Be wary of Sora Concept Stocks!

Author: Xiao Li Feidao, Editor: Xiao Shimei

Sora was born, swiping the screen around the world, earning the attention of investors.

On the first day of the Year of the Dragon, the concept of A-share AI broke out as scheduled, with more than 100 companies rising by more than 10%. Next, will AI repeat the beautiful story of the first half of 2023?

【Terrible Sora】

On February 15, Open AI officially announced Sora without warning, which caused a thousand waves with one stone.

Open AI calls itself "Sora" as a "simulator of the world".

Musk said bluntly: Human beings admit to gambling and losing.

Former Ali VP Jia Yangqing commented on Sora: It's really good.

Many fans exclaimed: the real world does not exist anymore.

What exactly is Sora?

To put it simply, it's a large model of a text-to-video video. This doesn't seem to be a big deal, long before the launch of Open AI, there were related large model products one after another, and everything was available. For example, PIKA, Runway, SVD, Genmo, Moonvalley.

But Sora is enough to reduce the dimension and beat the competition.

First of all, it increases the video duration from the current mainstream of a few seconds to 1 minute, cutting into the duration range of short video creation. Second, a single video with multiple angles and consistency. In addition, it is possible to generate complex scenes with multiple characters and specific movements, which can understand the physical world to a certain extent.

▲ Screenshot of the 60s video generated by Sora, source: Open AI

It can achieve such a shocking effect because it stands on the shoulders of giants and applies two more mature technologies on the market (Transformer + diffusion model). It is embedded in its own GPT model paradigm, that is, the prompt words entered by the user are first expanded in detail by GPT, and then the expanded content is handed over to Sora to generate more accurate videos frame by frame. This gives it a unique advantage.

Sora's competitors are stuck in Wensheng Wen and Wensheng diagrams because they do not have the underlying expansion processing model of GPT, and there is an insurmountable gap with Sora.

The birth of Sora marks a milestone in the development of Sora. In the future, it will profoundly change the field of content production such as short videos, games, entertainment film and television, animation, advertising and marketing, news and social networking.

For example, we see a video of a brand in just a few minutes on TV or the Internet, and the traditional advertising agency offers millions of dollars. The latter also requires significant cost and time to achieve. In the future, the use of Sora may significantly reduce the spending of its brands. For advertisers, it may mean a reshuffling of business models.

Therefore, the outside world has a high praise for Sora.

Even so, the striking Sora has its flaws. Although Sora's understanding of the real world has improved to a considerable extent, its essence is still based on training and computing based on massive amounts of data, rather than its own understanding of the world, and it is far from reaching the imagination of the human brain.

As a result, the videos generated by Sora have many limitations, including even factual errors, and the inability to understand the laws of cause and effect that require imagination. For example, a person bites a biscuit without a bite mark, a person runs in the wrong direction on a treadmill, and a glass is only deformed when it falls to the ground.

In recent days, some fans have exclaimed that Sora has become a "universal world model" by beating Runway and Pika. Obviously, this is too mythical Sora, which is still a long, long way from AGI (Artificial General Intelligence) and the world model.

The so-called "real world no longer exists" is by no means the present. No one can deny the vastness of the promise of artificial intelligence, but it is necessary to keep a clear head now.

[AI Sustainability?]

The terrible Sora detonated the A-share AI concept, and the highlight of the main line returned. But what about sustainability?

Taking history as a mirror, we might as well review the previous AI market.

At the beginning of 2023, ChatGPT swept the world, kicking off the artificial intelligence revolution. In this regard, the U.S. capital market is not stingy and has given a high valuation premium to the real beneficiary leaders. For example, NVIDIA's PE is as high as 95 times.

After the explosion of ChatGPT, the A-share artificial intelligence sector has a sense of déjà vu that implies U.S. stocks. U.S. stocks are soaring with AI, and A-share related AI concept stocks are also ready to move.

Especially in the first half of 2023, the A-share AI chicken dog will ascend to heaven. In just a few months, 86 related concept stocks have doubled. Among them, Hongbo shares rose by more than 500%, and Zhongke Information and Wanxing Technology rose by more than 400%. You must know that this is still achieved in the context of the continued sluggishness of the A-share market.

From the second half of 2023 to the present, the AI sector has been beaten back to its original shape, not only completely giving up the previous gains, but also much lower than the position before ChatGPT exploded. From this perspective, the AI market of A-shares is more of a storytelling and speculation concept.

The Cambrian period is a prime example. Since 2017 was financial, Cambrian has not been profitable. In the past four years, the cumulative loss has been 4.5 billion yuan, and its revenue in the same period was only 2.499 billion yuan, and the growth was extremely slow. In the first three quarters of 2023, revenue plummeted by 44.8%.

Poor fundamentals, Cambrian soon took shape, and the stock price ushered in a sharp decline, with a maximum decline of more than 60%. However, on the first day of the Year of the Dragon, Sora rose 16% again with the news, which was very eye-catching.

For another example, Kunlun Wanwei, which rose 17% on February 19, will be widely questioned as a "fake divorce and real reduction" in 2023 because of the cash-out operation of Li Qiong, the ex-wife of founder Zhou Yahui. After the company landed on the GEM in 2015, it extended more tentacles to the acquisition of assets and the expansion of overseas business, trying to tell the market a capital story of "game + platform + investment", but the stock price has always been tepid.

In 2023, due to the spring breeze of AI, the stock price will soar. However, the fundamentals behind it are weak, and profits have continued to decline sharply in the past three years. Later, the stock price plummeted by nearly 60%, trapping a large number of investors.

From February 19th to 20th, Chinese Online relied on its business to touch Sora, and there was a continuous 20CM up and down. The financial performance behind it is also unbearable to look at directly.

Therefore, the AI market detonated by Sora is more of a concept hype than last time, and its overall sustainability is questionable. Moreover, Sora is an important upgrade of ChatGPT, which does not have a greater impact on the market when ChatGPT was popular, and the persistence of the concept should not be too long, and it is more based on the news to carry out a wave of low valuation repair.

After all, in recent months, AI has been killed too hard, from 72 times to 40 times. Some core faucets PE even fell below 20 times.

▲人工智能PE走势图,来源:Chocie

Of course, the explosion of Sora will still help some excellent companies in A-shares to run out, and there are many investment opportunities.

["Selling shovels" is the most beneficial]

The advent of Sora indicates that the competition of large models has risen from text and pictures to video, and the amount of AI training has increased significantly, and the demand for computing power has increased exponentially. Therefore, the computing power company that "sells shovels" should be the first beneficiary.

Previously, some institutions predicted that the total computing power of global computers will increase from 615EFlops to 56ZFlops in 2030, an increase of nearly 1,000 times, with a compound annual growth rate of up to 65%.

This estimate may not be radical. A few days before Sora's launch, Open AI CEO Altman was rumored to be in talks with investors, including the UAE government, to raise up to $7 trillion to build a global supply of semiconductor chips.

What is the concept of $7 trillion? Altman can take Nvidia, AMD, TSMC, Broadcom, ASML, Meta, Samsung, Intel, Qualcomm and other companies in a package.

I don't know if the rumors are true or not, but from the side, it can reflect that AI computing power chips should have immeasurable demand in the future.

Of course, the domestic computing industry chain will also usher in a high degree of prosperity. According to the statistics of Southern Fortune Network, there are 52 A-share computing power-related enterprises, with a revenue scale of 1,043.58 billion yuan in 2022, a year-on-year increase of 8.5%, maintaining growth for five consecutive years. The net profit attributable to the parent company was 43.064 billion yuan, a year-on-year decrease of 2.6%, but an increase of more than 70% over 2018.

▲Source: Southern Fortune Network

This happened before ChatGPT exploded. In addition, Sora's excellent generative capabilities this time may trigger a boom in domestic Wensheng video AI large models, and it is expected that relevant domestic manufacturers will also quickly follow suit, thereby driving the demand for computing power to increase.

The AI computing power race is inevitable, especially the demand for optical components. In 2024, global AI investment will spread from several cloud giants in North America to Tier2 clouds, large enterprises and intelligent computing centers, driving the recovery of global optical modules.

In the A-share market, Zhongji InnoLight is the world's leading optical module and one of the companies that can make real achievements in the computing power track, with a net profit of 120 times in 6 years. The market share of high-speed optical modules ranks first, and it focuses on R&D investment, far exceeding its peers such as Xinyisheng and Accelink. Relying on the core competitiveness of products, it has won major customers such as Google, Amazon, Huawei, and ZTE, and its overseas revenue has remained at about 70%.

On February 19, market rumors said that after Nvidia tested 800G silicon photonics, it said that the effect of InnoLight was crushing and good, and placed a small batch order.

▲ Zhongji InnoLight net profit trend chart, source: Choice

In the first half of 2023, Zhongji InnoLight's share price rose by more than 500%. After that, the industry retraced to digest the valuation, but far outperformed the average performance of the industry. Since then, the stock price has rebounded by more than 80% since it bottomed out in November (up 20% on February 19), and is only one step away from a new all-time high.

The AI wheel rolls forward. Embracing a promising industry is not guaranteed to be the best choice, but it is certainly not the worst choice. However, from the perspective of the capital market, there are too few companies like Zhongji InnoLight that can show solid performance, and the vast majority of A-share AI companies are speculating on concepts, and the stock price will eventually return to dust and dust.

disclaimer

The content of this article related to listed companies is the author's personal analysis and judgment based on the information publicly disclosed by listed companies in accordance with their legal obligations (including but not limited to temporary announcements, periodic reports and official interactive platforms, etc.), and the information or opinions in this article do not constitute any investment or other business advice, and Market Value Watch does not assume any responsibility for any actions arising from the adoption of this article.

——END——