Aus 10 Millionen werden 1,6 Millionen abgezogen. Sie nutzen KI, um wahnsinnig Geld zu verdienen.
AI unicorns are continuously creating financing myths.
In September this year, Anthropic, the biggest competitor of OpenAI, completed its latest round of financing plan. It received $13 billion (approximately RMB 92.6 billion) at a valuation of $183 billion (approximately RMB 1.3034 trillion). Although this figure is still far behind that of OpenAI, in the entire artificial intelligence track and even in the whole Silicon Valley, actually only OpenAI currently has a higher valuation. By horizontal comparison with other current AI unicorns, Scale AI is valued at less than $30 billion, Perplexity at $18 billion, and Anysphere, the developer of Cursor, at $9.9 billion.
More importantly, Anthropic has only been established for about 4 years. The previous round of financing was announced to be completed in March 2025, and at that time, Anthropic's valuation was $61.5 billion (approximately RMB 441.1 billion). It can be said that achieving a valuation of over $100 billion in 4 years and a 275% valuation increase in 5 months, the smooth launch of this round of financing further confirms the extremely high enthusiasm of the capital market for artificial intelligence and also further confirms Anthropic's "leading position" in the artificial intelligence track.
(Dario Amodei, the founder of Anthropic, Image source: Wikimedia Commons)
However, the biggest highlight of this round of financing is not these record - breaking figures, but that Anthropic actually said "no" to a part of the money. According to reports from multiple media, Anthropic set very strict restrictive clauses in this round of financing and clearly informed potential investors that this round of financing requires the use of self - owned funds for participation and no longer allows investment through SPV (Special Purpose Vehicle) as before.
More importantly, almost at the same time, OpenAI also issued the same statement. In an official blog post on September 5th, they clearly stated that unless the share transfer obtains the written consent of OpenAI, all transfers are "invalid". It does not recognize or participate in any equity transactions related to SPV.
In the most capital - intensive startup track in human history, why do these companies unanimously say no to money? Moreover, can they even make the top competitors in the same track stand together and jointly resist? This sense of contrast quickly attracted the attention of the entire Silicon Valley and became the absolute headline in the past period.
It is also worth mentioning that people are not arguing. Instead, countless investors and entrepreneurs unanimously praised Anthropic and OpenAI, shouting "We've suffered from SPV for a long time" and "It's high time to deal with these swindlers who get something for nothing".
Invest $10 million and take a $1.6 million cut
Let's first briefly talk about what SPV is.
SPV is the abbreviation of Special Purpose Vehicle. "Vehicle" means that in essence, SPV is an independent company with its own independent balance sheet. "Special Purpose" means that the establishment of an SPV usually has a specific and clear goal, such as isolating financial risks when investing in high - risk projects to ensure that the parent company will not be affected by the potential impact of the invested project's bankruptcy.
Of course, in the venture capital industry, the more well - known use of SPV is to "securitize the invested company". VC/PE firms that obtain financing shares will establish an SPV and "resell" the financing shares by selling the shares of this SPV externally. The advantage of doing this is that the sources of funds for VC/PE firms can be greatly enriched. Usually, VC/PE firms set a relatively high threshold for LPs, such as requiring a minimum investment of $500,000 to obtain a share. However, the threshold to become a shareholder of an SPV can even be reduced to $1,000.
Secondly, since an SPV is an entity with an independent balance sheet, holding the invested company through an SPV can allow VC/PE firms to reduce the risk of cyclical fluctuations to a certain extent. Even if the development of the invested company fails to meet expectations, there is no need to worry that the valuation change will drag down core indicators such as IRR and TVPI.
Thirdly, in recent years, as the US dollar has been in an interest - rate hike cycle and the liquidity problem is serious, wealthy investors are increasingly reluctant to invest their money in VC/PE firms with long - term and uncertain returns. Under this premise, establishing an SPV with a clear investment target and a lower participation threshold is still a good "survival strategy" for current GPs. A survey conducted by Carta at the end of 2024 found that from 2021 to 2023, the number of newly established SPVs recorded by Carta reached 1,719, a 198% increase compared with the previous three - year cycle. The median asset size of SPVs also increased from $1.18 million to $2.17 million.
In contrast, the number of newly established equity investment funds from 2022 to 2023 decreased by 38% year - on - year.
The fourth and most direct benefit is that SPV can also become a source of income for VC/PE firms. When investors indirectly invest through an SPV, they need to pay a certain handling fee to the SPV's establisher. There is no fixed standard for the amount of the handling fee, which is usually determined by "market popularity" and "market scarcity". For example, currently, the artificial intelligence track has the highest market popularity, and the financing shares of unicorns such as OpenAI and Anthropic are the most sought - after. So the handling fee for investing in SPVs related to them is more expensive than that of other companies.
How expensive is it? Michelle Lim, a Silicon Valley angel investor, recently revealed on her personal social media that many of her investor friends, including herself, have "received a large number of SPV placement invitations for OpenAI or Anthropic" in the past period. The minimum investment threshold ranges from $100,000 to $1 million, and the handling fee reaches 16%.
Of course, strictly speaking, if one can really invest in projects of the level of OpenAI and Anthropic, a 16% fee is already a reasonable price.
Touzhong Jiachuan found in a recent survey that many VC/PE firms in Silicon Valley have raised their management fees to 20%. For example, Banyan, a VC firm, charges a 20% management fee. The largest font on their official website homepage reads "AI - Native Pre - Seed Venture Capital" - "We are a pre - seed VC focusing on the artificial intelligence native field". Another VC firm, Radix Ventures, also charges a 16.5% management fee. Their official website homepage says that Radix is a Latin root meaning "source", which means they mainly invest in "hard technologies" that "promote the underlying transformation of the industry".
In addition, there are two other smart uses of SPV. Firstly, it can be used as a "continuation fund". In the US venture capital market, a large number of VC/PE firms choose to establish SPVs to continue holding assets that cannot be exited in the short term or that they are reluctant to exit. Secondly, it can be used as a "special - purpose fund". Many VC/PE firms choose to establish some SPVs for start - up companies they are optimistic about but that do not meet the standards stipulated in the Limited Partnership Agreement (LPA) of the fund.
With all these benefits, SPV is no longer exclusive to small or unorthodox VC firms and has a very high popularity in the industry. For example, among Anthropic's existing shareholders, Menlo Ventures invested through an SPV. It is an established VC firm that has been around for nearly 50 years, and its current asset management scale is approximately $7 billion. Their representative works include Uber, a milestone product in the ride - hailing software industry, and Carta and Chime, the two hottest companies in the fintech circle, as well as the well - known robot unicorn Canvas.
In short, the venture capital circle is used to referring to the secondary market trading of unlisted companies' old shares as the "Secondary market", and SPV is the most typical operation method in the "secondary market". Artificial intelligence is currently the track with the most intensive SPV transactions.
"They are all swindlers, complete swindlers"
So why did Anthropic reject such a common investment method in the current industry? The answer is actually just three words:
It's too chaotic.
On the one hand, although an SPV is established by VC/PE firms, in terms of the equity structure, this SPV is the shareholder of the startup company, and one needs to penetrate the equity to find the real investor. On the other hand, an SPV is a company in the legal sense, so an SPV can also accept investment from another SPV, and the SPV that invests in an SPV can also accept investment from an SPV.
For entrepreneurs, once they allow SPV investment, it means that the equity structure will actually become quite complex. They will find it difficult to figure out who exactly invested in them, and it will be even more difficult to effectively restrict the behavior of shareholders, let alone urge shareholders to fulfill their necessary obligations. The frequent occurrence of events such as "the BP of a star project being leaked" in recent years is a typical negative example (for details, you can refer to our previous podcast "Why do fake BPs always appear"). After multiple levels of subcontracting and selling, entrepreneurs don't know who exactly leaked the internal documents. In the worst - case scenario, it may cause an important product data leak and disrupt the original strategic plan.
Moreover, for entrepreneurs who choose to enter the capital market, "valuation management" is a compulsory course. A reasonable valuation state should be "sustainable and healthy growth" to more easily attract long - term value investors with good brand effects, rich resources, and patience, rather than "speculators" who make short - term trades. Allowing unrestricted SPV transactions will directly lead to valuation out of control. Neuralink, Elon Musk's brain - computer interface company, is a very typical case.
After the FDA (US Food and Drug Administration) first approved human clinical trials in May 2023, the valuation of SPV transactions related to Neuralink was quickly adjusted to $7 billion. However, in the most recent round of financing before that, Neuralink's actual valuation was less than $2 billion. This stimulated many biologists and medical experts to "cross - industry" and call on investors to be calm. For example, Arun Sridhar, an assistant professor at the University of Washington and a cardiac electrophysiologist, publicly shouted on social media: "This is a study to evaluate safety and durability, and it cannot justify the current valuation in any way."
The bankruptcy of Enron, an energy giant in 2001, is the most classic tragedy caused by SPV. Enron was an energy company that seemed to be booming. At its peak, its annual revenue reached as high as $101 billion, and it was rated as "America's Most Innovative Company" by Fortune magazine for six consecutive years. However, around the turn of the millennium, as the Internet bubble drove the global stock market to soar, Enron's stock price also got a share of the "bubble dividend". Therefore, in the face of huge interests, the company did something extreme: it allowed SPVs to hold a large number of the company's stocks to achieve faster and more cash - out.
At first, everything went smoothly, and Enron got a considerable amount of money as expected. However, soon, potential risks emerged. Those SPVs used these stocks to hedge the assets on the company's balance sheet, and Enron needed to guarantee the value of the SPVs to reduce risks. Eventually, when the Internet bubble burst and the global stock market entered a dark period, Enron's stock price plummeted, and it also had to fulfill its guarantee obligations to the SPVs. Then, since Enron was unable to repay the large amount of money owed to the creditors and investors of these SPVs, financial collapse followed.
(The moment Enron was delisted, Image source: Medium)
For investors, the worst consequence is being cheated out of money. Venture capital is called "alternative investment" not only because of its poor liquidity, high risk, long investment period, and inability to obtain periodic returns like stocks and bonds. More importantly, venture capital is highly opaque and non - standard. Different industries have different valuation algorithms, and different companies have different strategic advantages. These characteristics determine that the implicit threshold for venture capital investors is very high. In addition to "money", high requirements are also put forward in terms of knowledge reserve, psychological preparation, and even physical health.
Therefore, when SPV can infinitely penetrate the equity structure and infinitely lower the explicit threshold of "venture capital", the probability of ordinary investors falling into traps also greatly increases. People find it difficult to confirm whether their funds are actually invested in the company described. After generating exit returns or investment losses, it is also unclear how to exercise the rights and interests of their funds.
Although the famous article "Beware of the Retailization of Venture Capital" by Bao Fan did not mention "SPV", it was actually talking about the same thing. The scenarios he mentioned, such as "seeing the sale of 360's privatization shares in street windows and claims on forums of obtaining investment shares in Meituan - Dianping from well - known investment institutions", can still be seen in the "investment message boards" of some websites today. In essence, these are all behaviors of "turning investment shares into fast - moving consumer goods" and "using the opacity between projects and retail investors, with intermediaries adding prices layer by layer". According to Bao Fan's records, during the mobile Internet boom, such transactions were at a 30% premium compared to the actual investment price.
To put it simply, although existence is reasonable and SPV is just a market tool, the tool itself has no values. However, the problem is that the tool is an extension of human will. As the saying goes, "One man's trash is another man's treasure". When SPV is unreservedly used in venture capital, the "opacity" and the emphasis on "imagination" in the venture capital industry itself are infinitely magnified, and the "gambling nature", "paranoia", and "fanaticism" in human nature are also infinitely magnified.
Moreover, compared with the boom of the mobile Internet era, the artificial intelligence era is even more intense, involving many factors beyond "transactions", such as great - power games, era changes, and human progress. Eventually, SPV transactions found the most suitable soil in the artificial intelligence track and entered a state of wild growth.
Michelle Lim, the angel investor who exposed the high handling fees of SPV, also mentioned in the same social media post: "People are creating SPVs on top of SPVs and charging management fees for this, which is like a pyramid. Obviously, people are eager to share the success of Artificial General Intelligence (AGI), but the opportunities are too scarce." This is a relatively euphemistic way of putting it. Sarah Guo, an early investor in Figma and the founding partner of Conviction, directly and seriously warned investors on her personal social media: "The fanaticism about artificial intelligence has given rise to a group of multi - layer SPV brokers who have nothing to do with the companies. They are complete swindlers."
Looking back at Anthropic's ban, it is not difficult to explain. Anthropic's goal is to develop a large - scale foundational model to support the future artificial intelligence society. The name "Anthropic" itself represents their ultimate vision. With the support of large - scale investors such as the Qatar Investment Authority and the Government of Singapore Investment Corporation (GIC), they have no need and should not bear the risks brought by SPV.
As Bao Fan mentioned back then, if investors do not have the investment ability, it will not only bring risks to a single project. Even worse, it will allow projects that should not get funding to obtain money, thus interfering with the effective resource allocation in the venture capital market. In the artificial intelligence track where resources are most in need and competition pressure is the greatest, Anthropic really needs to take action and directly say "sorry, please leave" to SPV.