Nvidia и OpenAI: почему сделка на $100 млрд пошла не по плану

Imagine you give a friend a hundred rubles so he can buy you a glass of lemonade from him. In your reports — sales growth, your friend has a full glass, and investors are thrilled with your resourcefulness. That's roughly what the potential $100 billion deal between Nvidia and OpenAI looked like, the one that had been whispered about in the corridors since last September.

But it seems the party is ending before guests have finished eating the cake. News that the agreement between Jensen Huang and Sam Altman is falling apart could hit the market harder than yet another delayed AI model release. This isn't just a contract cancellation — it's a serious crack in the foundation of the entire modern AI economy.

Circular financing has become Silicon Valley's secret sauce over the past couple of years. The scheme looks elegant: chipmakers invest in model developers, who in turn spend that money buying accelerators. As a result, Nvidia's revenue sets records, and startups display frantic activity and massive computing power. This created an illusion of endless growth, where money simply circulates within a closed ecosystem. But when you're talking about a $100 billion sum, the math starts to break down. Too many zeros, too much attention from tax authorities, and too many risks in one basket. At some point, someone had to ask: where will real money come from when this loop breaks?

Sam Altman has long been playing his own game, and his ambitions clearly go beyond simple partnership. His plans to create his own network of chip manufacturing plants worth trillions of dollars are not mere fantasies — they're an attempt to break free from Nvidia's "gilded cage." Why hand over $100 billion to a competitor, even if an ally, if you can try to build your own production?

OpenAI understands that dependence on a single supplier makes them vulnerable. On the other hand, Nvidia has no desire to subsidize a potential competitor who could switch to Microsoft tomorrow or start printing their own processors at TSMC's facilities. Friendship is friendship, but a monopoly on the hardware market demands a more pragmatic approach.

We shouldn't discount regulators either. In the US and Europe, such financial carousels are viewed with increasing suspicion. If antitrust authorities prove that such deals artificially inflate company valuations and distort real market demand, the consequences would be catastrophic for the entire tech sector. Perhaps the lawyers on both sides simply decided not to tempt fate and quietly shelved the deal before people in strict suits arrived with search warrants. In the world of big numbers, silence often costs more than any contract.

This entire episode of "vanishing billions" is a vivid symptom of AI hangover — the kind that inevitably follows a wild celebration. We're entering a phase where investors are demanding not just charts of computing power growth, but real business models. If circular financing dries up, many companies will have to learn to earn real money rather than through clever capital shuffling. This will cleanse the market of weak players, but for giants like OpenAI it means the urgent need to find new ways to monetize their technologies before electricity and chip bills eat up the last remnants of venture funding.

The bottom line: The era of "free" money and circular AI investments is coming to an end. Is the market ready for a reality where you have to pay for chips with real profit, not promises to take over the world?