Summary of Key Points
DeepSeek has announced that it will significantly increase the prices of its API services in the near future, causing widespread concern among developers. This move reflects the current shortage of computing power in the industry and the soaring costs associated with using these services. DeepSeek’s high volume of requests has made it the most popular platform globally, but its low-price strategy has led to long-term losses. The price increase marks not only a turning point for DeepSeek’s commercialization efforts but also the gradual end of a phase in which domestic large-model companies were willing to sacrifice profits to gain market share through low prices. Now, the industry is moving towards more rational pricing. At the same time, this change tests DeepSeek’s core competitiveness: will it be able to maintain its position with its cost-effective “floor price” or through the strength of its models?
1. Why the Sudden Price Increase? The Main Reason is a Lack of Computing Power
You can think of an AI model as a huge factory, where Tokens represent the products being produced. DeepSeek’s “factory” is currently under extreme pressure:
- Exponential Increase in Requests: Data from the OpenRouter platform shows that DeepSeek-V4-Flash generated 7.22 trillion Tokens in one week, exceeding the total usage of all other models on the platform combined. On the OpenCode platform, 5 trillion out of 8 trillion Tokens used daily came from free trials.
- Increasing Complexity of Tasks: Previously, users only engaged in simple question-and-answer interactions (about 2000 Tokens per request), but now there’s a surge in Agent-based applications that require millions of Tokens per request. This is like the factory shifting from producing small parts to manufacturing cars, resulting in a hundreds-of-times increase in resource consumption.
- Cost Inversion: DeepSeek’s previous pricing (1 yuan per million Tokens for input and 2 yuan per million Tokens for output) was below its actual costs, leading to negative profit margins. For example, its prices were only 1/105 of those of Claude Fable 5, meaning it was losing money on every transaction, relying heavily on funding to sustain operations.
In short, too many people are using the service too intensively, and the costs have become unaffordable.
2. How Much of a Price Increase is “Appropriate”? Different Rates Have Different Consequences
The market is particularly concerned about the extent of the price increase, as different ranges will have different impacts:
- 1-2x Increase (Mild Range): Users can generally accept this and are unlikely to switch to competitors.
- 3x Increase (Profitable Range): With input prices rising to 4 yuan and output prices to 8 yuan, profit margins could return to 30%-40%, allowing DeepSeek to eliminate low-value users (such as those using automated scripts) and improve system stability.
- 4x Increase (Critical Point): Although the prices are still lower than those of competitors (with domestic flagship models typically ranging from 10 to 30 yuan), the price difference would shrink from 10 times to 2 times. If the increase isn’t enough to cover migration costs, some high-frequency users might switch to other models (like GPT Plus).
Developer feedback supports this: Some say they can accept a 100% increase, while others are willing to tolerate a 200% increase given the previous low prices.
3. It’s Not Just DeepSeek; the Entire Industry is Seeing Price Increases
DeepSeek is not alone in this trend. This year, many domestic large-model companies have started raising prices:
- Zhipu AI: Raised prices three times this year, with prices increasing by 83% in the first quarter of 2026, while usage increased by 400%.
- Moon’s Dark Side: Kimi K3’s input prices have risen by 3 times, and output prices by 4 times, due to the higher resource consumption of its more advanced models.
- Cloud Providers: Alibaba Cloud and Baidu Smart Cloud have increased their computing power services by 5%-34%, and Tencent Cloud has raised prices twice, both due to global demand for computing power and rising chip costs.
The industry consensus is that long-term low-price competition is detrimental to development, and capacity upgrades inevitably lead to higher costs. DeepSeek’s price increase signals a shift in the industry’s approach: by raising prices, it frees other companies from the pressure of keeping them low, allowing everyone to breathe easier.
4. Can DeepSeek Still Compete After the Price Increase? It Depends on Its Core Competitiveness
In the past, DeepSeek relied on its excellent cost-effectiveness to attract users. Now, with reduced pricing power, its ability to retain users depends on two factors:
- Quality of Models: If its models are truly superior to competitors, users will continue to use it even after the price increase.
- Self-developed Chips: DeepSeek has started a project to develop its own inference chips, aiming to reduce computing costs and reduce reliance on external chips. This could give it more bargaining power in chip purchases and help manage long-term costs.
DeepSeek has also just raised 51 billion yuan and plans to raise another 50 billion yuan, indicating sufficient funding. Investors are concerned about the company’s ability to generate profits; the price increase is aimed at improving its financial sustainability.
5. What’s the Impact on Users?
The impact of the price increase varies depending on the user group:
- Small and Medium-sized Developers/Individuals: They are most affected, as they often face higher costs due to high Token usage. Some may switch to cheaper alternatives like GPT Plus.
- Enterprise Customers: The impact is relatively minor, as many companies use private deployments and pay for their own computing resources.
- Automated Script Users: A 300% price increase will likely eliminate those using automated scripts to obtain Tokens for free, allowing only genuine users to access stable services.
In summary, this price increase marks DeepSeek’s transition from a strategy of burning money to gain scale to one focused on making profits. It also signals a shift in the industry from aggressive pricing wars to more rational competition. Whether DeepSeek can retain its user base will depend on the strength of its models to justify the new prices.
(The entire text is written in plain language to make it easy for non-financial professionals to understand.)