Anthropic Makes OpenAI Look Cheap
How much should a well-regarded artificial intelligence startup be worth on paper?
That’s the question facing scores of sovereign wealth funds, mutual fund firms and other private equity investors looking at two marquee AI opportunities: OpenAI and Anthropic. OpenAI is in the middle of its second employee share sale in the last six months, and if it succeeds, the new share price could imply an enterprise valuation of at least $80 billion.
Anthropic, a rival whose founders previously worked at OpenAI, wants to raise $2 billion from Google and others at a valuation of at least $20 billion, Kate, Anissa and Stephanie reported yesterday. If that headline causes you deja vu, that’s because it came one week after Amazon invested $1.25 billion in the two-year-old developer of large language models as part of a product and sales partnership between the companies.
Our report contains new financial data and projections about Anthropic’s top line, so we can compare the two startups’ valuations and put them into perspective.
An $80 billion valuation on OpenAI implies a multiple of around 80 times its recent annualized revenue. Even if revenue has suddenly accelerated thanks to the launch of GPT-4 “vision” features, pushing down the valuation multiple to more like 60 times annualized revenue, that’s still a sky-high price. Publicly traded enterprise software stocks on average are trading around seven times trailing revenue, according to Koyfin.
There’s another wrinkle. The OpenAI shares being sold are effectively profit units, meaning shareholders can get a return without an acquisition of the company or an initial public offering, as long as OpenAI operates in the black. But the return on those units could be capped at 100 times the original investment amount, plus a 20% increase to that cap every year, starting in 2025, thanks to how OpenAI structured its latest Microsoft deal. That may not bother institutional investors as much as it does venture capitalists, who generally want home runs with 1,000%-plus returns or higher.
Even then, Anthropic’s hoped-for valuation makes OpenAI seem like a bargain. A $20 billion valuation represents a forward-looking multiple of 200 times annualized revenue—two to three times higher than OpenAI’s. And let’s not forget that Anthropic also has a funky structure: an independent, five-person group can hire and fire the company’s board. So investors may have limited power no matter how much stock they buy.
Corporate investors such as Google, Nvidia, Amazon, Microsoft and Salesforce seem happy to pay outrageous prices for stakes in hot AI developers like Anthropic, as we have continually noted. These investors don’t care about financial returns as much as they care about whether the startups can boost their sales or at least spend money to rent their cloud servers or use their chips. So don’t be surprised to see them participate in funding rounds with astronomical valuations, including Anthropic’s upcoming financing. And as sure as the sun sets in the west, OpenAI will use its employee share sale to raise equity financing (and maybe some debt) at an even higher price as it fulfills CEO Sam Altman’s promise to be “the most capital-intensive startup in Silicon Valley history.”
But corporate investors can only soak up so much of a startup’s cap table. Financial investors will have to fill the gap. Because not all of those investors are SoftBank’s Masayoshi Son, who seems to have never met an AI company he didn’t like, fund managers will face a tough choice.
LLMs made by OpenAI and Anthropic will either become the next operating system, perhaps elevating these companies to the trillion-dollar heights and overcoming competition from open-source AI and the likes of Google; anxious regulators and politicians; shortages of chips and compute-power sources; and the models’ natural limits. Or they won’t.
A safer choice for investors is to make the easiest AI bet of all: buying shares of Google and Microsoft.