Anthropic and $300 billion: when ambition outruns common sense

The world of AI venture capital has completely lost touch with gravity. While we debate new model features, a financial drama of hard-to-grasp proportions is playing out behind the scenes. Anthropic, the company former OpenAI employees founded in the name of "safe and ethical AI," now wants a $300 billion valuation. For scale: that is more than the market capitalization of most companies in the S&P 500. Dario Amodei and his team want to raise another $20 billion, and their previous rounds haven't even had time to "cool off" in the bank statements.

Let's recall how it all started. Anthropic has always cast itself as "the adults in the room." Its founders left Sam Altman because they considered his approach too commercialized and too risky. They promised us constitutional AI that wouldn't be rude and wouldn't take over the world. The irony is that building "safe" AI takes exactly as much compute as building "dangerous" AI. And compute means paying NVIDIA's Jensen Huang — up front, and a lot. That is what drives this aggressive hunt for cash.

Why would a startup that doesn't yet earn billions in profit need a $300 billion valuation? It's a game of survival. The LLM industry has reached a point where you either build your own supercomputer the size of a small city or you drop out. Google and Amazon have already put billions into Anthropic, but that clearly isn't enough. The new $20 billion goes beyond engineer salaries: it is fuel for the next leap. Training Claude 4 or Claude 5 will require infrastructure that only nation-states could previously afford. Anthropic is trying to lock in its status as the third power center alongside Microsoft/OpenAI and Google.

The more interesting question is who will put up the money. Traditional venture funds already eye numbers like these warily. Most likely, we'll once again see strategic investors or Middle Eastern sovereign wealth funds step in. For them, $20 billion is a bet on who will own the operating system of the future. If Anthropic manages to close this round, it will put enormous pressure on competitors. Players like Mistral or Cohere will look like dwarfs by comparison, forced either to find equally generous patrons or to retreat into narrow niches.

This story has a dark side, though. A $300 billion valuation saddles the company with obligations that are nearly impossible to meet in the short term. Claude is an excellent model, and many developers prefer it for its concision and reasoning. But does it bring in enough money to justify that kind of valuation? Not yet. We're watching a textbook case of bubble inflation, where an asset's value is set not by its current revenue but by investors' fear of missing the next technological era. If it turns out tomorrow that model scaling has hit a ceiling, those $300 billion will turn into a pumpkin faster than you can finish typing your prompt.

The bottom line: Anthropic is moving for good from the league of research labs into the league of financial titans. Does it have the talent to turn these billions into intelligence that genuinely transforms the economy, or are we simply watching the most expensive arms race in the history of silicon?