A new storm may be brewing in the AI industry, sparked by rumors surrounding a Chinese company called DeepSeek. It is reportedly preparing to launch a new AI model, ‘V4’, at a price so low it could reshape the entire market.
Just how cheap is it? Analyst estimates place the cost at around $0.25 per million output tokens. To put that in perspective, it’s about 60 times cheaper than Anthropic's Claude Sonnet 4.6 ($15) and 40 times cheaper than OpenAI's GPT-5 ($10). For a company spending $15,000 a month on an AI API, this could slash costs to just $250. It’s no wonder one analyst called it a “financial missile aimed at CFOs.”
This potential disruption didn’t happen in a vacuum. It’s the result of several converging trends. First, there's a clear and growing global demand for affordable AI solutions. Recent data showed that for the first time, the usage of Chinese AI models on platforms like OpenRouter surpassed that of US models, proving that users are actively seeking out cost-effective alternatives for large-scale tasks.
Second, this is a story of technological self-reliance. Ongoing US restrictions on exporting high-end AI chips, like NVIDIA's H200, have pushed China to accelerate the development of its own hardware and software ecosystem. Companies are increasingly building on domestic chips from Huawei (Ascend) and Cambricon, creating a fully independent stack that allows them to control costs from the ground up.
Finally, the fluctuating nature of US policy has inadvertently strengthened China's resolve. The back-and-forth between partial easing of restrictions and threats of new quotas has created an unpredictable environment, reinforcing the strategic importance of a domestic supply chain. This paradoxically gives Chinese companies the foundation to compete aggressively on price.
If DeepSeek V4 launches at this price point, it will force a market-wide response. While major US tech companies have diversified businesses that can absorb some impact, their high-margin AI API services will face intense pressure to lower prices, kicking off what could be the second round of the great AI price war.
- Token: A basic unit of data, like a word or part of a word, that AI models use to process and generate text.
- Multimodal Model: An AI model that can understand and process information from multiple types of data, such as text, images, and video, all at once.
- Inference: The process of using a trained AI model to make predictions or generate outputs based on new, unseen data.