DeepSeek Slashes AI Prices Permanently as China’s Chip Push Intensifies Pressure on Global Rivals

Chinese artificial intelligence startup DeepSeek has permanently reduced prices for its flagship V4-Pro model by 75%, signaling an aggressive escalation in China's AI pricing war as domestic firms race to challenge U.S. dominance while adapting to Washington's semiconductor restrictions.

The move sharply lowers the cost of accessing one of China's most advanced large language models and could intensify pressure on rivals across Asia and the United States already grappling with falling margins and soaring infrastructure costs tied to generative AI.

DeepSeek said on Saturday that V4-Pro application programming interface prices would now range between 0.025 yuan and 6 yuan per million tokens, down from 0.1 yuan to 24 yuan previously. Tokens are the units of text AI systems process when generating or analyzing content.

At the upper end, the reduction cuts pricing from roughly $3.30 per million tokens to less than $0.85, bringing DeepSeek's costs far below many Western frontier AI models and reinforcing China's strategy of competing aggressively on affordability and scale.

The announcement is also seen as an indication that China's domestic AI ecosystem may be overcoming at least part of the compute bottleneck that had constrained advanced model deployment following sweeping U.S. export controls on high-end chips.

While DeepSeek did not directly confirm the reason for the permanent cut, the decision is likely to fuel speculation that supplies of Huawei's Ascend 950 AI processors are improving ahead of broader deployment later this year. When the company launched V4 last month, it acknowledged that the more advanced Pro version was significantly more expensive because of "constraints in high-end compute capacity." At the time, DeepSeek said V4-Pro could cost as much as 12 times more than its lower-tier Flash model due to limited access to advanced hardware.

The company had also indicated that prices would likely fall once Huawei's Ascend 950 supernodes entered large-scale production in the second half of the year. The timing of the announcement, therefore, points to broader developments inside China's semiconductor supply chain, where domestic technology firms are accelerating efforts to reduce reliance on American chipmakers such as Nvidia.

Washington's tightening export restrictions have barred Nvidia from selling its most advanced AI processors to Chinese customers, creating a vacuum that domestic firms have rushed to fill. Huawei has emerged as the central beneficiary of those restrictions, with its Ascend AI chips becoming a strategic alternative for Chinese cloud providers, AI startups, and state-backed technology projects.

Still, China's chip ambitions continue to face serious constraints. Separate U.S.-led restrictions on semiconductor manufacturing equipment have complicated Huawei's ability to scale production of cutting-edge processors, particularly at the most advanced nodes needed to compete directly with Nvidia's latest Blackwell and Hopper systems.

That tension has created a two-track AI market inside China: rapid software innovation paired with persistent hardware limitations.

DeepSeek's pricing move suggests the balance may be beginning to shift. By driving inference costs sharply lower, Chinese AI companies are attempting to accelerate adoption across enterprises, developers, and government agencies at a time when the economics of generative AI remain under scrutiny globally. The cost of running advanced models has become one of the industry's most important competitive battlegrounds, especially as businesses increasingly prioritize practical deployment over experimental AI spending.

However, there is projection that the cuts could also deepen competitive pressure on China's own AI sector. Domestic rivals, including Alibaba Group, Baidu, and Tencent, have all reduced AI model pricing over the past year as competition intensifies for developers and enterprise customers. Several firms are increasingly treating AI models as low-margin infrastructure services designed to drive cloud computing, data and software ecosystem growth.

That mirrors a broader trend already visible in the United States, where frontier AI companies are spending tens of billions of dollars on compute infrastructure while simultaneously facing mounting investor pressure to prove sustainable monetization.

DeepSeek's move is significant because it comes as global AI companies grapple with escalating infrastructure expenses tied to increasingly powerful models. Industry leaders including OpenAI, Anthropic and xAI are collectively committing hundreds of billions of dollars to data centers, chips and energy infrastructure to support next-generation AI systems.

China's approach appears focused on compressing costs and accelerating scale rather than maximizing near-term profitability. The strategy could prove influential across emerging markets where businesses and governments want advanced AI capabilities but remain sensitive to infrastructure and operating costs. Lower-priced Chinese AI services may become especially attractive across Asia, Africa, and the Middle East, regions where Chinese cloud and telecom infrastructure already maintain a strong presence.

The pricing cuts also reinforce how the AI race is becoming inseparable from the geopolitical contest over semiconductors. As the United States attempts to restrict China's access to advanced chips, Beijing is accelerating efforts to build a vertically integrated AI ecosystem spanning models, semiconductors, cloud infrastructure and industrial deployment.

DeepSeek's aggressive reductions suggest Chinese firms are increasingly confident that domestic hardware alternatives, even if less advanced than Nvidia's top systems, may now be sufficient to support commercially viable large-scale AI deployment.