Sora Ignites A-Share Concept Stocks! Unveiling the Truth Behind 68 Sora-Related Stocks
-
OpenAI's text-to-video model Sora has emerged, reigniting the generative AI frenzy.
On February 20, Zhidongxi reported that today, at least 66 Sora-related stocks surged in A-shares and Hong Kong stocks, with over 20 brokerages releasing hot research reports on Sora.
On February 16, OpenAI重磅推出其首款文生视频大模型Sora, which can generate up to 1-minute videos based on prompts, achieving a generational leap compared to AI video startups like Runway and Pika. Shortly after, OpenAI was reported to have surpassed an $80 billion valuation in its latest funding round. Sora has recreated the frenzy seen when ChatGPT was launched a year ago: social media platforms both domestically and abroad are flooded with videos generated by Sora across various themes; tech giants such as Elon Musk, Yann LeCun, Jia Yangqing, Jim Fan, Xie Saining, Zhou Hongyi, and Li Zhifei have engaged in heated debates; at the time of Sora's release, it even managed to divert almost all attention away from Google's Gemini Pro 1.5, which supports a context of up to 1 million tokens.
Among the 68 concept stocks, five players saw increases exceeding 40%, with Huichang Communications up by 44.05%, Chinese Online by 44.04%, Danghong Technology by 44.03%, Insight Group by 43.99%, and Annoqi by 43.91%; 18 players saw increases between 20%-40%, while 21 players experienced gains between 10%-20%; only SenseTime and EZVIZ Network saw slight single-digit declines in their stock prices. From the top five gainers, it can be seen that the popularity of Sora concept stocks is concentrated in the video industry, media industry, and computing power players. Among them, Huichang Communication and Danghong Technology are video solution providers, Insai Group and Chinese Online are related to the media industry, and Annoqi was included in the concept stocks possibly because it plans to acquire computing power player Shanghai Gencong.
It is worth mentioning that even companies completely unrelated to the AI industry have appeared in the concept stocks, such as Dongcai Technology, whose main business is high-speed resin material manufacturing. The company responded that its high-speed resin materials have been scaled up for use in OpenAI's AI servers. Additionally, financial technology player Xinyada was included in the concept stocks because one of the founders of AI video generation startup Pika, Guo Wenjing, is the daughter of Xinyada's chairman Guo Huaqiang. The following shows the latest stock performance of companies labeled as 'Sora concept stocks' on mainstream trading platforms such as Tonghuashun and Tencent's self-selected stocks:
▲List of Sora concept stock companies and their stock price changes from February 8th to 20th (covering A-shares and Hong Kong stocks)
On the first trading day after the Spring Festival, Sora concept stocks became the focus of this capital feast. Since yesterday, multiple concept stock players have publicly responded through channels such as investor interaction platforms. Among the 68 Sora concept stocks, there are mainly four types of players, including companies like ArcSoft and Wondershare that have explicitly responded to developing text-to-video technology and products, as well as firms like CloudWalk and Hanwang Technology that have revealed related technical expertise. Additionally, stocks of companies like Hikvision, Dahua Technology, and Sinodata that have explicitly stated they are unrelated to Sora are also included. Beyond these, stock prices of upstream and downstream companies in the text-to-video large model industry, such as those in the film, media, and AI computing power sectors, have also risen.
Among the 68 Sora concept stocks, several have explicitly responded to ongoing research and development of text-to-video-related technologies and products.
Companies like ArcSoft, BlueFocus, and Wondershare have already released corresponding products, with their stock prices rising by 11.96%, 10.76%, and 24.64%, respectively. Computer vision algorithm service provider ArcSoft announced yesterday the upgrade of its core AI model engine ArcMuse, which will support automated commercial video generation for business shoots. BlueFocus responded by stating it had previously released an AI video production platform called BlueFocus Avatar based on TTS speech synthesis technology.
Last year, Wondershare launched its billion-parameter multimedia AI model "SkyCanvas," with core features including one-click video generation.
Some companies have revealed that they have corresponding R&D plans, but no products have been released yet.
The securities department of brand marketing company Insai Group told China Securities Journal·Zhongzheng Taurus reporter that its AIGC project team will develop text-to-video functionality in March, followed by the launch of a public beta version in the second quarter of this year to officially commence commercialization. Additionally, Insai Group responded to investors on the investor interaction platform that its self-developed InsightGPT belongs to a vertical application-layer model for the marketing industry, which is not in direct competition with foundational general large models like Sora. Instead, it can leverage general large models such as Sora to enrich its video material library and improve video generation efficiency and quality. Several companies currently have not released any text-to-video products and have not disclosed whether they have a clear connection with Sora-related technologies or products, but they mentioned having technical accumulation in related fields.
AI vision company CloudWalk Technology responded on the investor interaction platform that it has deep accumulation in the field of vision. The Congrong large model has already implemented generative technologies such as Diffusion and GAN to model character images and video data, enabling the generation, control, and editing of image and video content through text and audio.
iFlytek, a major player in the domestic large model sector, revealed on the investor interaction platform that its Spark large model already has relatively basic text-to-video capabilities. Yesterday, a representative from Hanwang Technology's securities department responded to China Securities Journal reporters, stating that the company possesses multimodal recognition technologies including text recognition, image recognition, and video analysis.
Danghong Technology, a provider of intelligent video solutions and video cloud services, revealed in an interview with Beijing Business Today that they have made progress in text-to-video generation technology, with ongoing breakthroughs in video duration and quality.
Similar to previous ChatGPT concept stocks, multiple companies among Sora concept stocks have begun issuing statements to distance themselves from the association since yesterday. Zhongke Information, an IT service provider specializing in intelligent recognition and analysis technologies, has explicitly stated that it is not engaged in related business.
Yesterday, security industry leaders Hikvision and Dahua Technology told The Paper that their multimodal large model developments differ from Sora's direction. While Sora generates videos from text, Hikvision focuses on smart IoT applications with perceptual intelligence, and Dahua specializes in AI-based video analysis.
Huitong Communications, a one-stop cloud video platform and solution provider, clearly stated that the company currently has no involvement in OpenAI's Sora intelligent video model business. AI vision technology provider Jiadu Technology stated on an investor interaction platform that the company does not have products similar to Sora.
Apart from directly related companies, concept stocks in the AI application and computing power sectors have also risen.
For example, stocks of several media-related companies such as Chinese Online, Guomai Culture, and Zhangyue Technology have increased, with Chinese Online surging by 44.04%, and the other two by 25.01% and 21.05% respectively. Many of these companies have responded that they are actively monitoring the development of Sora and exploring its implementation and application in the media sector. Today, Mango Excellent Media responded on an investor interaction platform, stating that it will actively explore the practical application of emerging AI technologies such as text-to-video in the media field. Yidian Tianxia looks forward to the potential improvements and benefits Sora could bring to its video advertising business once it becomes widely available.
Another category includes domestic AI computing power companies like Hygon Information, Sugon, and Digital China. The reason is that compared to text and images, the large-scale adoption of video generation applications may lead to a substantial increase in computing power demands.
Within just three days of its release, Sora has become a hot topic in the secondary market. The current surge in related stock prices can be compared to last year's ChatGPT concept stocks, when Hygon Information's stock price rose by as much as 200% in a month. It's possible that the current hype around Sora concept stocks will continue for some time. ▲2023 ChatGPT concept stock companies list and stock price changes
In addition to the soaring stock prices of concept stocks, within just five days, 20 securities firms have released 26 research reports.
Almost all securities firms have given Sora positive evaluations. Guotai Junan called it a "milestone in AI-generated videos," Western Securities mentioned "Sora ushering in a new era of AI," and Tianfeng Securities believes this signifies "the arrival of the iPhone moment for the AI industry." ▲Statistics of titles from 26 research reports on OpenAI's text-to-video large model Sora by 20 brokerages
In the research reports, brokerages analyze the stunning effects of videos generated by Sora while also focusing on the impact of Sora's release on domestic AI applications and the computing power industry.
Among them, Tianfeng Securities' report mentions that with OpenAI's significant improvement in Sora's text-to-video model capabilities, content creation workflows are expected to be disrupted, and the prototype of the next billion-user internet platform has emerged. The $200 billion short video creation ecosystem is expected to be the first to be disrupted, and the substantial progress of generative AI in video creation and world models will penetrate downstream application scenarios such as videos, 3D, and games. Huatai Securities has explicitly mentioned that Sora is expected to lower production costs in fields such as film, short videos, and gaming.
In terms of computing power, due to the significantly larger data requirements for video training compared to text training, Guotai Junan Securities predicts that the Sora model will further widen the computing power gap.
Similarly, CITIC Securities believes that the large-scale application of multimodal models like Sora will not only require higher standards for computing power upgrades but also pose significant challenges to the network capacity within existing AI computing clusters, leading to stronger demand for related network equipment upgrades. In China's generative AI industry, according to incomplete statistics, there are currently at least 13 companies advancing the research and development of text-to-video related technologies or products. These include internet giants like Baidu, Tencent, Alibaba, and ByteDance, as well as video industry companies such as New Y Tech, ArcSoft, and Wondershare.
▲ Statistics on the progress of some domestic companies in text-to-video large model development
ByteDance previously launched the text-to-video model Boximator, which can precisely control the movements of characters or objects in generated videos through text. Today, ByteDance responded that Boximator is a research project on motion control methods in the field of video generation and is currently unable to be implemented as a mature product. There is still a significant gap compared to leading foreign video generation models in terms of image quality, fidelity, and video duration. Yesterday, ArcSoft announced that its core AI model engine, ArcMuse, has been upgraded to support automated generation of commercial videos for product photography. In terms of product functionality, ArcMuse has achieved two core features: one is the automatic generation of product showcase videos, and the other is the automatic generation of fashion model videos.
Additionally, many prominent figures in the AI industry have previously been reported to venture into text-to-video model startups. For example, in December last year, IEEE Fellow and University of Hong Kong professor Xu Dong's startup, Xutu Intelligence, emerged, offering services such as text-to-video based on large visual models.
Wang Changhu, former head of visual technology at ByteDance, founded Aishi Technology, which launched the video generation tool PixVerse, capable of producing 4K resolution high-definition videos. Furthermore, using PixVerse, AI film creator Xianren Yikunsheng generated the viral video Shan Hai Qi Jing (Mirror of Mountains and Seas). ▲PixVerse by Aishitech generates 'Mystical Mirror of Mountains and Seas' video (Image source: Aishitech official account)
Since last year, the text-to-video sector has been gaining increasing attention. The emergence of Sora has quickly positioned it at the center of the generative AI competition, making this year the inaugural year for AI video generation.
Undoubtedly, the debut of OpenAI's first text-to-video model, Sora, instantly outperformed numerous products from previous players in this field like Pika and Runway. It excels in terms of video length, realistic effects, and text comprehension capabilities. However, even with its powerful capabilities, Sora's generated videos still contain numerous errors that defy biological common sense, sparking debates among experts about whether it truly understands the physical world.
At the same time, even before Sora has officially opened for trials, a flood of fake videos claiming to be generated by it have appeared on social platforms, muddying the waters and raising concerns in the industry about the future development of video generation.
Therefore, from the perspective of product maturity and the prospects for large-scale commercial application, the development of the text-to-video field still requires a long time. Amid the current hype around concept stocks, it's even more crucial for people to enhance their discernment and avoid being blinded by short-term stock market fluctuations.