Anthropic за 350 миллиардов: когда осторожность становится самым дорогим товаром в мире

Remember when Anthropic was called "that startup that's afraid of AI"? A group of OpenAI defectors distanced themselves from Sam Altman to build safe models, unburdened by corporate greed. Forget about it. Today, Anthropic is discussing a valuation of $350 billion. That's not just a lot, that's the market capitalization level of the world's largest corporations that took decades to build. If the deal goes through, Dario Amodei's company will officially secure its position as the leading contender for the throne in the world of large language models.

The essence of what's happening is simple: Anthropic is preparing a tender offer. This is a mechanism that allows company employees to sell their shares and convert "paper" wealth into real money. Usually, such steps are taken to retain the best minds — when your engineer sees seven zeros in their bank app, they're much more willing to stay in the office until midnight. But the scale is striking. Along with selling employee shares, the company plans to attract another $20 billion in fresh capital. For comparison: that's more than the annual research budget of many developed countries.

Why is this happening right now? The answer lies in Claude. The third version of their model hasn't just caught up with GPT-4, it has surpassed it in some aspects — for example, in programming and working with massive context windows. The corporate sector, which has always been wary of OpenAI's unpredictability, saw in Anthropic the "adult in the room." Amazon and Google have already poured billions into the company, and now they seem ready to double down. For cloud giants, Anthropic isn't just a startup, it's critical infrastructure that guarantees their servers will be loaded with computations for years to come.

This deal can't be viewed in isolation from what's happening with competitors. While OpenAI is experiencing internal conflicts and organizational restructuring, Anthropic is demonstrating remarkable stability. They methodically build what they call "Constitutional AI," and investors are ready to pay a huge premium for this predictability. $350 billion is the price of trust that AI can be controlled. Or, more likely, it's the price of fear of missing out on dividing the biggest pie in the history of technology.

However, such a valuation also places enormous responsibility. At $350 billion, Anthropic can no longer afford to be a "lab." Now it's a profit-generating machine. Every Claude update will now be evaluated not by the quality of poetry it writes, but by how many billions it saved or earned for Fortune 500 clients. Dario Amodei has long tried to avoid publicity and the image of a "tech prophet," but with that kind of money behind him, he'll have to step into the light. The question is only whether there will be room for that very "safety" that the whole thing was supposed to be about, when stakes this high are on the table.

The main point: The era of romanticism in AI is officially over. Anthropic is now a full-fledged financial leviathan. Will they be able to live up to investor expectations, or are we witnessing the biggest bubble since the dot-coms?