Source: Yaya Hong Kong Stock Circle

This year's champion in AI stocks has recently changed.$Palantir (PLTR.US)$and$Applovin (APP.US)$It has been surpassed, and the current top gainer is an AI voice model company. $SoundHound AI (SOUN.US)$ This company recently reached an all-time high, and this year's increase has reached ten times, rising to the top position in the annual growth rankings for AI application stocks in the USA.

What this company does is not very new and is quite common in the eyes of investors in China - the AI model for converting speech to text. The current explosive growth in SOUN's performance mainly comes from starting to build customer service systems to assist Restaurants in taking orders, and entering Software interactive applications in the Autos and Financial Industries.

Similar voice control AI is everywhere, $Apple (AAPL.US)$ SIRI from $XIAOMI-W (01810.HK)$ Xiao Ai from $TENCENT (00700.HK)$ voice-to-text feature in WeChat.$BYD Company Limited (002594.SZ)$/ $BYD COMPANY (01211.HK)$ Hello, Xiao Di. Many Chinese Technology products have similar functions. Behind this,$Iflytek Co.,ltd. (002230.SZ)$it is a company that has long focused on the voice-to-text track.

It can only be said that this thing has a relatively low penetration rate overseas, but the market is huge. Currently, SOUN is regarded as an American version of Iflytek Co.,ltd. Under the fervent hype of AI applications and with the company's explosive performance, it has undergone a complete revaluation, with a year-on-year growth rate of 70%. However, the valuation has risen from 10 times PS to 100 times.

But what kind of imagination lies behind the 100 times PS?

1. The leap achieved through CHATGPT.

Looking at the product model, SOUN focuses on speech-to-text conversion, then relies on GPT large model to process information and achieve automation. Nowadays, using large language models requires typing, and to communicate using speech, a speech-to-text plugin is needed. However, the research and development iteration of plugins is currently neglected by most large model companies, as they concentrate on the development of multimodal conversion for images and videos in Agi. Therefore, small companies specializing in speech-to-text have found room to survive.

This plugin cannot be said to lack technical content; if the conversion quality is poor, interacting with the large model through speech becomes very inconvenient. The faster the conversational interaction, the better.

However, compared to large models, the technical difficulty is too low. Many Chinese companies can replicate it after casually developing for a few months because every language, letters, words, and pronunciations are limited. There are only about 3,000 Chinese characters and approximately 0.4-0.6 million English words, so the database for speech processing models does not need to be very large. Speech-to-text conversion is very simple; the technical challenge lies in understanding and controlling the language.

For example, Iflytek Co.,ltd. is a leading company in this field in China. The small models for speech-to-text are well-developed, but without the GPT era, their self-developed language large models are inadequate, leading to shallow business coverage. Even if their business expands into various industries, there are no good commercial returns, and ultimately, they still need to resort to making money with learning machines and translation machines.

Simply put, people feel that Siri and Xiao Ai understand too little and respond too clumsily. But now that CHATGPT has emerged, the brains of Siri and Xiao Ai have been replaced with the latest large models, greatly enhancing their processing and control capabilities, increasing usability. This is likely the foundation for SOUN's performance starting to grow.

At the performance press conference, SOUN mentioned that the company cooperates with GPT, and information processing still relies on external technology. From this, it can be seen that the company's so-called moat, simultaneous speech conversion and natural language understanding, essence is still benefiting from the improvement of GPT technology. If the GPT technology had not improved to its current level, SOUN's exclusive technology would likely just be empty talk. From many product examples, SOUN's speech-to-text technology does not have superior technical content compared to Iflytek Co.,ltd.

Of course, relying on GPT for growth is not shameful; many so-called AI application stocks are essentially utilizing GPT well, penetrating into various industries for specific applications, as long as GPT does not take over everything and snatch away everyone's jobs.

Previously, the global penetration rate of voice AI technology was very low, with China using it relatively more, which created a national gap. A simple example is:$Tesla (TSLA.US)$The voice recognition control capability of the car's system is extremely poor. Basic voice operations like closing windows and asking about the weather are not feasible, while domestic automakers can control most vehicle functions. For example, BYD can now generate AI images using voice commands in the car system. In terms of voice control, China significantly outperforms the rest of the world, while Tesla is already considered relatively advanced among overseas tech companies.

If we look at other global automakers like Toyota and Volkswagen, their capabilities lag far behind Tesla. This is why BYD's overseas users find its voice control function amazing, while Chinese consumers think it's strange that they haven't experienced these capabilities.

Conversely, this indicates that there is still a large development space for such businesses overseas. Voice control in cars, voice control in home appliances, ordering food, customer service, and automated call-backs—all of these areas contain significant iteration potential for many companies in Europe and the USA. Furthermore, the B-end payment capacity is much stronger than that of Chinese companies, which provides SOUN with considerable market imagination, leading to high valuations.

In summary, the previous low level of language models led to a narrow application of speech-to-text technology, which was merely seen as a flashy feature, causing B-end users to be unwilling to invest. With the development of large models today, the performance of these tools has greatly improved, prompting various industries to start taking notice and being willing to spend money on applying this technology to their products. From relatively simple and single-function automotive control to food ordering, customer service, and voice interaction, AI is flourishing across various sectors.

In the field of speech-to-text, Europe and America have not been as competitive as China, but the market is larger than that of China. The earliest tech companies in China to implement these solutions, even with more mature options, may find it difficult to share in this vast overseas market.

The reason lies in the sanctions and exclusion amid US-China confrontations—such as OpenAI being unable to supply to Chinese companies, or even using large models like Doubao for processing, as the USA could claim that they would extract sound data from speech, impacting National Security. Thus, the immense competitive forces in the industry have been eliminated, while the market remains substantial, providing SOUN with significant imaginative space.

SOUN has also described its market target as 140 billion dollars. Looking at the current Market Cap, there seems to be a lot of room for growth, major imagination, and a whole bunch of strong competitors unable to participate in the competition, which is a key reason for its high valuation.

2. Concerns about the next Chegg?

However, excluding factors like political competition, a PS of 100 times is still too high. The real concern for SOUN does not come from China, but from the cross-industry ventures of large model companies.

Previously, the development of GPT has already caused many small model companies to exit the historical stage, such as the listed company $Chegg (CHGG.US)$ , an online platform providing student chatbots, where the large model constructed by GPT inadvertently completely replaced this functionality. How can AI models made by small companies compare with the high versatility trained by Internet giants in large quantities? There are plenty of replaced AI startups, such as Grammarly, which corrects grammatical errors.

Many AI applications are still on the brink of life and death. Whether they can survive entirely depends on whether the large models are willing to give them a lifeline.

Of course, GPT is not intentionally trying to eliminate them; it just inadvertently replaces these intricacies on the road to AGI. If one day, OPENAI, Google, and Meta decide not to compete in images and videos but rather in voice and text conversion, then obviously, SOUN's survival space would become zero. However, SOUN can seek to be acquired by large companies. But this expectation of acquisition cannot be overly relied upon; supporting a high valuation based on the expectation of being acquired is also unreasonable. Because once there is no interest, the greater the expectation, the greater the disappointment.

Therefore, the reliance on external suppliers for core technology remains the biggest flaw of this model company. In fact, one can observe the trend: in 2023, with the advent of CHATGPT and the hype around AI Software and Hardware, SOUN's stock price has been falling continuously in the second half of 2023, almost returning to its historic low. At that time, people's views were similar to Chegg's, as they felt they could be easily replaced. Meanwhile, SOUN's valuation was not high, and its performance was still growing, indicating that the decline was clearly not due to poor performance.

The decline at that time was definitely incorrect, but today's rise definitely has problems.

3. The truth behind the performance.

Returning to the fundamentals of SOUN, there are still some obvious flaws.

The reason for the stock price rising by 10 times is partly due to the company's goal of a 140 billion target market. Once the guidance for doubling next year's performance came out, the 140 billion figure was instantly believed by everyone. Additionally, the market is currently fervently pursuing AI application Stocks, pushing valuations to historical highs. Moreover, let’s not forget the significant drop last year due to technological replacements, which has also provided ample rebound potential for the company's stock price.

However, upon seeing the disclosure of this year's latest earnings reports, although performance has significantly increased, it is still driven by the acquisition of Amelia. Amelia contributed a quarterly performance increment of 10 million. Excluding this factor, the organic growth rate of performance is not high, remaining in single digits; the truly remarkable aspect is the revenue growth expectation of doubling next year, which seems to indicate that the company has secured some Orders for next year, providing a bit of confidence.

Looking at the profit margin, the core of the company's products is based on GPT. Without the advancement of GPT technology, it cannot be implemented to expand revenue. This creates a contradiction; GPT also has to make money. Since the beginning of this year, large models have started their monetization journey. Large models are not charities and will not be free forever. It is inevitable that prices will continue to rise. Before OpenAI achieves good profit margins, the pace of price increases is unlikely to slow down.

As a purchaser of GPT, the issue SOUN faces is that future costs will continue to rise. We see that the company's gross margin for Q3 2024 has significantly dropped compared to last year, even when excluding the acquisition factor.

A manufacturer that does not master core technology does not have the foundation for high profit margins.$NVIDIA (NVDA.US)$The more you earn, the less favorable it is for PC assembly companies and datacenters.

The current status of SOUN is that the market is only focused on revenue growth, overlooking the issue of declining profit margins. For example, Applovin has shown strong performance while its profit margin increased from 10% to 30%. Technology companies with declining profit margins should theoretically have their valuations lowered rather than increased, which is the biggest problem with SOUN's valuation currently.

The market seems to have fallen into an 'AI application asset shortage', where companies with AI software applications receive unlimited funding for purchases, while ignoring the fact that there are many good symbols among non-listed companies, which are still at low prices.

SOUN also has competitors overseas, such as Rev.com, which is currently the number one speech-to-text company, with annual revenue reaching 0.8 billion USD, nearly 10 times that of SOUN. However, this company is not publicly listed, while the largest publicly listed speech-to-text technology company is SOUN.$Microsoft (MSFT.US)$、$Amazon (AMZN.US)$There is also a speech-to-text tool department under its company, but generally this business is too small, and it does not affect the stock valuation.

Now the question arises, if Rev.com goes public, what will its valuation be? Does the market think that there is only one speech-to-text company, SOUN, and that it will own the entire 140 billion in the future?

It is not difficult to see the phenomenon of "AI application asset scarcity"; even the scraps of the industry have received outrageous pricing. If all companies went public, they could possibly have a market cap of billions, but the problem is, can such a small technology really have that much market cap space?

The larger possibility is to replicate the trend of Zoom, which during the pandemic had an "SaaS asset panic," causing a video conferencing software with low paid space, low customer stickiness, and not high technical content to be boosted into one of the top ten software companies by market cap globally, but ultimately it had to return to reality.

Conclusion

The rise of SOUN seems to have a certain degree of randomness, and compared to the APP, it is not that convincing. The core technology is not exclusive; with the price increase by OpenAI, the path to profit is opening up. The problem this company will face in the future is that while revenues may rise, profit margins are decreasing, and it may never be able to achieve profitability.

Gross margin is expected to fall from 70% to 60%, or even lower in the future. Theoretical price-to-sales ratio should drop from 10 times to 5 times or even lower, but today the market sees the target market figures and doubled performance expectations, giving a 60 times price-to-sales valuation based on 0.15 billion revenue in 2025.

Especially since the company's revenue is less than 0.1 billion, a substantial order can completely change the growth rate for several years; a 100% increase in one year does not mean a 100% increase every year, and the future remains full of uncertainty. The performance of this company's stock price vividly reflects the current "AI application assets scarcity."

The AI application boom is upon us! Make good use of the "Investment Themes" feature, open Futubull > US Stocks > Investment Themes >AI application software stocks, and seize investment opportunities.

编辑/jayden