CCV Capital’s Wei Zhou: AI Winter Coming in Second Half of Year, Remaining Genuine as Fake Ones Eliminated, Silicon Valley Startups Only Need to Survive to Next Cycle

SANTA CLARA, Calif., July 10, 2024 — “It’s very popular for Silicon Valley startups nowadays to say they only need to survive until the next cycle,” said Wei Zhou, who witnessed multiple cycles in venture capital industry, from his time at KPCB to now as Founding Managing Partner of CCV Capital.

This year, the primary market remains bleak, many VCs are facing the fundamental problem of capital shortage. However, Wei emphasized that in the entire VC investment field, AI is undoubtedly the direction where all investment institutions have invested the most and largest amounts this year, both in terms of number of deals and dollar amount.

In addition to “hot” this year, many “pseudo-entrepreneurs” and “pseudo-demands” have also emerged. Having witnessed the ups and downs of cutting-edge technology venture capital industry over the years, Wei realized: the AI winter is imminent.

“I predict that AI will cool down starting from the third quarter of the second half of this year, and remain cold until the second half of next year.”

Wei recalled the now classic case of Shukun Technology. The years 2015-2017 were when AI medical imaging applications in China were at their hottest. Many companies received huge financing. However, after 2017 the enthusiasm cooled sharply. Due to the previously overvalued AI imaging companies unable to generate revenue, the primary market remained cold from 2017 until early 2019.

“Nothing can grow in the winter cold, but it will also retain some particularly robust seedlings. So when winter passes, these few remaining seedlings will grow very fast. The business environment has also improved, the products are more mature, and meanwhile the competition has reduced.” 

Always daring to invest in cold sectors, Wei seized the timing in 2018 when AI medical imaging was at its coldest, and led Shukun Technology’s US$200 million Series A+ round. At that time, he judged that China’s AI medical imaging industry was ready, the technology was mature enough, and commercialization was the only thing lacking.

As represented by Shukun Technology, after 2019, the business environment and real application scenarios of AI imaging companies began to emerge. Sure enough, a number of companies that survived the winter saw their revenues rise.

The once wildly popular metaverse concept also basically fits this cyclic curve.

“It was hot for a year and a half before cooling, and still hasn’t warmed up,” he predicts that AI will follow the same pattern to cool for a year or year and a half. “So we’re not worried about AI. It’s very popular nowadays in Silicon Valley for startups to say they only need to survive until the next cycle.”

In Wei’s words, although everyone knew from the beginning that AI was a great direction, exactly where and how to land commercial applications was still unclear to all.

This reminds him of 2000 when everyone was investing in the Internet. Looking back, everyone now certainly knows the best entrepreneurial direction at that time was portal sites. But if you went into e-commerce and other sectors that rose later at that time, because the overall environment did not match and the market was not yet ready, it actually could not take off.

Wei bluntly said, AI entrepreneurs today also face the problem of timing. 

“Choosing what to do at this point in time is most appropriate. You may have great vision, but if you’re too early and the surrounding environment is not ready, that can be terrifying.”

In Wei’s view, AI is still just at the beginning application stage, and too many people are trying in different directions, but over the past year and a half, the overall AI environment has gradually shown some of the industry disorders of the cryptocurrency field back then.

“It became too hot, and many of those who rushed in were not thinking about how to use AI to solve actual existing problems, but were using various concepts to speculate. We call these pseudo-entrepreneurs, people chasing waves.”

From March last year to February this year when Sora came out, AIGC instantly became a star that sparked public enthusiasm around the tech arena.

Text-to-image, text-to-video, image-to-video, seem to have become buzzwords in the AI circle. 

Wei bluntly said, “Some money was invested in places where it was wasted.”

The current AIGC direction is like Internet startups back then. Take online stores for example, anyone who understands the basics of the Internet can participate. 

“Hard to say if you can win, but you can play in it.”

In his opinion, the current entrepreneurial environment in the AIGC field is similar.

“Basically, if you understand AI, just like Internet startups back then, as long as you notice some market demand, you can dive in to do it. You can enter just with your so-called ordinary business sense.”

He emphasized that generative AI technology is crucial, but “without differentiated technology far surpassing others, achieving technological monopoly, it will eventually become a so-called mass consumer product – every site will have it, everyone uses different vendors, the products won’t differ too much.” If this continues, companies with high value will be very few, “there will be a large batch of barely surviving companies.”

Under such circumstances, Wei believes it will be great if AI cools down in the second half of this year. “It will eliminate some so-called pseudo-entrepreneurs and pseudo-demands, making the environment cleaner. Some very solid entrepreneurs who are using technology to solve real problems will succeed.” 

Specifically, he believes that B2B applications that can influence industry cost structures and entire industry workflows are truly valuable AI applications. “For B2B, if your understanding of the industry is superficial, you can’t even get your foot in the door.” He analogized such applications to the “shield machines” used to build highways. 

But B2B applications also have higher requirements for founding teams.

Not only must they have sufficient understanding of the AI industry and progress, they also need to know the roles of each participant in the workflow, and be able to combine their own AI products with the entire workflow to produce a “shortcut effect”.

“They can sensitively feel which part of the costs the AI application can chop out of the workflow – not reducing, but making the participating side that previously took a big piece of the pie disappear, then your value is huge.”

In the interview, Wei mentioned Huski.ai, a Silicon Valley B2B startup he invested in. Huski.ai mainly provides cost-effective solutions for IP lawyers and their staff to help them review, monitor and enforce trademarks for clients.

Wei said law firms previously handled at most 10 cases a year, because the economic and manpower costs of evidence collection for law firms and brands to deter counterfeiters were very high. After Huski.ai intervened with AI, evidence collection costs were infinitely reduced, even directly replacing the evidence collection process, no longer subject to manpower constraints.

For entrepreneurs, Wei introduced three advantages of B2B entrepreneurial directions to Sohu Science and Technology. First, there are currently fewer teams with cross-border capabilities, so competition is relatively less fierce. Second, revenue comes relatively fast. Third, it can cause thorough changes to an industry.

Therefore, Wei frankly said that after investing in large model and platform companies, he now prefers “companies that cut very deep into a vertical field”.

Chinese AI companies cannot afford to expand globally only after becoming big, otherwise they will be “rolled to death” domestically, Wei said.

Although from the perspective of the AI field, he currently believes it has not reached the point of “unable to survive without going global”. But he emphasized, “Competition among Chinese companies domestically is extremely fierce, like the dark forest in The Three Body Problem. In the age of deglobalization, Chinese companies must still adhere to globalization.” 

“From day one, Chinese AI companies must consider and prepare in advance. Not like before, first growing big domestically then expanding globally. By then you will encounter a lot of obstacles and guard mentality. That is to say, from day one, as long as you have sufficient confidence and resources, begin implementation in parallel.”

For entrepreneurs, he advised focusing on future global target markets early on in the initial stages, establishing localized teams from partners to services with a “local image” in the home countries. 

“But you actually have close ties with Chinese companies and supply chains.” That is to say, while fully integrating with local markets, culture, regulations and so on to build a localized image, also maintain connections with China in resources, talent, supply chain and technology to form a distributed structure. “Becoming like many mushrooms sprouting simultaneously from different places, but with a certain degree of mutual support between them.”

Wei emphasized looking for investments to support Chinese innovation when making investment decisions. “In the past, 15-20% of our projects were overseas, certainly mostly related to China, either the founders were Chinese, or linking closely with China’s supply chain and market or talent.”

Wei pointed out that if a company can obtain sufficiently unique data in a vertical field to train models, it can create barriers. Take Shukun Technology as an example. After years of R&D and exploration, Subtle Medical used its unique medical data to self-research and train the multimodal medical model ShuKunGPT. 

In addition, if you can understand and grasp an industry deeply enough, you can also form leading advantages and barriers. In such vertical scenarios, “others can hardly catch up with you overnight, even if you have a lot of money, this is still an advantage.”

In summary, have to focus on subdivided industries. Once relatively unique vertical opportunities are identified, quickly build closed data loops. 

“In any event, I feel like under today’s environment, if entrepreneurs want to avoid the situation of ‘nothing grows under huge trees’, they still have to think clearly – exactly what is the direction and advantage of my startup.”

SOURCE CCV Capital

Leave a Reply

Your email address will not be published. Required fields are marked *