GuShiio.com Article source: Chop Pepper Spicy
Image source: Generated by AI
DeepSeek,It has taken the lead in opening up a closed business loop.
Source: DeepSeek
On March 1, 2025, DeepSeek released a technical article stating that from 24:00 on February 27 to 24:00 on February 28, the GPU rental cost is calculated at US$2/hour, and the average daily cost of the platform is US$87,072 (approximately RMB 630,000); based on R1 pricing for all input/output tokens, the platform’s single-day revenue is US$562027 (approximately RMB 4.09 million), the net profit is US$474955 (approximately RMB 3.46 million), and the cost-profit margin is as high as 545%.
However, it should be noted that the financial data displayed by DeepSeek is only the result of highly simplified and idealized processing, which deviates greatly from the complex situation in actual operations. DeepSeek admitted thatBecause V3 ‘s pricing is lower, only charges for some services, and discounts during off-peak hours, the company’s actual revenue is not so much.
Source: WeChat
Perhaps because of the huge commercial imagination contained in the DeepSeek model, recently,Many technology companies have spared no effort to promote their own products connected to DeepSeek。For example, on February 16, Baidu search officials announced full access to the latest in-depth search functions of DeepSeek and Wenxin Model. On February 27, Tencent’s new product “Yuanbao” entered the life service section of WeChat.
In fact,The popularity of DeepSeek cannot hide the embarrassment of limited commercialization capabilities of AI models。Since ChatGPT, which was launched in November 2022, set off a trend for large models, AI large models have not found an ideal commercial implementation scenario, and most AI large model companies have been mired in losses.
Because C-side products are difficult to charge and need to continuously burn money to promote them, the commercialization of C-side AI large-scale model products faces great challenges.
One possible commercialization path is the commercialization of AI search.
For example, some netizens recently discovered that Xiaomi Super Xiaoai was answering “How is the weather today?” When the problem was raised, commercial advertisements were added at the end.
Photo source: Weibo
On February 25, 2025, the dark side of the moon made industrial and commercial changes to add “Drug Internet Information Services”, which is suspected to be preparing for Kimi to launch medical advertisements.
At the 2024 Q4 earnings call conference, Baidu said,”We expect our advertising revenue to gradually increase, thanks to our efforts to monetize the results of artificial intelligence transformation.”
ButBecause the results provided by AI search are highly unique, the richest advertisers may not be able to provide the best solution。On the contrary, due to the need to balance high marketing costs, the solutions provided by advertisers may lack value for money and even harm the interests of consumers. After asking people around him, Spicy found that many people are cautious about AI search and advertising, worried that AI search will lose fairness after adding advertising.
Another business path is for technology companies to act as “water sellers” to generate revenue by providing AI infrastructure to B-end customers。For example, on February 20, iFlytek launched two full-stack domestically produced AI infrastructures, the Spark DeepSeek Tower All-in-One Machine and the Spark DeepSeek All-in-One Machine, committed to providing downstream customers with efficient and reliable large-model deployment solutions.
However, not all technology companies have the technical heritage that can support the business story of “water sellers”. Referring to the experience of the PC and mobile Internet industries, if the AI model wants to open up a closed loop of business, it seems that it should have a deep insight into user demands and create new interactions and transaction scenarios, rather than minimally innovating along existing business models.
The reason why DeepSeek has swept the technology industry recently is of course because it adheres to the spirit of openness and is completely open source, but this is also inseparable from its more efficient reasoning and lower cost. Official data shows that the reasoning cost of DeepSeek-R1 is only 17% of that of the GPT-4 Turbo.
After DeepSeek is open source, technology companies no longer need to spend huge sums of money to build closed-source models. By directly connecting to DeepSeek-R1, they can obtain first-class AI large-scale model capabilities at a lower cost and naturally welcome them.
However, it should be noted thatAlthough DeepSeek’s reasoning costs are low, they are not completely free。With the increasing number of downstream users, many companies that have difficulty getting through the closed business loop have begun to fall into the “scale trap”.
Photo source: Liang Bin’s Weibo
On February 14, 2025, Liang Bin, Chairman and CEO of Beijing Bayou Technology Co., Ltd., issued a document on the social media platform stating that if a 2C service running the full-blood version of DeepSeek uses AMD MI300X to provide 100 billion tokens of output per day, then the total monthly cost is approximately 360 million yuan.
An AI practitioner told Spicy,Since cost is directly proportional to the amount of calculation, current AI products do not have obvious marginal effects like traditional Internet businesses.”But the better the product effect, the higher the subsequent cost.”
Given that most of the products currently connected to DeepSeek have not broadened revenue channels, most companies ‘AI-related businesses are actually mired in losses.
For example, the “Consensus Crusher” on the Weixin Official Accounts revealed thatIn order to support WeChat’s AI search, Tencent added 100,000 – 200,000 H20s, and the hardware value alone was as high as US$2 billion.
Perhaps it is because upstream costs are extremely high, AI search results are limited by the WeChat ecosystem, and commercialization capabilities are limited. To this day, WeChat has not opened AI search functions to all users.
In this context, Tencent is committed to promoting Tencent Yuanbao, which has access to the full-blooded version of DeepSeek and can be searched on the entire network.
Thanks to the large number of streaming on WeChat, Station B, Zhihu and other platforms, on February 22, Tencent Yuanbao’s single-day downloads surpassed Doubao, ranking second in the App Store’s free App download rankings in China.
Coincidentally, precisely because it is realized that AI model technology contains huge imagination, in 2024, 360 will also begin to increase its AI-related business, and successively launched 360 AI search, 360 Children’s Watch A9 AI Red Dress Edition, AI Assistant and other products.
Source: 360 Financial results for the first three quarters of 2024
Unfortunately,Due to the high R & D and operating costs of AI-related technologies, and the difficulty of “open source” downstream, 360 ‘s performance is deteriorating day by day.。The financial report shows that in the first three quarters of 2024, 360’s operating costs were 2.608 billion yuan, a year-on-year surge of 31.92%. In the same period, 360’s revenue was 5.609 billion yuan, a year-on-year decrease of 16.76%; the net loss was 579 million yuan.
In stark contrast to Internet companies that face users, it is difficult for them to open up commercial closed-loop operations in the AI model businessThe performance of “water sellers” such as chip manufacturers and cloud service vendors upstream of the AI industry is steadily improving.
Source: Nvidia’s third-quarter financial report for fiscal year 2025
Take Nvidia as an example. In the third quarter of fiscal year 2025, its data center business revenue reached US$30.8 billion, a year-on-year increase of 112%, a record high. Thanks to this, Nvidia’s net profit reached US$19.309 billion, a year-on-year increase of 109%; gross profit margin was 74.6%, a year-on-year increase of 0.6 percentage points.
The reason isMainly because Nvidia’s AI chips have unmatched performance and are extremely irreplaceable。As the AI industry thrives, Nvidia can charge a significant premium from downstream customers.
In response, Louis Navellier, founder of market consulting firm Navellier Associates, said that regardless of Nvidia’s financial performance, he would not sell Nvidia’s shares because “I have never seen a stock with such strong monopoly and such strong influence.”
Not only Nvidia at the top of the pyramid,Relatively downstream cloud service providers such as Alibaba Cloud and Baidu Cloud have also relied on AI technology to get out of the quagmire of losses.
Alibaba’s third-quarter financial report for the fiscal year 2025 disclosed that Alibaba Cloud’s revenue was 31.742 billion yuan, a year-on-year increase of 13%. Adjusted EBITA was 3.138 billion yuan, a year-on-year increase of 33%, mainly due to “public cloud revenue driven by AI-related products.”
Recently, Morgan Stanley also stated in a research report that starting from Q4 of 2024, Baidu Cloud’s profitability and profit margins have improved, and the trend should be sustainable. In Q1 of 2025, Baidu’s cloud business revenue will increase by 20% year-on-year.
Due to the promising prospects of AI technology, Alibaba Cloud is committed to increasing AI-related investment. On February 24, 2025, Alibaba Group CEO Wu Yongming announced that in the next three years, Alibaba will invest more than 380 billion yuan to build cloud and AI hardware infrastructure, with a total investment scale exceeding the total of the past ten years.
In this regard, Wu Yongming said,”The AI outbreak far exceeded expectations, and the domestic technology industry is in the ascendant and has huge potential. Alibaba will spare no effort to accelerate the construction of cloud and AI hardware infrastructure and promote the ecological development of the entire industry.”
As He Xiaopeng, chairman of Xiaopeng Motors, said,”Basically no one can make money in the language model. Except for Nvidia or Microsoft, Microsoft does not strictly make money. For example, OpenAI integrates money, not makes money, including Xiaochuan.” Combing the commercialization process of the AI model industry, Spicy found that there is a significant gap in the commercialization progress of upstream and downstream companies in the AI model. The downstream and closer to users, the more difficult it is to make money.
Since AI big model technology has become a major outlet in the technology industry, many downstream companies are scrambling to deploy related technologies. The computing power infrastructure provided by upstream chip manufacturers, cloud service vendors and other “water sellers” is highly sought after. Therefore, it is possible to generate revenue by providing computing power to downstream customers, and a closed loop of commercialization has been opened early.
For upstream “water sellers”, it is indeed a blessing to be the first to open up the closed loop of commercialization, but the healthy operation of the industrial chain is based on the positive cash flow of upstream and downstream enterprises. It has been difficult for downstream AI model companies to commercialize, and it has also been decided that the business story of upstream “water sellers” is just a rootless tree.
Because the AI model has strong comparative advantages in information retrieval, it can not only generate content, but also focus on semantic understanding and personalized recommendations, and the commercialization capabilities of traditional search engines have been confirmed. Therefore, many Internet companies have added more AI search, trying to commercialize the AI model through search functions.
In this regard, Minsheng Securities stated in its research report thatAI search is expected to become the first commercial C-end super application,”the ‘first light’ for the commercial implementation of large models.”
As a result, Internet giants such as 360, Baidu, and iFlytek, as well as latecomers such as Dark Side of the Moon, Smart Spectrum Qingyan, and Mita Technology, have begun to compete in the AI search market.
Source: AI product list
Facts have proved thatBecause it meets the needs of users, AI search products have indeed become the vanguard of the implementation of AI technology.Among the top 20 AI product list and website list in January 2025, AI search products accounted for three seats, namely New Bing, Nano AI Search and Perplexity AI. Among the top 20 domestic rankings during the same period, AI search products occupied four seats, namely Nano AI Search, Mita AI Search, Zhihu Direct Answer and C Know.
A typical representative of these is nano-AI search. As an “old cannon” in the Internet industry, as early as 2012, 360 entered the search market and launched a comprehensive search service in an attempt to divert Baidu’s influence. However, due to limited search capabilities and brand awareness, 360 ‘s search business has not been able to open up to the market.
As AI technology matures, 360 decisively increases its AI search and launches nano-AI search. At present, this product has become an industry leader.The domestic version of the AI Product List shows that in January 2025, the number of visits to the nano AI search web was 308 million, ranking first.
In parallel with the expansion of market influence, a number of AI search companies are also testing the water for commercialization. After sorting out the chopped pepper Spicy, it was found thatCurrently, there are three major orientations for the commercialization of AI search companies, namely subscription, advertising and selling APIs.
As one of the first players to get involved in the AI search business, Perplexity chose the subscription model as the first stop for the commercialization of AI. If users want to get more professional searches, use their own AI models, and analyze documents, they need to subscribe to the Pro version for $20 per month.
The Information reports that Perplexity AI is expected to have 550,000 premium subscription users in 2025, with annualized revenue of US$127 million. Even so, Perplexity AI is still difficult to turn around. In its official blog, Perplexity AI said,”The current subscription fee of $20 per month or $200 per year is not enough to support ambitious goals and growing revenue sharing plans with publishers.”
Yang Zhilin, founder of Dark Side of the Moon, believes that “charging based on the number of users will not create greater commercial value with the product. Subscription will not be the ultimate business model.”
Given the strong payment habits of European and American users, it is difficult for AI search companies to turn losses through subscription models. In the China market,Due to consumers ‘lack of payment habits, it is naturally difficult for AI search products to open up a commercial closed loop through the subscription model.
Source: Perplexity AI
As a last resort, in November 2024, Perplexity AI launched an advertising strategy. When users use Perplexity AI to search for questions, the “Follow up Questions” column at the bottom of the search results page displays advertisements provided by sponsors.
In fact,At present, a number of AI search companies in China have also targeted advertising models.
However, it should be noted that unlike traditional search engines, which can feed back large amounts of content in order, bid and rank the top-ranked content, and then let users choose their own differences, AI search has the ability to integrate information and the displayed results are extremely accurate. If manufacturers excessively pursue commercialization, it is likely to conflict with the interests of users.
For example, if a user searches for medical-related issues, AI search will put a huge amount of money spent by advertisers from Putian Hospital into the feedback results, which can easily trigger the “Wei Zexi Incident.”
In fact, this is also a key incentive why Internet companies dare not rashly let AI search advertising business enter commercial deep water areas. At the end of 2024, Zhou Hongyi, founder of 360 Group, said to the public,”In fact, it is very painful for us to do AI search. We also have self-subversion, and we can’t find a place to put advertisements inside for the time being.”
In addition to revenue-generating models such as subscriptions and advertising related to C-end users, AI search can also generate revenue by selling APIs to B-end customers.。For example, Bing provides Bing Search APIs for corporate customers, and companies can call relevant APIs to provide search functions for their own users.
Source: Bing
However, due to the high cost of AI models, the cost of using Bing Search API is not low。In May 2023, Microsoft raised the pricing of the Bing Search API. The price of the Bing API that supports ChatGPT is US$28 per thousand calls for less than 1 million requests per day; US$200 per thousand calls for more than 1 million requests per day. In response, Microsoft said that “the new pricing model more accurately reflects Bing’s continued technology investment to improve search.”
Due to the high cost of use, in April 2023, Brave Search cut off all connections to the Bing API, used its own index to display search results, and plans to launch its own Brave Search API.
Given that DeepSeek is open source and can be accessed at will by any organization, it is difficult for AI search to persuade downstream customers with certain technical capabilities to pay higher costs and purchase their own APIs.
In this regard, in May 2024, Wang Xiaochuan, founder and CEO of Baichuan Intelligent, said,”API revenue and the revenue of the big model itself are not the sexiest business models, and we are not going to catch up with and compare them.”
All in all,Because user needs, technical routes, and result presentations are completely different from traditional search engines, the three major commercial orientations of AI search, namely subscription, advertising, and selling APIs, are facing considerable challenges. It is difficult for related companies to achieve it overnight and become a search giant in the AI era.
In fact, the problem of poor commercialization faced by the AI model is not an isolated case. In the early days of PC and mobile Internet, many Internet companies also faced similar challenges.
Take Tencent as an example. Wu Xiaobo’s “The Legend of Tencent” records that at the end of 1999, OICQ had more than 1 million registered users. Because OICQ costs a lot every day and is difficult to make money, Tencent once had only 10,000 yuan in cash on its accounts and was on the verge of bankruptcy. Helpless, Tencent founder Ma Huateng plans to sell the company for 3 million yuan.
However, Ma Huateng visited a number of companies but failed to negotiate the sale of Tencent. Most companies only give Tencent a valuation of several hundred thousand yuan based on tangible assets. Lin Jun, author of “Fifteen Years of Boiling”, recalled that OICQ may be a project that seems to be growing rapidly,”however, no one in the world knows how it makes money.”
I believe many people are familiar with the story after that. Based on the new IM interaction method, QQ created a usage scenario for online chat, extending to advertising, QQ shows, QQ games and other services, making Tencent a fortune.
He Xiaopeng believes thatAfter interactive changes, the scene will change, and this process will bring huge commercialization opportunities.PC, QQ, Didi, Meituan and other products under the mobile Internet all rely on new interaction methods and transaction scenarios to improve efficiency and create more user value, thereby opening up a closed business loop.
Unlike the transition from the PC Internet to the mobile Internet, where only the carrier of content has changed, the AI model may involve new interactive carriers, methods, and scenarios, and the commercialization model may be very different from the traditional Internet.
Regarding the commercialization prospects of AI in the automotive field, He Xiaopeng is full of expectations and believes that “if L4 does it, it will definitely increase the software revenue or ecological revenue of the car.” For example, the automatic car washing of your car is one of the ecological income; automatic parking is the second ecological income; and automatic charging is the third ecological income.”
Of course, smart cars are just a vertical scene where AI technology is implemented. Beyond that, AI has more imagination. For example, Yang Likun, chief scientist of Meta, believes that as technology matures, AI glasses may integrate AI Agent functions, allowing users to handle daily things through just one pair of AI glasses, greatly improving work efficiency, creativity and productivity.
In the process, AI glasses may usher in the “iPhone moment”, and AI technology may also explore new commercial models as application scenarios widen. For example, shopping, navigation, watching immersive videos, etc. through AI glasses.
Clearly,If we look at the AI model from a development perspective, we can find that it is similar to previous technologies such as the Internet and mobile apps, and faced the challenge of commercialization at the beginning of its inception.
Fortunately, the needs of users are certain. A number of Internet giants and start-ups actively carry out functional and interactive innovation based on the needs of users. Once new usage scenarios are explored, the AI model may open up a business closed loop.
