DeepSeek, a Chinese AI startup, has announced a permanent 75% price reduction on its V4-Pro AI model. This significant price cut places the model at a quarter of its previous cost.
The V4-Pro model now costs substantially less, a move linked to the increasing availability of Huawei's Ascend 950 AI supernode systems. This change marks a notable shift for DeepSeek, as the V4 series is optimized to run on Huawei's Ascend AI accelerators, diverging from a primary reliance on Nvidia hardware. The reduction, particularly for input cache-hit pricing, which is critical for tasks involving repeated prompts and extended AI agent operations, has been even more drastic, sometimes reaching one-tenth of prior expenses.
The company has not explicitly stated whether the permanent price reduction is a direct consequence of increased Huawei Ascend 950 chip supply. While Huawei's AI chip sales have seen a boost due to U.S. export controls limiting Nvidia's access to advanced semiconductors in China, separate restrictions on chipmaking equipment exports have concurrently hindered Huawei's capacity to ramp up Ascend production.
Read More: Internet Speed Awards 2026: Users Report Real Speeds vs Promises

This aggressive pricing adjustment is particularly striking for a frontier-grade AI system like the V4-Pro. Previously, DeepSeek cited "constraints in high-end compute capacity" as the reason for the Pro version's higher cost, up to 12 times more than the less capable Flash variant. The company indicated that V4-Pro pricing was anticipated to decrease significantly with the large-scale launch of Huawei Ascend 950 supernodes expected in the latter half of the year.
The declaration of a permanent price cut, however, raises questions about its long-term sustainability, especially if chip supply faces renewed limitations. DeepSeek has not provided details regarding profit margins, future pricing strategies, or concrete confirmations of Huawei's chip supply stability. No new benchmark comparisons have been released alongside this pricing announcement. The notion of "permanence" in AI pricing remains a rare commodity.
Read More: AWS MCP Server for AI Agents Now Available in US East 1 and EU Central 1