Nvidia и OpenAI: брак по расчету на 20 миллиардов долларов

The AI market has long resembled a closed club where everyone is both friends and creditors to each other. The news that Nvidia is about to pour $20 billion into OpenAI looks like a logical conclusion to this chain of events. Jensen Huang is not simply writing a check; he is buying insurance for a future in which his chips will be needed by everyone, but first and foremost by the most advanced player on the field. If this deal closes, we will see the largest investment in the company's history—a company that just a couple of years ago was associated exclusively with video cards for gamers.

Let us recall how we got here. OpenAI burns money faster than GPT-4 generates text. Training models requires thousands of H100 and B200 accelerators, each costing as much as a decent car. Sam Altman has long been hinting at securing hundreds of billions to build his own chip manufacturing plants to escape the shortage. And then Nvidia enters the stage. Why? The answer is obvious: so that Altman doesn't have to build his own plants. Why reinvent the wheel when you can simply become a co-owner of the one riding the fastest bike of all.

This deal creates a remarkable closed loop in Silicon Valley's economy. Nvidia reaps record profits from GPU sales. Then it takes part of that profit and invests it in its primary customer—OpenAI. OpenAI, in turn, uses that money to buy even more GPUs from Nvidia. This is not just business; it is a perpetual motion machine that leaves competitors like AMD or Intel far behind. While they try to prove their chips are just as good, Nvidia simply takes over the entire market, becoming not just a supplier but a strategic partner.

It is interesting how Microsoft sees this. Satya Nadella has spent years building a special relationship with OpenAI, investing more than $13 billion. Now the "iron king" enters the game, and his bet trumps everything done before. This could create some tension within the alliance. On the other hand, if OpenAI becomes even more powerful, all participants in this closed party will benefit. But there is a flip side: antitrust regulators. In the US and Europe, they already view Nvidia's dominance with suspicion. A direct investment of this scale in a software market leader is a red flag for any official dealing with competition protection.

We are entering an era when the boundaries between hardware production and algorithm development blur. Nvidia understands that simply selling cards is risky. Tomorrow, someone could invent a more efficient architecture or find a way to get by with less memory.

But if you own a stake in a company that dictates industry standards, you will always be one step ahead. Jensen Huang is playing the long game, turning Nvidia into something like the Standard Oil of the digital age, except instead of oil, it is computing power. Right now, every country wants its own sovereign AI, and OpenAI is the main asset in this race.

Nvidia, by investing such sums, is effectively taking on the role of financial guarantor of technological superiority.

Bottom line: Nvidia is closing the AI production cycle on itself. If the deal goes through, OpenAI will effectively become Nvidia's in-house lab for chips, and Sam Altman will have to temporarily forget about his dreams of his own factories. The only question is whether this union will become too toxic for regulators?