news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Author: Yoky
Email: [email protected]

Dark Side of the Moon has raised funds once again.

On August 6, Bloomberg reported that Dark Side of the Moon had just completed a new round of financing of more than 300 million US dollars, pushing its post-investment valuation to 3.3 billion US dollars (about 23.6 billion RMB), making it the unicorn company with the highest valuation among the "Four Little Dragons" of the big model.

Interestingly, on July 25, Baichuan Intelligence completed a RMB 5 billion A2 round of financing, with a post-investment valuation of RMB 20 billion; on May 31, Zhipu AI completed a US$400 million B+ round of financing, with a post-investment valuation of US$3 billion, about RMB 21 billion.

If we compare these three sets of information horizontally, we will find that the top three large-scale model unicorns in China have joined the "20 billion" club within two months, and only MiniMax is left with the final push. After strategic financing, its valuation has reached US$2.5 billion, or approximately RMB 17.5 billion.

What does a 20 billion yuan valuation mean? Compared with the "Four Little Dragons of AI" in the previous era, SenseTime, which has the highest valuation, has a market value of about 34.4 billion yuan, Megvii Technology has a market value of about 26.7 billion yuan, Fourth Paradigm has a market value of 18.8 billion yuan, and CloudWalk Technology has a market value of only 10.4 billion yuan, less than half of the Dark Side of the Moon.

According to an insider, Baichuan Intelligence had completed the settlement of its previous financing around November last year, but it was not disclosed until July. Dark Side of the Moon's latest financing had also been completed before that, but it chose to disclose it at this point in time to follow closely and not fall behind.

An investor analyzed: "It only shows that half a year ago, everyone still had expectations for the large-scale model of the base. The key is whether it can raise money in the next round. The recent disclosure is probably because there is no major breakthrough in technology and products, and the only way to gain market confidence is to rely on financing news."

Overseas unicorns that could not raise funds have collapsed one after another in the first half of 2024. On August 3, Google officially announced the "acquisition" of Character AI (hereinafter referred to as C.AI) and incorporated the team. Adept and Inflection AI were sold to Amazon and Microsoft respectively not long ago.

Coincidentally, their valuations are also in the range of $2.5 billion to $4 billion.

Under high valuations, startups can no longer rely solely on dreams. If they cannot find a visible profit model, investors will lose patience and the founding team will easily pack up and leave.

Behind the “valuation of 20 billion yuan” is not a milestone for the startup company, but rather a danger zone for the startup company.

1

“20 billion is just a number”

More than one investor told us that a valuation of 20 billion yuan actually means nothing. "In the past, there was no such a high bubble in the Internet era. A valuation of 10 billion to 20 billion yuan could reflect the actual commercial value of the company, but now it is just a number."

"Now these four companies may have reached 20 billion, and it is possible that one of them will return to zero next year or the year after, so it is meaningless."

It is not a coincidence that the collective valuation of 20 billion is the result of the "rolling" of startups. "In the same echelon of companies, when one reaches a new height, all will follow and cannot fall behind." Some investors have analyzed that the valuation is not so important now, the important thing is not to fall behind and to keep up with the competition.

"Generally, the valuation of startups like this is done by selecting companies in the same industry or with similar business models as the target company as a reference, and comparing their financial data and market performance to make an estimate. In other words, if only one company reaches a certain value, it will be used as a reference, and the valuation will quickly be brought into line," said another investor.

An investor from Source Code Capital also explained the investment logic behind it: "In the early stages, investors hardly look at the technology, products and business models of large-scale model companies. Assuming that it will take 10 billion yuan to get results, investors are thinking about who can raise 10 billion yuan. Only the company that survives to the end has a chance." Especially under the premise that large-scale model projects are scarce today, the more funds raised, the more financing they can get, and the higher the valuation will be.

This is nothing new, especially in the early "money-burning" fields, where robotics, autonomous driving, chips, and energy companies have all grown up with a lot of capital. But the difference is that the speed at which large model companies reach this threshold is almost "rolled" to the extreme.

The "Four AI Dragons" that were once highly sought after by capital took nearly three years to reach their "20 billion" threshold. Amid the ups and downs of technology in autonomous driving companies, it took Pony.ai and WeRide nearly four years to reach the above valuation level.

In contrast, for the big model companies, Dark Side of the Moon took 350 days, and Baichuan Intelligence only took 300 days, and both rounds were very early. Baichuan and Dark Side of the Moon had just reached the A+ round, and they became super unicorns when their business models had not even emerged.

Silicon Stars learned from several investors that there are several key nodes in the development of startups: when the valuation reaches US$1 billion, it means that it has reached the entry threshold for unicorn companies, and the startups begin to emerge and turn into a seller's market; when the valuation reaches US$3-5 billion, it means that a milestone breakthrough has been achieved in the development process, and the valuation reflects the expectations for future development; once the valuation exceeds this level, the target company is required to demonstrate a viable and efficient business model, and continuously verify and optimize its business model in practice to achieve sustained profitability and growth.

Simply put, before this, startups could survive on investment, but after this, all-round challenges were posed to the founding team, business model, and ecosystem.

Therefore, "20 billion" is often seen as a watershed for a startup company, either taking off or falling.

1

The “Danger Zone” of Mid-Value Valuations

After entering the 20 billion yuan queue, startups begin to face the test of commercialization: capital begins to expect returns, corporate competition is on the table, the growth rate of user scale slows down, retention becomes a test, and they fall into comprehensive considerations of technology, cost, and market.

Startups in this valuation range have entered the most difficult danger zone. Not making money, or having no possibility of making money, has become a sin.

C.AI's valuation eventually shrank from $4 billion to $2.5 billion (17.5 billion yuan) and was acquired by Google, becoming stuck at the 20 billion yuan mark.

Although it was once the most popular mobile app with 4.2 million users and 16 million AI bots created, the huge user scale did not bring a visible business model. Among the 6 million monthly active users, only 100,000 were paying users, accounting for only 1.6%. Instead, it brought high computing power costs. In April this year, C.AI was exposed to cash flow problems. The extremely low payment rate made profitability a pipe dream.

Taking Character.AI as an example, almost all domestic large-scale model products are free for C-end users. Even though Dark Side of the Moon has firmly chosen the to C product form and is actively exploring the user payment model, the backlash from users just proves the difficulty of to C commercialization.

However, startups are not good at providing services in the to B sector. The above investor shared: "Many cloud vendors offer large model services for free. Startups have a weak advantage in the to B sector, so integration is a good choice."

“A company with a valuation of 20 billion must generate at least 200 million in profits in the next 2-3 years, but now almost no large-scale startup has a visible profit model.”

Another investor also said: "Companies at this stage are essentially returning to business. The current valuation and business potential are seriously mismatched, and there is even no hope. If C.AI DAU (daily active users) increases from 4 million to 50 million, even if it is not profitable now, according to past experience, it is easy to make 100 million or 200 million with 50 million DAU. But it stops at 4 million, which means your ceiling is so high, and it will be the same no matter how much money you burn."

If a company does not make money, investors will lose patience. In the process of constantly pushing startups to commercialize, the actions of startups will gradually change, which may lead to companies sacrificing long-term development in pursuit of short-term performance. Compared with the founders' initial goals, investors are more willing to see the company's valuation continue to expand in order to increase profit margins.

According to Silicon Star’s understanding of several large model companies and startups in China, some startups have been “forced” to charge users, develop new business models, or push products overseas, even if the founders themselves do not think now is the best time.

At the same time, some management problems of the entrepreneurial team will also erupt at this stage, just like the management of Stability was full of dissatisfaction with the founder and eventually got rid of his "notorious" CEO, but in the final analysis, it was because of the lack of profit, and all the problems that could be whitewashed were brought to the table.

Faced with internal and external troubles, startups are faced with a new "difficult copy" in the midst of such a tug-of-war.

1

Who can continue to vote?

Then the biggest question is, how will the company, which has a high valuation, continue to raise funds?

In Dark Side of the Moon’s latest financing, another very interesting phenomenon is that Tencent emerged as a new investor, taking over together with old shareholder Alibaba.

In this regard, the above-mentioned investor said: "For large companies with annual profits of tens of billions, this amount of money is not much. It may be to buy a cutting-edge position, to have some exchanges with the academic community, or to have some technical cooperation in some terms, etc. In short, the reasons behind it are very diverse."

But what is certain is that with strong capital injections from the giants, it has become very difficult for startups to carve out their own independent path.

Judging from the current domestic financing situation, two directions have emerged: one is capital injection from local industrial funds, and the other is to accept investment from venture capital funds of large foreign conglomerates such as Saudi Aramco.

In Baichuan's latest A+ round, investors include Shanghai Artificial Intelligence Industry Equity Investment Fund Partnership and Beijing Artificial Intelligence Industry Investment Fund. According to public information, Beijing Artificial Intelligence Industry Investment Fund has already invested in many star startups in the field of large models, including Baichuan Intelligence, Shengshu Technology, Zhipu AI, RealAI, Deepin Technology, and Mianbi Intelligence.

In addition to accepting the Beijing Artificial Intelligence Industry Investment Fund, Zhipu began accepting investments from more than ten local industrial funds including Beijing, Shenzhen, Nanjing, Zaozhuang, Qingdao, Suzhou, and Kunshan as early as 2021.

The above-mentioned investor speculated that if the valuation continues to rise, funds with state-owned backgrounds may come in later. This will further centralize the entire large model industry, intensify polarization, and form a new pattern.

Another direction is to turn the target to the Middle East consortium. In the latest round of financing of Zhipu AI, Prosperity7 Ventures, a venture capital fund under Saudi Aramco, invested $400 million in the B+ round. Prosperity7 Ventures is a diversified venture capital fund under Saudi Aramco and has invested in Chinese AI companies many times, including autonomous driving solution company Hongjing Zhijia and vector database company Zilliz.

"It is normal for autonomous driving companies to introduce international capital in the later stages of financing because the amount of financing is so large. Pony.ai introduced the China-UAE Joint Investment Fund in the D round, and the D+ round was also led by a Saudi-backed venture capital."

As for why Saudi venture capital has frequently invested in China in recent years, an investor with a US dollar fund background explained: "Some RMB investors cannot afford to invest, and some cannot wait. This kind of long-cycle industry requires some old money to have enough patience. It is normal to see returns in ten years."

The rally starts, and ultimately, whether a startup can develop in the long term under the halo of high valuation will still be a test of the startup’s own innovation capabilities, internal management, and strategic determination.

In other words, the competition for China's large models will only truly begin in the second half of 2024.