI projects in China’s tech sector, where domestic chips are promoted as alternatives to U.S. suppliers. Analysts note that higher costs may push some firms to source foreign accelerators, despite ongoing trade restrictions.
Both Huawei and Cambricon said they are working with memory manufacturers to secure additional HBM capacity, but warned that the shortage is likely to persist through the end of the year.
For full details, see the Reuters coverage of China’s AI chipmakers price hikes.