For the first time, the number of visits to ChatGPT showed negative growth, and the chill was transmitted to the securities market.
A group of AI concept stocks that experienced skyrocketing in the first half of the year died down, and well-known stocks such as BlueFocus, 360, and Kunlun Wanwei all fell by more than 30% from their high points. Microsoft and Nvidia were not spared either, and there were also callbacks to varying degrees.
Looking back at this roller coaster-like market performance, many AI concept companies have reduced their holdings of stocks, and even as many as 9 shareholders of listed companies have equity changes due to divorce. They exited in such a dramatic way in the rapidly blowing AI bubble market.
Funds fled, the aura of ChatGPT faded, and pessimistic arguments followed. But there are more signs that the big language model is moving towards the B side.
On July 7, OpenAI fully opened the GPT-4 API, allowing developers and companies to create more powerful generative AI applications with the help of large models. Baidu and Ali have also aimed their efforts at the B-side of their large models, allowing AI to not stop at dialogue, but to participate in the transformation of various application scenarios.
As the tide recedes, more killer apps are brewing.
Over 300 of 473 AI concept stocks fell
The excitement brought to the world by the big language model is declining, and even the popularity of the "top stream" ChatGPT has also gone downhill.
According to data from the website analysis tool SimilarWeb, in June, ChatGPT’s global visits fell for the first time, down 9.7% month-on-month. In fact, in the first five months of this year, the month-on-month growth rate of its visits has gradually declined, being 131.6%, 62.5%, 55.8%, 12.6%, and 2.8%, respectively.
The traffic and popularity continue to decline, and the time users spend on the site has also decreased. The data shows that the average access time of ChatGPT users dropped from 8 minutes and 32 seconds to 7 minutes and 48 seconds.
ChatGPT cooled down, and the crazy rise of AI concept stocks came to an abrupt end.
According to Choice data, from June 21 to June 30, the A-share artificial intelligence sector fell by 5.38% as a whole. Among the 473 concept stocks, more than 300 stocks fell. In the U.S. stock market, Microsoft, Nvidia, and "chip rookie" AMD, which are highly related to the AI industry, have also fallen to varying degrees from their highest point.
"Being stuck" has become the status quo of many investors chasing AI concept stocks. According to the statistics of "Metaverse Daily Explosion", as of the stock market close on July 7, many AI concept stocks that had skyrocketed in the first half of the year have ushered in a tragic correction, and there are many well-known stocks among them. HKUST Xunfei, Tiandi Online, BlueFocus, 360, Kunlun Wanwei and other former "monster stocks" have all fallen sharply, with corrections of 16.33%, 29.39%, 37.52% and 44.27% from the year's highs, respectively. Funds come in fast and go out quickly.
Many AI concept stocks experienced a sharp correction
With the cooling of the concept of AI, cracks began to appear in the rapidly blown bubble. What is striking is that when the market is full of voices, the wave of shareholding reduction of listed companies is already undercurrent. According to statistics, among the 203 listed companies in the AI sector of the A-share market, 79 companies have issued shareholding reduction announcements during the year, accounting for nearly 40%.
The directors of the chip giant Nvidia also reduced their holdings and cashed out at the high point in June. According to documents disclosed by the US Securities and Exchange Commission (SEC), several directors of Nvidia intensively reduced their holdings of nearly 290,000 Nvidia shares within 4 days, worth about 120 million US dollars.
At the same time, the "divorce wave" is also staged in A shares. Since the beginning of this year, 9 A-share listed companies in the AI sector have disclosed changes in company equity due to divorce, including 360, Tongcheng New Materials, Kexin Technology, Tiandi Digital, Tonghe Technology, Huitian New Materials, Saiteng, Zhuo Victory Micro.
Shareholders of listed companies get divorced, and most of the results point to a disguised reduction in holdings. Take 360 as an example. On April 4, 360 issued an announcement stating that Zhou Hongyi, the actual controller, and his ex-wife Hu Huan reached a property division agreement. Zhou Hongyi plans to divide 6.25% of the company’s total shares directly held by him to Hu Huan At that time, the market generally questioned that he used divorce to reduce his holdings, and the stock price of 360 also peaked that day.
Zhuo Shengwei announced on the evening of June 20 that Tang Zhuang, one of the company's actual controllers, divorced his wife Yi Gebing. According to the "Divorce Property Division Agreement", Tang Zhuang transferred 32.7575 million shares of the company he held to Yi Gebing occupies 6.14% of the company's total share capital. Calculated based on the stock price at that time, the market value corresponding to this "breakup fee" was as high as about 3.4 billion yuan.
Listed companies were busy cashing out during the bubble period, and shareholders in the secondary market unfortunately became "takers". In the first half of the year, the AI fever that was hyped vigorously left a lot of chicken feathers in the blink of an eye.
The flow ebbs and the big language model goes to the B side
CITIC Securities believes that the decline in ChatGPT user traffic is mainly due to the decrease in freshness of individual users.
According to SimilarWeb data, ChatGPT has exceeded 100 million active users two months after its launch. This is because its leapfrog experience has won the popularity of users all over the world. According to the Google Popularity Index, from October 2022 to April 2023, the popularity of the keyword "AI" increased by 267%.
Since May, the popularity of the "AI" keyword has stagnated, and in the week from May 14 to 21, there has been the largest decline in nearly a year. After many users have experienced ChatGPT, they found that the main usage scenarios of this chat robot with a wide range of knowledge are still partial office purposes such as text summarization and basic code writing. Most individual users do not have strong demand stickiness for it.
Will ChatGPT be cool?
In the view of CITIC Securities, the decline in its popularity is not only related to external reasons, but also due to the active adjustment of its main company OpenAI's operating strategy-constrained by the cost of computing power resources, it shifted its focus to the B-end market.
This view coincides with Chen Jia, an independent strategy researcher. Chen Jia said that OpenAI itself is very worried about the overspeeding of ChatGPT's early traffic, because its background server is overwhelmed. Therefore, OpenAI took advantage of the opportunity of GPT upgrade to 4.0 to not only launch paid services, but also increased the rate, hoping to divert out non-core development-level consumer users who are chasing popularity and traffic.
From a product perspective, for more than half a year, OpenAI has not made more optimizations in the scenarios for C-end users except for the launch of the GPT Plus service. And there are more and more signs that OpenAI is focusing on the B-end market.
Recently, OpenAI CEO Sam Altman revealed in an interview that the work content of the latest quarter will focus on the development of API services for enterprises and continue to deepen cooperation with Microsoft.
On July 7, new trends came out, and OpenAI announced on its official website that the GPT-4 API is fully open for use. All paid API users now have direct access to GPT-4. OpenAI is expected to open the GPT-4 API to new developers by the end of July.
GPT-4 API is fully open
Previously, in order to ensure the provision of safe and stable services, developers have been able to use GPT-4 only through supplementary applications. With the full opening of the GPT-4 API, developers around the world can use it to enhance their own applications or develop A new generative AI application.
Compared with ChatGPT's own capabilities, the addition of large language models and industry-side applications allows users to experience the power of AI more conveniently and efficiently. Just like Microsoft integrated ChatGPT technology into Office 365, allowing AI to directly exert its capabilities in office scenarios, including summarizing meeting minutes, efficiently creating PPT, drafting emails, and more.
ChatGPT and the large model behind it are like a base, which needs to be combined with scenarios of different vertical industries to produce applications with an intuitive experience. Not only ChatGPT, but Baidu's Wenxin Yiyan and Ali's Tongyi Qianwen are also more focused on transforming the B-side.
After the release of Wenxin Yiyan, the Baidu Smart Cloud Large Model Platform was launched immediately, providing not only large-scale model services including Wenxin Yiyan, but also a complete tool chain for large-scale model development and application. Zhang Yong, the leader of Ali, "Xiaoyaozi" also said that all Alibaba products will be connected to the large model of Tongyi Qianwen for a comprehensive transformation.
After experiencing the cheers and departures of C-end users, the big language model turned to the B-end. Under the nourishment of the wisdom of developers around the world, it has blossomed. When ChatGPT is no longer thriving, the AI application revolution will begin.