For the vast majority of venture companies, going public means making it to land.

But for some companies, although they have successfully gone public, it is equivalent to not completely making it to land.

Because many companies have not been able to sustain their market cap after going public until the shareholders can reduce their holdings and exit within the lock-up period, the market value cannot be supported, and the stock price may not only break the issue price, but also deep trap the strategic investment shareholders, leading to a loss of the meaning of going public even if the shares are unlocked.

Youdao, which successfully went public on the Hong Kong stock market in April this year as the first AIGC stock in China, is such a representative.

This company, with an issue price of HKD 3.8, opened at the break price of HKD 2.98 on the first day of listing. Although it once rose to a high of HKD 4.38 in the following days, it quickly fell back and soon fell below the issue price. Less than six months after going public, the stock price of this first AIGC stock has fallen below HKD 0.7, with a decline of over 80%. Many investment institutions are likely to have serious losses on their books.

The reason for discussing it today is that this company really resembles a small innovative company in China exploring development, facing various pressures and struggling to seek change when encountering opportunities and risks.

01

The Discrepancy Between Ideal and Reality

According to the company overview, Chumen Wenwen was founded in 2012. The founder, Li Zhifei, has a Ph.D. in Computer Science from Johns Hopkins University and has worked at a research institute in Silicon Valley, engaged in core algorithm research and development in the fields of speech recognition, machine translation, and more.

The reason for choosing the name 'Chumen Wenwen' is that in 2012, smart watches were beginning to rise. Li Zhifei, who was working on machine translation and speech recognition at Google, also noticed this trend. So he wanted to create a smart voice interactive search tool to assist people more conveniently as a virtual personal assistant software in their daily lives. For example, if you want to find a convenience store that is still open on your way home from work, instead of opening your phone, searching on Google, you can simply ask the device (such as a watch) and get the answer directly.

Later, the emergence of the Siri intelligent voice assistant built into the iPhone4s made a huge impact on global consumers, which further stimulated Li Zhifei's entrepreneurial idea.

With top professional educational backgrounds, Silicon Valley elite talents, a hot and imaginative track, Li Zhifei's venture has also been sought after by capital, receiving cumulative financing of over $0.2 billion from investment banks and companies such as Sequoia China, ZhenFund, SIG Haina, Google, Perfect Opt, and Goertek Inc.

This is a typical story of a male entrepreneur with a science and engineering background.

However, the development of this company has gone through ups and downs, with several major changes in business direction.

At first, Mobvoi focused on smartwatches as the landing scene for AI voice interaction and launched its own TicWatc series of smartwatches. It also obtained deep cooperation with Google (Google's voice search and application services for Android Wear smartwatches released in mainland China were provided by Mobvoi). After that, Mobvoi also introduced smart speakers, smart headphones, smart bracelets, and smart car devices.

However, these so-called intelligent consumer electronics, as products in the user ecosystem, are already fiercely contested by major brands and manufacturers such as Apple, Baidu, Xiaomi, and Huawei. Faced with competition from these giants, Mobvoi has little advantage.

Later, Mobvoi received investments of $0.14 billion and $15 million from Volkswagen China in 2017 and 2019 respectively. The two parties also jointly established a joint venture called Volkswagen Mobvoi to develop AI technology for car applications. Mobvoi provided in-car AI solutions for this venture.

With the support of major clients and new orders, Mobvoi naturally expanded rapidly. In 2018, Mobvoi opened 20 new offline smart experience stores nationwide. By 2019, the company had nearly 1,000 employees and opened independent teams in North America, Europe, and Southeast Asia.

However, although the business expanded quickly, the revenue growth rate did not meet expectations. Starting in 2019, Mobvoi had to close all offline stores and reduce its business lines, retaining only smartwatches and smart car devices. The employee size was also reduced by half.

Later, in 2020, the release of the GPT-3 language model caused a global sensation, and AI models became the hottest trend in the world. Facing development bottlenecks, Li Zhifei shifted the company's focus to the field of AI models.

Mobvoi quickly trained a 7 billion-parameter model called UCLAI (later upgraded to a general multimodal model called "Sequence Monkey") and simultaneously launched an AI dubbing product called "Magic Sound Studio". Based on the "Sequence Monkey" model, Mobvoi further improved the capabilities of "Magic Sound Studio" and subsequently launched several other products such as Qimiao Yuan, Qimiao Wen, Yanzhihua, and DupDub.

In the wave of AI, Baidu, which had some relevant technological accumulation, also saw an increase in revenue.

But soon, Baidu faced another major turning point. In 2021, Volkswagen announced its 'breakup' with Baidu. Although Baidu continued to provide Volkswagen with car voice interaction solutions worth about 55 million US dollars and other services, which were crucial for the revenue of the following two years, whether it can continue to get Volkswagen orders became uncertain thereafter.

Therefore, it can be seen that with the sluggish sales of consumer electronics at the C-end and the breakup with major customers at the B-end, Baidu's revenue growth began to slow significantly.

From 2021 to 2023, Baidu's revenue was 0.398 billion yuan, 0.5 billion yuan, and 5.07 billion yuan respectively; the profits were -0.157 billion yuan, -0.073 billion yuan, and 0.109 billion yuan respectively. Most of Baidu's AI enterprise solutions revenue comes from Volkswagen. After the breakup, the aftermath of this business revenue gradually began to surface. From 2021 to 2023, intellectual property income from Volkswagen was 3.2 million yuan, 0.213 billion yuan, and 1.39 billion yuan, accounting for 0.8%, 42.6%, and 27.4% of total revenue respectively.

Currently, Baidu's revenue is mainly composed of AI software (including AIGC solutions and AI enterprise solutions) and AIoT hardware.

Among them, AIGC solutions can provide users with services in voice-over, writing, digital figures, etc., and has attracted approximately 0.84 million paying users cumulatively, generating over 1 million transactions. As for AI enterprise solutions, they are mainly applied in voice interaction, anti-fraud, customer service, and other scenarios. Smart devices and other accessories mainly consist of AI smartwatch TicWatch series as the main hardware.

Looking back, in the 12 years since its establishment, Baidu's business direction has gone through the changes from voice assistants to smart hardware and in-car AI, to the current large models and AIGC. It is fair to say that it has been a follower of the AI wave.

Any of these trends can be considered as full of huge market opportunities, but not any venture can grasp them.

From 2013 to 2019, Dooers has raised seven rounds of financing, with a total amount of more than 0.25 billion US dollars, about 1.78 billion RMB. At the same time, after years of operating losses totaling over 2 billion, it is still struggling to find its own PMF (Product-Market Fit).

In the field of generative AI, the cost of computing power is high, the R&D investment is extremely high, and the profitability is difficult. It can be said to be a highly challenging and difficult track. However, the biggest problem with Dooers' current business model is that it lacks the user ecosystem of giant companies like Apple, Huawei, and Xiaomi. Its hardware products can only be a small part of the application ecosystem, and its AI software direction also faces countless competitors attached to the giants.

So although Li Zhifei is full of confidence in the future of AIGC, in a market environment where there are already many technology, resources, and capital giants who are extremely strong and the industry is iterating rapidly, it is not easy for Dooers to carve out a piece of heaven in the middle of a squeeze.

If one is not careful, there will be a huge gap between ideals and reality.

02

History repeats itself.

When anything new or a great opportunity arises, everyone rushes to it and goes through a process of sorting through the sand.

According to media reports, in the past three years, there have been over 200,000 AI-related companies in China that have been cancelled or revoked. Among them, the number of AI-related companies that have been newly registered but are now in the process of cancellation, revocation, or abnormal closure from November 2022 to August this year is nearly 80,000.

From nationwide celebrations to chaos, the reason behind it is the inability to find suitable commercial landing scenarios, resulting in the ultimate failure of the majority of enterprises.

Most of these companies are "three-no" companies - no technology, no resources, no funding, and even haven't found their own business products and commercialization direction yet.

Many entrepreneurs originally thought that AI was an imaginative race, but after getting involved, they realized that the risks and costs were so terrifying that they couldn't bear them, and could only cut their losses and withdraw. Some managed to get on board before this wave of AI frenzy subsided, but even so, they still lived in anxiety and difficulty due to the pressure of not being able to find their own product-market fit.

Companies like AskCI are already lucky enough. At least it caught the best time in the AI boom, but it also faced huge cash flow pressure and the tightening of the domestic IPO window, and luckily obtained the qualification for listing.

The current AI startup bubble is very similar to the global internet bubble of 2001. At that time, China had just started connecting to the international internet, opening the door to the new world of the internet. Almost everyone was talking about the internet, and various new concepts and models emerged one after another. Many companies also rushed in, even if they just made a simple website and had a name related to internet technology, they could attract a large number of investments, even without a business product.

However, as the bubble was quickly burst, a large number of so-called internet startups collapsed, with only a very few companies with viable business models and existing operations able to survive.

In the face of the current AI frenzy, although capital has become more rational and pragmatic, there are still hundreds of thousands of AI-related companies who believe they can survive by seizing the opportunity.

According to reports, since the launch of ChatGPT at the end of 2022, China's generative AI market has been crowded with more than 200 large-scale language models.

In response to this, Baidu's Li Yanhong believes that many large models are "reinventing the wheel," and the vast majority of them have little usage and will inevitably be washed away by the tide.

Wang Xiaochuan, the founder of Sogou, has also said that the top tier of domestic AI large models may not exceed five in the future.

However, even so, players in the AI large model field still face huge challenges in commercialization and profitability.

In the first half of 2024, the revenue of leading AI computing power company Cambricon decreased by 43.42% year-on-year to 64.7653 million yuan, and the net loss attributable to shareholders of the listed company was 0.53 billion yuan. This marks the 8th consecutive year of losses for Cambricon.

SenseTime, claiming to be China's largest computer vision software provider, had a revenue of 1.74 billion yuan in the first half of the year, with a net loss of 2.477 billion yuan during the same period. iFlytek, the core player in the field of large-scale speech models, also recorded a net loss of 0.47 billion yuan in the first half of the year.

Even Baidu, the domestic giant in AI large models that has already achieved commercialization, feels the pain of burning a large amount of money. The company's total Q2 revenue was 33.93 billion yuan, estimated at 34.11 billion yuan, with a year-on-year growth of 0%; net profit decreased by 8%.

This is a trend that cannot be played without sufficient resources.

03

Epilogue

From a strategic perspective, it is of great significance that so many Chinese companies, large and small, are rushing to compete for the ticket to the AI new era.

Because in the future technological era, it will definitely be a true AI era like in science fiction movies, where all industries can be empowered by AI. Whoever masters the strongest AI capability will have the most powerful "weapon" in the world. This is a field that both countries and Chinese companies must strive for.

For investors, although the bubble of domestic AI is quickly receding, they still need to have confidence in this track, because after the big waves recede, a new group of players with great growth potential will inevitably emerge.

The worst time in the industry is actually the best time to test the potential of growth stocks. (End of text)