- DeepSeek is revising its pricing strategy.
- The Chinese startup announced a “significant price increase” for its API.
- This move is being linked to a recent funding round and a possible IPO later this year.
Chinese AI startup DeepSeek announced an upcoming “noticeable” increase in the cost of its models’ API. The company did not disclose the new rates, but urged customers to adjust their service usage in advance.
The company’s statement says:
“The specific pricing plan will be subject to official notice.”
The price hike comes amid a major funding round and DeepSeek’s preparations for a potential listing as early as this year.
Recall that after the funding round closed in June 2026, the company’s founder, Liang Wenfeng, became the world’s richest entrepreneur among AI model creators.
DeepSeek Moves Away From an Ultra-Cheap AI Strategy
Right now, the DeepSeek V4 Flash model costs just $0.14 per million input tokens and $0.28 per million output tokens.
For comparison:
- Moonshot Kimi K3 — $3 and $15 per million tokens, respectively
- Anthropic Fable 5 — $10 and $50
It was precisely this price gap that made DeepSeek one of the key drivers of the price war in the generative AI market. The rates were so low that they even spawned the term “DeepSeek death zone” — a situation in which more expensive or less capable models lose competitiveness, Bloomberg writes.
According to the outlet, the new approach could become a turning point for China’s artificial intelligence market. While the company previously focused on making models as cheap and open as possible, investors now expect a shift toward a more profitable business model.
Another signal was the change in pricing policy just three weeks after introducing time-of-day tariffs. If the July model was aimed at balancing the load on infrastructure, the current decision points to a rethink of the service’s positioning.
According to Bloomberg, the company also plans to build a large data center in Inner Mongolia as part of a project to create 1 GW of infrastructure for training and running artificial intelligence models. DeepSeek may lease part of its compute resources from third-party operators.
The Market Reacted to the Decision With Mixed Feelings
Higher tariffs could directly affect developers and companies that use the DeepSeek API, although users will still have alternatives. Since the weights for the V4 Flash model are available under an open license, other providers can deploy it on their own servers and offer access at their own prices.
At the same time, some industry participants questioned whether the move made sense.
AI developer Michael Guo wrote:
“DeepSeek choosing to raise prices at this time—isn’t this just asking for trouble? American models with equivalent capabilities, Muse Spark and GPT Luna, are at DeepSeek’s price point, or even cheaper!”
Robert Li, Bloomberg Intelligence analyst, meanwhile, noted that Chinese users’ interest in AI services remains high, and that DeepSeek, together with ByteDance, continue to dominate the domestic chatbot market, increasing pressure on smaller competitors.
As a reminder, in May 2026, DeepSeek cut the cost of using its flagship V4-Pro model by 75%, one of the most high-profile moves in the price competition in the generative AI market.
Сообщение DeepSeek Announced ‘Significant’ API Price Increase Amid Potential IPO появились сначала на INCRYPTED.