Nvidia and the 70 Billion Dollar Question

There is a question making the rounds on the board that deserves a real answer, not a meme: if every hyperscaler is spending 50 to 70 billion dollars a year on AI infrastructure, who actually makes money other than Nvidia? I want to take it seriously, because it is the most important question for anyone holding this name.

Start with where the spend lands

When a cloud provider commits tens of billions to AI infrastructure, a large slice of that goes straight into accelerated compute, and Nvidia still sits at the center of that spend. Oracle guiding FY27 capex toward 70 billion is not a warning sign for Nvidia, it is the demand signal. The capex of the buyers is the revenue of the sellers, and Nvidia is the prime seller. That is why the bears and the bulls are really arguing about the same chart from two ends.

The honest risk

The bear case is not silly. It is concentration and digestion. At some point hyperscalers will want to show returns on this spend, the same way the market just punished Oracle for spending without yet proving the payback. If that scrutiny ever turns into a capex pause, Nvidia is the most exposed name on the way down because it was the most exposed on the way up. Owning the leader means owning the beta in both directions.

How I hold it through days like today

I hold Nvidia as a core position, not a trade. On a macro driven red day I do not touch it, because the session told me nothing about accelerator demand or the software moat. What I do is keep a watch list of leading indicators: hyperscaler capex guidance, any sign of order pushouts, and competitive traction from custom silicon. Those are the things that would actually change the thesis.

Until one of them flashes, the 70 billion dollar question answers itself. The more the buyers spend, the more the main seller earns. A Hormuz headline does not change who is collecting the cheque.