Nvidia и OpenAI: сто миллиардов оказались красивой сказкой

Sam Altman has long accustomed us to the idea that in the world of artificial intelligence, numbers are measured in trillions, but reality sometimes reminds us of itself with a cold shower of smaller figures. The story of how Nvidia was about to pour an astronomical hundred billion dollars into OpenAI suddenly changed its tone. It turned out that even the main beneficiary of the AI boom has limits to generosity, or, at minimum, a very clear understanding of exactly how this money should work.

Let's recall the context we've been living in for the past few months. OpenAI has turned into a black hole for capital, consuming billions faster than they can be printed. Jensen Huang and his team at Nvidia, in turn, have become the main dealers in this digital fever. Seemingly, a union was made in heaven: some build the future, others sell the shovels in the form of H100 chips for it. When the hundred billion figure started appearing in headlines, the market perceived it as OpenAI's final victory. They say if Nvidia is giving that much money, it means the question of dominance is settled once and for all.

However, fresh data from Bloomberg puts different emphasis. It turned out that hundred billion is not a check that Jensen will write to Sam at the next presentation. It's rather a long-term memorandum of intent, stretched over years, and possibly decades. In the near term, Nvidia is ready to direct about twenty billion dollars to OpenAI's capital. Certainly, this is still an enormous sum, capable of overturning any other market, but in the context of Altman's ambitions, it looks like a switch to a strict diet.

Why did Nvidia decide to slow down? The answer lies in Jensen Huang's strategy. Nvidia today is not just a hardware manufacturer, it's a full-fledged architect of a new economy. To invest a hundred billion at once in a single player, even if the most promising one, is a strategic mistake. Nvidia has obligations to other partners, from Google to Meta, and excessive rapprochement with OpenAI could raise unnecessary questions from regulators and competitors. Moreover, twenty billion today is a way to keep a hand on the pulse, without taking on all the risks associated with possible slowdown in large language model development.

For OpenAI, this news sounds like a warning bell. The era when you could come to an investor with a presentation about saving humanity and leave with an unlimited credit card is gradually ending. Now even the closest friends and partners need to prove efficiency. Every new iteration of GPT becomes more expensive to train, and the quality improvement no longer seems as explosive as the transition from the second version to the fourth. Investors, including Nvidia, want to see not just beautiful demos, but a real business model that won't collapse at the first market change.

What does this mean for the entire industry? First, we'll see more cautious distribution of resources. If even Nvidia, which literally prints money from selling chips, chooses tactics of gradual investments, it means the time of crazy hype is being replaced by pragmatic construction. Second, it gives competitors like Anthropic or xAI a chance. The market remains open, and OpenAI's monopoly, backed by Nvidia's endless money, is postponed for now.

The main thing: Jensen Huang reminded everyone that he is first and foremost a businessman, not a patron. Will this shortfall of eighty billion be critical for Sam Altman's plans to create AGI?