Nvidia and OpenAI: Why the $100 Billion Deal Went Off the Rails

Imagine you give a friend a hundred rubles so he can buy a glass of lemonade from you. Your books show sales growth, your friend has a full glass, and investors are thrilled by your entrepreneurial spirit. That is roughly what the potential deal between Nvidia and OpenAI, worth an almost unimaginable 100 billion dollars, looked like.

It has been whispered about in backrooms since last September. But the party seems to be ending before the guests have finished their cake. News that the agreement between Jensen Huang and Sam Altman is falling apart could hit the market harder than yet another delay to a new model's release.

This is not just a canceled contract. It is a serious crack in the foundation of the entire modern AI economy.

Circular financing has been Silicon Valley's secret sauce for the past couple of years. The scheme looks elegant: chipmakers invest in model developers, and those developers in turn spend the same money on accelerators. As a result, Nvidia's revenue sets records, while startups show feverish activity and enormous computing capacity. It created the illusion of endless growth, with money simply circulating inside a closed ecosystem. But when the sum reaches 100 billion, the math starts to fail. There are too many zeros, too much attention from tax authorities and too much risk in one basket. At some point, someone had to ask: where will the real money come from when this circle breaks?

Sam Altman has long been playing his own game, and his ambitions clearly go beyond a simple partnership. His plans for a network of chip fabs costing trillions of dollars are not just fantasies. They are an attempt to break out of Nvidia's "golden cage." Why hand 100 billion to a competitor, even an allied one, when you can try to build your own production? OpenAI understands that depending on a single supplier makes it vulnerable. Nvidia, for its part, is not eager to subsidize a potential rival that could walk away to Microsoft tomorrow or start printing its own processors at TSMC's facilities. Friendship is friendship, but a monopoly in the hardware market demands a more pragmatic approach.

Regulators should not be written off either. In the US and Europe, such financial carousels are viewed with growing suspicion. If antitrust authorities prove that such deals artificially inflate company valuations and distort real market demand, the consequences for the entire tech sector would be catastrophic. Perhaps the lawyers on both sides simply decided not to tempt fate and quietly let the matter drop before people in dark suits arrived with search warrants. In the world of big numbers, silence is often worth more than any contract.

This whole episode of the "vanished billions" is a vivid symptom of the AI hangover that inevitably follows a wild party. We are entering a phase in which investors demand not only growth charts for computing capacity but also real business models. If circular financing dries up, many companies will have to learn to earn money for real, not through deft movement of capital. That will clear weak players out of the market. For giants like OpenAI, though, it means urgently finding new ways to monetize their technology before electricity and chip bills eat up what is left of the venture funding.

The key takeaway: the era of "free" money and circular investment in AI is coming to an end. Is the market ready for a reality in which chips must be paid for with real profit, not promises to conquer the world?