Why the Current Business Model is Not Sustainable for AI Native Companies 🤖

If we look at the current benchmarks, Anthropic and OpenAI are usually leading the charts when it comes to the most powerful AI models, but this kind of growth is not sustainable.

So far, Anthropic has raised over 100 billion dollars, and most likely they will not be able to raise more in the private market. When you raise capital in the private market, you have a very limited number of rounds until all the investors are so diluted that the only option to raise additional capital is going public. It is not just about dilution. Every private investor invests in a company at an early stage because they want to cash out at some point. Investors usually cash out in two ways: either when the company goes public or when it is acquired by another company.

🧱 The Acquisition Blockade

With the latest round of raising capital, Anthropic was valued at more than 900 billion. With such a valuation, acquisition of the company is almost impossible. Even if one of the tech giants like Google would try it, Anthropic has the ability to bypass the investors’ voting rights and block that offer.

Anthropic actually has a non-standard legal setup that is very unique. It is a Public Benefit Corporation (PBC), meaning it is legally allowed to balance profit with the public good. Being a PBC is not unique, but they have a built-in mechanism called the Long-Term Benefit Trust (LTBT). The LTBT is composed of five voting trustees with deep expertise in AI safety, national security, and public policy. Most importantly, these people have to be completely independent from Anthropic with no financial or equity interest in the company.

These trustees have the power to elect and remove members of Anthropic’s corporate board of directors. The idea is that in all companies, the board of directors serves the investors. With the LTBT, they are serving people who have no interest in the company. The trust can also protect the firm from a hostile takeover. Even though Google owns around 15% of Anthropic and Amazon also owns in the high teens, their ownership is capped and they will not be able to take over.

📉 The Limits of Going Public

Okay, so what is left for Anthropic? The only way to raise additional capital is to go public. When a company goes public, they raise capital initially, but after that, there is no money flowing into the company from shares. After that, you need debt.

Technically they can issue more shares, but doing so continuously dilutes existing shareholders and drives down the stock price. Public markets have little patience for companies that issue new equity just to cover operating losses (like ongoing compute costs).

Let’s see how much they can raise when they go public. When a company goes public, they issue new shares (free float). SpaceX just raised a record amount of capital at 75 billion, and they build rockets, have satellites, and also their own powerful AI model. Even if Anthropic would be valued around 1 trillion and they issue a 5% free float, the most they will be able to raise is an extra 50 billion. And what happens next?

If a company’s only source of income is shipping the best AI models, it will always be capital hungry. Anthropic is not alone in this race and their prices will always be competitive. The only companies that can play these capital-intensive games are companies with already profitable existing business models like Alphabet, Meta, and others.

🔄 The Race and the Need for a Pivot

I find it very difficult to imagine that one model will win the race and the race will be over. The race will continue for decades and multiple models will survive. Computing costs will go down, but it is not really good news for Anthropic. If computing costs go down in general, prices will drop as well due to competition.

What is the scenario where Anthropic survives and thrives? I think a complete change of business model is needed at some point, turning into a more licensing-based model and expanding more into hardware and software.

What I mean, for example, is what happens if an Anthropic model helps to find a cure for a certain disease? The company should get royalties for that groundbreaking medicine. They have to expand into hardware systems where they would license their models, like robotics and smart devices. The only other option is that they have such a massive breakthrough in technology that other companies cannot catch up, but this is almost impossible. Basically, rather than selling “intelligence by the token,” they need to capture the downstream value their intelligence creates.