The mystique of AI company valuations: Anthropic at 200 times market-to-sales ratio, OpenAI at 80 times.

How should the skyrocketing valuation of AI companies be evaluated? Many sovereign wealth funds, mutual funds, and other private investors face this question when dealing with top AI companies like OpenAI and Anthropic.

Currently, OpenAI is conducting its second round of employee stock sales, and if successful, the company's valuation will reach $80 billion to $90 billion.

Anthropic, founded only two years ago by a former OpenAI employee, aims to raise $2 billion from Google and several other companies at a valuation of $20 billion to $30 billion. Just a week ago, Amazon invested $1.25 billion in Anthropic and reached a product and sales partnership.

OpenAI's valuation of $80 billion means its price-to-sales ratio is around 80 times. Even if the company's revenue skyrockets due to the launch of the new GPT-4 feature, even if the price-to-sales ratio is lowered to around 60 times, the valuation is still sky-high. In comparison, the average price-to-sales ratio of many software companies listed on the market is only around 7 times.

Moreover, the current shareholders of OpenAI are in a profitable position, which means that as long as OpenAI continues to be profitable, shareholders can reap substantial returns compared to their initial investments, even without acquiring or IPO-ing the company. However, if these stocks are to be sold, the selling price may be limited to 100 times the initial investment, and starting from 2025, the multiple restriction will increase by 20% each year. Although this has little impact on institutional investors, later-stage venture investors may remain cautious because they demand a return rate of over 1000%.

But Anthropic's requested valuation might even make OpenAI seem cheap. If we calculate based on the lower end of the company's sought-after valuation of $20 billion, it means the price-to-sales ratio is over 200 times, more than three times that of OpenAI. Let's not forget that Anthropic's organizational structure is also quite unique, with an independent five-person team that can hire or dismiss the company's directors, so no matter how many shares investors buy, they won't have a say.

Enterprise investors like Google, Nvidia, Amazon, Microsoft, and Salesforce may be willing to pay a high premium for hot AI companies. These investors are not concerned about financial returns but mainly consider whether these startups can increase the revenue of the investing companies or at least absorb their own business, such as renting cloud servers or using their own chips.

Therefore, it is not surprising that these giants join the financing and push the valuation of AI companies to the sky. OpenAI will use its employee stock sales to raise a sky-high amount and fulfill the CEO's commitment to "make OpenAI the most capital-intensive startup in Silicon Valley."

However, enterprise investors cannot buy all the equity of AI companies, and the remaining portion needs to be filled by financial investors. But not all financial investors are like Masayoshi Son, who is open to any AI company. Fund managers will face difficult choices. OpenAI and Anthropic, the large language models created by them, have the potential to become the next generation operating systems. They could potentially lead these two leading companies to trillion-dollar heights, gaining an advantage in competition with open-source AI projects such as Google. This development may also cause headaches for regulators and politicians, as well as potentially leading to a shortage of chips and computing power.

Some industry insiders suggest that the safest choice for investors is to directly buy stocks in Google and Microsoft.