"Tencent is the biggest beneficiary of DeepSeek's Open Source" Zhu Xiaohu appears and does not discuss embodied intelligence: AI application entrepreneurs must dare to admit that they have no barriers.

After the remarks "The commercialization path of embodied intelligence is unclear" and "Recently exiting some embodied intelligence projects" triggered several days of market controversy, on March 31, Zhu Xiaohu, the managing partner of Jinsha River Venture Capital, appeared at the "AI Future Forum: Innovation · Investment · Globalization" held during the "2025 Zhongguancun Forum Annual Conference."

At this forum themed around AI (artificial intelligence) investment, Zhu Xiaohu did not mention embodied intelligence, and his speech still focused on the AI applications he has been most optimistic about in recent years. Shortly after the speech ended, Zhu Xiaohu hurriedly left the venue.

The emergence of DeepSeek has brought about a shift in the landscape of the large model field. Zhu Xiaohu believes that Tencent has become the biggest beneficiary of DeepSeek's open source. At the same time, AI application companies are experiencing explosive revenue growth, with some companies achieving a month-on-month increase of over 20%, and small teams realizing tens of millions of dollars in revenue within six months to a year.

In the face of the AI application startup wave, Zhu Xiaohu stated, "Entrepreneurs must dare to admit that they have no technological barriers." The core of competition in AI applications lies not in the foundational models, but in product, market, and operational capabilities. He advised domestic AI application companies to seize opportunities in overseas markets and to layout in regions with less competition, such as Japan, Southeast Asia, and the Middle East.

Organizer provided photo

Many AI application companies are beginning to achieve explosive revenue growth.

"In fact, over the past 20 years, the internet has gone through many cycles of opportunity. Today, these opportunities have proven to have viable business models, but at the time they all experienced a bubble period, faced ridicule, fell into a valley of death, and then developed again from the valley of death, ultimately achieving success." Zhu Xiaohu stated at the beginning of his speech.

For example, during the era of portal websites, the attention economy also faced questions about how to monetize, but today traffic has become a recognized basis for business models. In the long video industry, the focus of discussion during the Youku and Tudou era was whether users were willing to pay, while today, the number of paid subscribers for long videos in China has exceeded 100 million. Similarly, industries such as ride-hailing, food delivery, bike-sharing, and shared power banks have all experienced initial controversies but ultimately established their business models.

Zhu Xiaohu believes that the past five or six years have undergone a similar journey. Concepts such as the metaverse, SaaS (Software as a Service), and large models have successively become hot topics, but quickly cooled down in a short period. "At the beginning, there was a lot of hype, but it was quickly disproved. Today, no one is looking at the metaverse, and no one is looking at SaaS anymore."

However, similarly, new hope may also be孕育 in the "valley of death."

Taking the metaverse as an example, Zhu Xiaohu believes that although Apple's Vision Pro is technologically advanced, its commercialization performance is poor. In contrast, Meta's Reverb series products have achieved sales of over 2 million units by simplifying and focusing on basic functions. In light of this trend, Zhu Xiaohu's judgment is that China may see a large number of AI glasses released by 2025, and the metaverse may return to the market in another form.

SaaS software has also gone through a process from feverish pursuit to market coolness. In 2020 and 2021, Chinese investors frantically sought after SaaS companies, even giving system revenue and customer customization revenue valuations of 10 times and 20 times. However, as revenue growth slowed, the SaaS industry rapidly cooled down.

"However, after the launch of DeepSeek this year, we also feel the possibility of spring." Zhu Xiaohu stated that in March of this year, he observed that some SaaS companies had a year-on-year revenue growth of over 3 times. Companies that actively embrace AI and DeepSeek, if they can maintain such growth rates, are expected to usher in a true spring. "Software companies that raised a lot of money at high points and burned a lot of money may not see spring. I believe all software companies that strictly adhere to business discipline might see the dawn of an end."

The same is true for the large model track. Zhu Xiaohu stated that the large model has been extremely popular over the past two years, but after the emergence of DeepSeek, the market has clearly started to cool down. Many AI application companies have begun to achieve explosive revenue growth. He believes that the revenue growth rate of AI application companies is comparable to that of the group buying industry back in the day, "This is very impressive." Some companies have seen month-on-month growth exceeding 20%, and some have achieved millions of dollars in revenue within half a year to a year. These companies often have streamlined operations, achieving rapid growth with just a team of a dozen people.

Call on AI application entrepreneurs to dare to admit that they have no barriers.

"The emergence of DeepSeek has indeed brought about a tremendous change to the entire ecosystem of large models." Before the emergence of DeepSeek, Zhu Xiaohu's view on open source and closed source was that when closed source models hit a wall, open source models would gradually catch up, ultimately leading to two ecosystems: iOS and Android. "However, DeepSeek is completely different; it is even more open than Android. It not only does not seek commercialization, but it doesn’t even want users."

Zhu Xiaohu believes that this is also the reason why Tencent is fully embracing DeepSeek. In his view, Tencent is the biggest beneficiary of DeepSeek's open-source model, which, thanks to the advantages of the open-source model, allows its Yuanbao large model to quickly surpass other competitors in the market.

"No one may have expected that the biggest beneficiary after DeepSeek's launch is Tencent," said Zhu Xiaohu. He pointed out that a major characteristic of technology investment in China is that it is difficult to commercialize before the technology matures, but once it matures, it will quickly become widespread, leading to technological equality. In the future, both investors and entrepreneurs need to consider a question: who will be the biggest beneficiary when technology becomes widespread? "Therefore, we have always advised startups not to waste a single penny on training underlying models, but to fully embrace open-source models and fully embrace APIs (Application Programming Interfaces)."

"Although many people say that the Scaling law may still move forward, it is no longer important." Zhu Xiaohu believes that currently, foundational large models already possess commercial capabilities. Whether it's DeepSeek, Tongyi Qianwen, or Doubao, they have all demonstrated commercial capabilities in areas like customer service, sales, meeting minutes, marketing copy, and voice teaching, with a month-on-month growth rate of over 20%. For entrepreneurs, the foundational model is not important; what matters is the product and user demand. Understanding customers, creating products that meet their needs, and enabling users to quickly transition from trial to paid.

Zhu Xiaohu mentioned that many investors today are hesitant to invest in AI applications, believing that AI applications lack technological barriers. He thinks this viewpoint is similar to the doubts raised during the mobile internet era more than a decade ago. At that time, companies like Didi, Ele.me, and Xiaohongshu were all considered to have "apps without barriers."

AI application companies are no exception. "Today's AI application entrepreneurs are the same; they must be brave enough to admit that they have no barriers. Any technical barriers are just a deception; there are no barriers at all." Zhu Xiaohu believes that the core competitiveness of AI applications lies not in the base models, but in product, market, and operational capabilities. "All applications are just shell applications, but I can build barriers in many non-AI capabilities," such as workflow integration, editing capabilities, proprietary hardware, and manual delivery.

Moreover, both the Chinese and American markets are witnessing the growth of AI application companies. Zhu Xiaohu observed that American VCs (venture capitalists) are no longer investing in large models, but are instead concentrating their funds on AI applications, which has intensified competition among startups and even led to price wars. The competition in the Chinese market is equally fierce, and he advises Chinese AI application entrepreneurs to "refine their products and teams domestically, and then go overseas"; currently, Chinese AI application companies have almost no competitors in Japan, Southeast Asia, the Middle East, and other regions.

"China is indeed very strong in the implementation of AI applications," said Zhu Xiaohu.

(Source: Daily Economic News)

Source: Eastmoney.com

Author: Daily Economic News