Anthropic: 20 миллиардов долларов на попытку переиграть OpenAI

Twenty billion dollars. Let that sink in. This is not just the budget of a small nation or the cost of a major airline. Now this is the entry ticket into the club of those claiming to create "divine" artificial intelligence. Anthropic, a company that for a long time positioned itself as an "ethical and cautious" alternative to OpenAI, is completing a deal that will forever change its status. According to sources close to the process, this massive funding round will be closed literally any day now. This event marks the end of the era of romantic AI enthusiasm and the transition to total war between capital interests.

To understand how we got here, we need to recall the prehistory. Anthropic was founded by defectors from OpenAI who left Sam Altman precisely because of disagreements over safety and excessive commercialization. The Amodei siblings wanted to build AI that wouldn't go crazy and destroy humanity. However, reality proved ironic: to make AI safe, you first need to make it intelligent. And to make it intelligent, you need hundreds of thousands of Nvidia graphics cards and billions of dollars for electricity. As a result, the "safe" startup today is forced to seek the same astronomical sums as those from whom they once fled.

Why are investors willing to write such checks right now? The answer lies in the recent successes of the Claude 3.5 Sonnet model.

For a long time, Anthropic was in the role of a chaser, but their latest iteration made many acknowledge: Claude is not just on par with GPT-4o, but in some aspects — for example, in programming and linguistic naturalness — it actually surpasses it. Investors saw in this a chance to break OpenAI's monopoly. When you have a product that actually works and developers like it, 20 billion dollars no longer seems insane.

It seems like a reasonable bet on the second horse in a race where the prize is control over the future of global software.

However, this financial windfall has a downside. By receiving such money from giants like Amazon and Google, Anthropic inevitably loses some of its independence. Now the company is not just a laboratory of scientists, but a powerful lever in the hands of cloud providers. Most of these billions won't even settle in Anthropic's accounts. They will practically immediately return to the same investors as payment for renting server capacity. This is an elegant scheme where money makes a circle and returns to big tech business, leaving the startup only the right to continue the race.

What does this mean for the industry as a whole? First, the barrier to entry into the premier league of LLMs has become practically unreachable for new players. If you don't have tens of billions behind you, you can't even start a conversation. Second, we are seeing consolidation around two or three centers of power. The AI market is turning into a duopoly or triopoly, where Anthropic plays the role of the main intellectual counterweight to OpenAI. And although the Amodeis still talk about safety, their actions show: right now, scale matters most. Without massive computing power, any talk of ethics will remain just conversations in empty San Francisco offices.

The key point: Anthropic has finally become a mirror image of OpenAI. Will the company be able to preserve its "soul" and safety principles when tens of billions of investor dollars are at stake and the need to show multiple growth every six months?