Nvidia and OpenAI: The $100 Billion Turned Out to Be a Fairy Tale

Sam Altman has long taught us that in artificial intelligence the counting is done in trillions, but reality sometimes reminds us of itself with a cold shower of smaller numbers. The story of how Nvidia was going to pour an astronomical $100 billion into OpenAI has suddenly changed tone. It turns out that even the biggest beneficiary of the AI boom has a limit to its generosity, or at the very least a very clear idea of how exactly that money should work.

Let's recall the context of the past few months. OpenAI has turned into a black hole for capital, swallowing billions faster than they can be printed. Jensen Huang and his Nvidia team, in turn, became the main dealers of this digital fever. It looked like a match made in heaven: some build the future, others sell the shovels for it in the form of H100 chips. When the figure of $100 billion flashed across headlines, the market took it as a final victory for OpenAI. If Nvidia is giving that much money, the thinking went, the question of dominance is settled once and for all.

But fresh data from Bloomberg puts the emphasis elsewhere. It turns out the $100 billion is not a check Jensen will write to Sam at the next presentation. It is more of a long-term memorandum of intent, stretched over years and possibly decades. In the near term, Nvidia is ready to put about $20 billion into OpenAI's capital. Admittedly, that is still an enormous sum, one that could upend any other market, but against Altman's ambitions it looks like going on a strict diet.

Why did Nvidia decide to hit the brakes? The answer lies in Jensen Huang's strategy. Nvidia today is not just a hardware maker; it is a full-fledged architect of the new economy. Putting $100 billion at once into a single player, even the most promising one, would be a strategic mistake. Nvidia has obligations to other partners, from Google to Meta*, and getting too close to OpenAI could raise unwanted questions among regulators and competitors. Besides, $20 billion today is a way to keep a finger on the pulse without taking on all the risks tied to a possible slowdown in the development of large language models.

For OpenAI, this news sounds like a warning bell. The era when you could walk into an investor's office with a pitch about saving humanity and leave with an unlimited credit card is gradually ending. Now even the closest friends and partners need proof of efficiency. Each new GPT iteration costs more to train, and the quality gains no longer look as explosive as the jump from the second version to the fourth. Investors, Nvidia included, want to see not just polished demos but a real business model that won't collapse at the first shift in market conditions.

What does this mean for the whole industry? First, we will see more cautious allocation of resources. If even Nvidia, which literally prints money selling chips, picks a tactic of gradual investment, then the time of frenzied hype is giving way to a time of pragmatic building. Second, it gives a chance to competitors such as Anthropic or xAI. The market remains open, and OpenAI's monopoly, propped up by Nvidia's endless money, is for now postponed.

The bottom line: Jensen Huang reminded everyone that he is a businessman first, not a patron of the arts. Will this $80 billion shortfall prove critical for Sam Altman's plans to build AGI?

*Meta has been designated an extremist organization and is banned in Russia.