CITIC Securities: Why is Sora so popular?
The artificial intelligence industry is currently in a rapid development stage, with continuous technological advancements expected by 2024. NVIDIA's launch of Gear further emphasizes embodied intelligence, and future investment opportunities will gradually shift from hardware and tools to software products. CITIC Securities recommends continued focus on AI-related tech companies in the US stock market. The release of Sora indicates that the American AI industry has entered the stage of mental dominance and topic marketing after reaching a certain level of technological and product accumulation. Sora's video "Snowy Tokyo" has already reached 92.5 million views, demonstrating high duration, high quality, and high relevance to user preferences.
Zhitong App has learned that CITIC SEC has released a research report stating that the artificial intelligence industry is currently in a high-speed development stage. The capabilities of large basic models and their applications are continuously iterating rapidly. Although there is still some distance between the current industrialization stage of the industry and the explosion of consumer products, it is expected that there will be endless technological advancements by 2024. Recently, NVIDIA launched Gear, further pointing towards embodied intelligence. It is predicted that as the AI industry gradually matures, future investment opportunities will shift from hardware and tools to software products. The industry's development is showing a trend of pure software applications transitioning to a combination of software and hardware, connecting the virtual world with the real world. In the future of artificial intelligence, whether it is the current leader OpenAI, or the continuous followers like Google Gemini, Facebook Llama, and Chinese AI industry companies, they are all expected to continue to follow up. It is recommended to continue to pay attention to AI-related US technology companies.
▍Key points from CITIC SEC:
Sora: Surprisingly stunning performance "Three Highs" has attracted widespread attention in the global capital market and even the Internet world.
According to the technical paper of OpenAI, Sora is built on a solid foundation of OpenAI's visual understanding and the emergence of large models. Compared to diffusion models like Pika in video games, Sora replaces the backbone U-Net convolutional network with a more scalable Transformer, allowing the model's performance to be improved by increasing the parameter size and training data volume. In addition, during the video generation process, Sora may also introduce GPT to improve the workflow. The performance improvement driven by architectural innovation and massive computing power investment enables Sora to generate videos with high duration, high quality, and high instruction fit, leading the market to believe that Sora's capabilities exceed expectations.
At the same time, the AI-generated video lasting 1 minute is not only visually impactful but can also achieve viral spread in the global Internet world through social platforms and short video platforms. As of February 26, 2024, the Sora-generated video "Snowy Tokyo" released by OpenAI's official Twitter account has reached 92.5 million views.
Sora's breakthrough and timing of release indicate that the US artificial intelligence industry has entered the stage of mental dominance and topic marketing after accumulating to a certain level of technology and product.
A few hours before the release of Sora, Google released Gemini 1.5, which extends the context length to 1 million tokens; Meta also released the unsupervised video prediction model V-JEPA. All three are the latest technological achievements, but there is a huge difference in popularity. According to Google Trends, on February 16, the global search popularity of Gemini 1.5 was only 3% of Sora, and V-JEPA was even less than 1% of Sora. A week after the release of Sora, it has not yet officially launched, but a few hours after the release of Gemini 1.5, it quickly released and started topic marketing. OpenAI's timing of the release on Sora, along with its $7 trillion financing plan and the Google Clips demonstration video in December 2023, indicate that both first and second-tier companies in the United States have made significant technological and product advancements. However, they are still on the eve of an industry and consumer application explosion, requiring substantial resources such as funding, talent, data, and user base. The competition for resources ultimately propels the American artificial intelligence industry into the stage of mental dominance and topic marketing.
Why is Sora so popular in the Chinese internet world: The Spring Festival traffic gap + OpenAI's mental dominance over users.
Sora was launched on the sixth day of the Chinese New Year holiday, during a period of low traffic after the Spring Festival. Since its release, the popularity of the Spring Festival Gala and New Year films has been declining, creating ample space for this topic. According to Baidu search index, the day after Sora's release, it surpassed the popularity of Honor of Kings, and by the third day, it exceeded the popularity of "Pegasus" 2. A week later, its popularity was close to "Hot Summer".
The high popularity of Sora in the Chinese internet world is supported by the nation's high attention to AI and the stunning effects of Sora-generated videos, as well as OpenAI's mental dominance over users. Before and after the 2023 Spring Festival, ChatGPT's education of Chinese users has led many to view Sora and OpenAI as successful filters, believing that Sora is the GPT-3 moment of text-to-video, with the commercialization of text-to-video products on the horizon. However, considering actual technological progress, costs, and commercialization, it will still take time for the text-to-video industry to launch truly popular products.
OpenAI after the release of Sora: Under the spotlight, competing with giants, leading in technology, facing ongoing challenges.
OpenAI is leading the current wave of generative artificial intelligence. The release of products and models such as ChatGPT, GPT-4, and Sora has put the company in the spotlight, with its technological advancements and product releases continuously imitated by giants like Google. Comparing OpenAI and giants like Google in terms of personnel, product lines, and revenue, it is evident that as a startup, OpenAI faces significant competitive pressure due to its heavy resource consumption in parallel technological research and development paths and its lack of self-sustaining ability.
At the same time, its unique organizational structure also exposes the company to higher operational risks. To maintain its leading position, OpenAI's need for financing is urgent. As of February 2024, OpenAI has raised over $13 billion, and its latest $7 trillion financing plan reflects the current stage of the American artificial intelligence industry reaching mental dominance and topic marketing. The subsequent actions of OpenAI in financing, being both technologically advanced and trending, are worth paying attention to.
Risk factors:
Risks of AI core technology development falling short of expectations; Continued tightening of regulatory policies in the technology sector; Risks related to policy regulations on private data; Global macroeconomic recovery falling short of expectations; Potential ethical, moral, and user privacy risks of AI; Risks of corporate data leaks and information security. Global cloud computing market development falling short of expectations poses risks; intensifying industry competition also poses risks.