Summary of Key Points
This news article focuses on three significant developments at AI company DeepSeek: the stealthy launch of its new flagship model V4-Pro (still at an extremely low price), announcements of upcoming price increases, and persistent rumors about potential listings and financings. Behind these moves lies DeepSeek's commitment to two crucial battles that will determine its future success. The first battle is for control over the pricing of AI models globally—by offering competitive prices and strong capabilities, DeepSeek aims to make itself indispensable to developers worldwide. The second battle is for a share of the profits from the open-source AI ecosystem, ensuring that the value created by this ecosystem ultimately benefits DeepSeek. The outcome of these battles will determine whether DeepSeek becomes a “small but powerful model company” or a “global platform with significant influence,” and whether its valuation of 50 billion yuan is just the starting point or the ultimate goal.
Detailed Analysis
1. The New Model: Inexpensive and Efficient, but Still Needs Official Recognition
DeepSeek’s new V4-Pro model has been released at an exceptionally low price: $0.435 per million text tokens input, compared to much higher rates charged by American companies like OpenAI. Developers have already begun testing it; for example, one person used the model to create a complete 3D shooting game with 8 levels and 4 weapons for a total cost of only $0.19. Tools have also indicated that the model’s performance has improved by 15.8% compared to previous versions. However, these are all unofficial tests. Previously, the US National Institute of Standards and Technology (NIST) rated DeepSeek’s older models as the strongest in China, but they were still about 8 months behind the most advanced models like GPT-4. It’s important to note that a low token cost does not necessarily mean lower overall costs; more tokens may be required for complex tasks, potentially making the final cost higher than with more expensive models. DeepSeek’s current goal is not to top rankings but to demonstrate its ability to continuously reduce the costs associated with using its models, providing Chinese developers with a more cost-effective alternative compared to American models.
2. Price Increase Announcements: Not for Profit, but to Test User Loyalty
What’s most intriguing is that DeepSeek has announced a price increase shortly after the new model’s release. This is not just a random decision; it’s a way to assess user loyalty. People have been using DeepSeek’s models because of their low cost, but perhaps also because of their quality. True pricing power lies in the ability to raise prices without losing users. If users continue to use DeepSeek’s models even after the price increase, it will prove that its models are truly valuable. Investors are closely watching this development; despite a valuation of 50 billion yuan, DeepSeek’s annual revenue is only $400–500 million (140 times lower than its valuation). Capital is interested in whether DeepSeek can turn its technical influence into stable profits in the future. A price increase serves as a test of this; if users do not reduce their usage after the price rise, it will strengthen DeepSeek’s pricing power.
3. Listing Pressure: The Transition from a “Technological Leader” to a “Profit-Generating Company”
There have been persistent rumors about DeepSeek’s potential listing and financings (including a valuation of 50 billion yuan that was later postponed). Domestic competitors such as Zhipu and MiniMax are already active in the public market, while companies like ByteDance and Alibaba have advantages in cloud services and traffic. Although DeepSeek is the most well-known, it must prove that its global popularity can translate into actual revenue. Its strategy is to build an open ecosystem by collaborating with third-party cloud providers to distribute its models and promote the use of domestic chips. However, this raises questions about profit distribution: if third parties profit from selling DeepSeek’s models, how much will DeepSeek receive? A company preparing for a listing must clearly explain where its revenue comes from, how much computing power it invests, and how much profit it expects to make—otherwise, investors may not be convinced.
4. Two Crucial Battles That Will Determine Its Valuation
DeepSeek is engaged in two battles that will significantly impact its future success:
- Battle for Pricing Power: Can DeepSeek continue to offer powerful models at competitive prices, ensuring that they are considered by developers worldwide? This will determine its place among the leading AI companies.
- Battle for Profit Sharing: Will users continue to use DeepSeek’s models even after price increases? Will companies incorporate them into their long-term budgets? The success of this battle will also affect how much profit DeepSeek can generate from its ecosystem.
The future of DeepSeek depends on four key factors:
- The official evaluation results of the new model.
- Whether DeepSeek makes the new model more widely available to users.
- How the price increase is structured (with discounts for large customers and higher prices for small ones).
- The stability of usage after the price increase.
These factors will provide insights into whether DeepSeek’s valuation of 50 billion yuan is a realistic goal or just a starting point.
Conclusion
DeepSeek’s current efforts aim to demonstrate its capabilities with the new model and its potential value to users through price increases. If it succeeds in these areas, it can transform from a company that sells inexpensive models into a platform that influences global AI pricing, potentially achieving a valuation in the trillions of yuan. Otherwise, it may remain a company with strong technology but limited financial performance.
(End of article)