DeepSeek makes 75% price cut on flagship AI model permanent

DeepSeek turned a steep discount into standing policy, locking in a 75% price cut on its flagship V4-Pro model and leaving it at a quarter of its original price. The move is more than a temporary promotion: it signals that the Chinese artificial intelligence startup is betting on scale, developer loyalty and wider adoption rather than premium margins.

The decision lands at a moment when AI competition is increasingly being fought on price as much as performance. For smaller firms, startups and individual developers, a permanent cut on a frontier model can lower the cost of building products on top of advanced AI. For rivals, it raises the pressure to match lower prices without sacrificing the expensive computing power needed to train and serve top-tier systems.

That makes DeepSeek’s move strategically important well beyond a single product. A permanent reset in pricing can change user expectations, especially if customers begin to assume that the most capable models should rapidly become cheaper. It also puts a sharper focus on the economics behind AI platforms: who can afford to keep models available at scale, who can subsidize usage long enough to win market share, and how long premium pricing can survive in a market where adoption itself is becoming the main prize.

DeepSeek did not say whether the permanent price cut was linked to any increase in supply of Huawei’s Ascend 950 chips, which it used to maximize V4 performance. That chip question matters because hardware access remains a bottleneck in China’s AI industry, where model launches increasingly ripple into the domestic supply chain.

An earlier report in late April showed how quickly that effect can spread. Demand for Huawei’s Ascend 950 chips surged after the release of DeepSeek’s V4 model, and major Chinese internet companies including ByteDance, Tencent and Alibaba were said to be reaching out to secure orders. The chip scramble showed that one model launch could already move buying behavior across China’s biggest platforms.

Against that backdrop, DeepSeek’s permanent cut looks like a calculated wager on market share over margin. It may make advanced AI more accessible, but it also sharpens the central question now facing the industry: when frontier models are priced at a fraction of their original level, who can still afford to train them, serve them and compete at the top?