Image source: Generated by Unbounded AI

Large model companies are increasingly disclosing their financing information with half-hidden views.

Two recent financing news come from Zhipu and Leap. Although Zhipu announced the 3 billion yuan D round of financing, it has never publicly disclosed the specific list of investors. It only mentioned in media reports: According to multiple insiders, the new investors include a number of strategic investment and state-owned institutions, while old shareholders such as Legend Capital continue to follow in the investment.

This financing of the step is the only one disclosed in 2024, but the specific financing amount has not been announced. The public information shows "Series B financing of hundreds of millions of dollars". At the same time, It has not announced its valuation, making it the only basic large-model company among the current "Six Small Companies" whose valuation is unknown.

To disclose, but not fully disclose, large model companies have begun to use various methods to complete "pipa-style" financing announcements. Behind this is the current financing situation of large model companies:

On the one hand, they are in urgent need of continued financing to support expensive investments; on the other hand, they hope to retain some "mystery" and avoid Revealing the truth too early or too much.

"After the withdrawal of the US dollar funds, there is only so much money in the domestic market. It is basically a fixed pool of funds. If one company gets more, the other companies will get less. Squeezing the opponent's survival Only with space can we survive to the end. "Some investors told us that in 2025, China's large models will start a brutal battle. This battle for China's large models has become a survivor game, and when will it be revealed? Financing information and how to disclose it have become important means of competition.

1. After 8 months, enter the survivor game

Let Let’s quickly review several important nodes in the development of China’s basic large-scale model in 2024.

At the beginning of the year, after OpenAI released Sora, domestic large model companies began to follow up; in May, Volcano Engine took the lead in starting a "price war" in the large model industry; in June, Kuaishou released Keling, and then MiniMax, Zhipu, and Doubao have gradually launched video generation products; at the same time, foreign star unicorn companies Stability, Inflection AI, and Character AI began to suffer setbacks due to various reasons; in August, fed by several rounds of high-speed and huge financing, China's leading model companies entered the super unicorn club with a valuation of 20 billion almost at the same time; but followed closely However, due to OpenAI's delay in releasing Sora and new products, domestic large model technology progress has also stopped. The "Six Little Tigers" have begun to be questioned whether they have pre-trained large models, and there are also rumors that some companies have given up on pre-training.

Until September, OpenAI released o1 again, has extended the life of the craze for large models, but it also proves that it is still difficult for domestic prototype companies to talk about innovation independently.

The turning point may occur in August. On the one hand, when the valuation reached 20 billion, it posed a challenge to the business model; on the other hand, the market from February to August has given enough patience to large model companies in the absence of OpenAI, but it still No satisfactory results were obtained.

I was still confident in the first half of the year, but in the second half of the year, I began to have doubts.

A senior investor told Silicon Star: “When market risk appetite is high, capital pours in intensively, and companies disclose more financing information to shape ‘industry popularity.’ But when market risk appetite declines, Companies may intentionally reduce or delay disclosure to avoid exposing the market to unfavorable cash flow conditions or to weaken the credibility of competitors. "

This is not an individual problem of a certain company, but a "characteristic" of the entire large-scale model industry, which is full of serious problems of "technical investment first" and "commercial returns last" Misalignment, huge technology research and development and computing hardware expenses require continuous infusion of cash, but it is difficult to obtain substantial returns on the user or customer side in the short term. Enterprises can only rely on continuous financing "blood transfusion" to survive. Whether you can continue to obtain financing has become the basis of life and death.

“In fact, semiconductor, materials, and energy industries also require ten or twenty years of investment patience. For example, in industries like hydrogen energy, although the time is also very long, we know that the first step is to The state heavily subsidizes; the second step is to achieve market parity between hydrogen energy and fuel; and the third step is to find application scenarios on a large scale. In comparison, the large model seems very vague and it is easy to have no technological breakthrough in a short period of time. Loss of market confidence," another investor said.

So, the big model companies quickly entered a cruel survivor game. The above-mentioned investor said bluntly: "In this game, who can survive to the end often does not rely on technical strength, but on who can always get money."

This forms a certain The "chicken-and-egg" cycle is that in the early stages of technology, since there are no business indicators to test, the scale of financing is regarded as the only indicator of "potential", and the status of "capital darling" becomes proof and endorsement of competitiveness. .

2. "Pipa-style" financing has become a means of competition

< p> In addition, the first reason for continuous financing is to ensure a "long enough life" to survive the next round of technological explosion; the second reason is to absorb the existing funds in the market to squeeze the survival space of competitors.

However, as valuations and financing amounts become larger and larger, it becomes more difficult for startups to raise funds and the range of options available becomes smaller and smaller. On the one hand, the ownership structure of large model companies that have just been established will become extremely complicated, and conflicts of interest and even disputes will begin to arise. On the other hand, many timesAt this time, these financings do not have clear round boundaries and corresponding technical or business milestones as in the past.

Therefore, large model companies often fall into a contradiction when disclosing their financing progress: they need to build confidence in development to attract the next round of investment; but they cannot disclose too much information to become a "target" for all parties. ", thus forming such a situation of "still holding on to the pipa".

Moreover, in the large-model track where funds are crucial, the financing progress of enterprises is no longer a simple financial matter, but has become an important marketing node.

The disclosure of financing information and the control of timing will directly affect whether the company can continue to gain favor from capital and influence competitors' judgment of their own strength.

As a result, we are increasingly witnessing the difficult game between large model companies between "integrating" and "hiding". Financing went from a real thing to an on-demand announcement.

Many investors in the industry revealed that at present, after confirming financing, large model companies will basically choose to disclose it to the outside world after half a year, and some will even wait until full delivery or half of delivery. What's more, they have to disclose financing news only when they have almost run out of cash and are strapped for funds, in order to extend their lives and attract the next wave of financing.

For companies, the delay in disclosing financing may also be due to the fact that the current enthusiasm for capital has cooled down, and they urgently need to rely on the financing results of the previous period to regain the confidence and recognition of investors.

A senior investor put it bluntly: "The money you can get, and whose money you get, has almost become the most important weight in the next round of financing. As for technological progress and commercialization, progress has slowed down significantly. , no one can call it a breakthrough.”

This is becoming a common challenge in the large model industry, and we are likely to continue to see more “pipa-style” financing.