China-based artificial intelligence startup DeepSeek announced that it has made a permanent discount on the usage fees of its flagship model called V4-Pro. According to the company’s announcement, prices have fallen by approximately 75 percent compared to previous levels. Thus, the operational costs that developers face when creating artificial intelligence applications and services are expected to decrease significantly. Recently, the biggest problems in the artificial intelligence industry have been high infrastructure costs and lack of access to advanced artificial intelligence processors. For this reason, DeepSeek’s such a drastic price change in a short time has attracted attention throughout the industry.
According to the new tariff shared by the company, the cost of using the V4-Pro model varies between 0.025 and 6 yuan per million tokens, depending on the workload. Previously, pricing for the same model was between 0.1 and 24 yuan per million tokens. This difference means a serious cost advantage, especially for companies developing artificial intelligence services that require intense processing power. Despite this, DeepSeek did not directly explain the technical or business reason behind the price drop. Still, evaluations in the industry have turned their eyes to Huawei’s artificial intelligence-focused Ascend processors.
In its previous statements, DeepSeek admitted that limitations in accessing high-level processing capacity increased the cost of the V4-Pro model. It was stated that compared to the company’s lower-cost Flash model, the usage fee of the Pro version increased up to 12 times in some scenarios. One of the main reasons for this was that advanced AI hardware had limited availability in the Chinese market.
Huawei Ascend processors are becoming more critical for Chinese companies
NVIDIA’s inability to sell its most advanced artificial intelligence accelerators to China due to US export restrictions has made Huawei’s Ascend series processors more important for local companies. In particular, the recent prominence of the Ascend 950 platform brings with it comments that the artificial intelligence infrastructure in China has begun to strengthen through alternative solutions. Although there is no official confirmation regarding DeepSeek’s price reduction, it is thought that the bottlenecks on the infrastructure side may have been partially alleviated.
On the other hand, this decrease in the cost of artificial intelligence models may not be limited only to the Chinese market. In recent months, it seems that price and performance competition has accelerated between companies such as OpenAI, Google, Anthropic and Mistral. Reducing transaction costs, especially in large language models for corporate customers, enables developers to test more models and provide services with lower budgets. DeepSeek’s move may similarly put price pressure on other premium AI service providers.
In addition, it is stated that the problems on the hardware side have not been completely resolved. Huawei still faces several challenges in production capacity due to restrictions on advanced chip manufacturing equipment. Despite this, the ability of China-based companies to establish lower-cost artificial intelligence infrastructures with local solutions may lead to a change in the balance in the global market. DeepSeek’s latest price update is considered one of the early signs of this transformation.
In order not to miss the technology agenda, 📰 add it to Google News, 💬 join our WhatsApp channel, ▶ subscribe to YouTube, 📷 follow us on Instagram and 𝕏 X.