Nvidia just projected $673 billion in sales. The tech world is popping champagne. They should be checking the exits.
You’ve seen the headlines. AI demand is “widening.” The projections are going parabolic. We are supposed to believe this is the new reality, a permanent plateau of explosive growth. But if you look closely at the math, a different picture emerges.
The projected AI capital expenditure for this year and next is crossing $1 trillion annually. Nvidia projecting $673 billion in sales doesn’t sound like an anomaly when you realize the industry is lighting a trillion dollars on fire to get there.
We aren’t building an AI economy; we’re building a $1 trillion bonfire and calling it progress.
The real story isn’t Nvidia’s sales numbers. The real story is the systemic risk embedded in the AI supply chain. This isn’t a traditional market. It’s a positive feedback loop. Nvidia’s massive revenue is used as proof that AI is working, which triggers more investor FOMO, which unlocks more capital for tech giants, which immediately flows back to Nvidia for more GPUs.
It’s a cycle that feeds on itself. And it works beautifully—until it doesn’t.
Consider the margins. Nvidia is operating with profit margins above 50%. Before this AI craze, the only entities known to command those kinds of margins were drug trafficking cartels. When big tech is handing over that much premium for silicon, they aren’t buying efficiency. They are buying survival.
When your supplier operates with cartel-level profit margins, you aren’t a partner—you’re a hostage.
The tech giants buying these chips are not doing so because their AI applications are generating massive downstream revenue. They are buying them out of fear. Fear of missing out. Fear of being left behind by competitors. It is speculative infrastructure spending on a scale we have never seen.
But the ground is already shifting beneath their feet. While Nvidia projects $673 billion in sales, the underlying technology is undergoing rapid commoditization. Small AI models are growing in capability at a staggering pace, and they require a fraction of the compute power to train and serve. The competition is also waking up, with suppliers in China and the US rushing to provide alternatives.
You don’t need a $673 billion GPU monopoly to run a model that fits on a smartphone.
A positive feedback loop is just a polite way of describing a system that eats itself.
If the downstream economics don’t materialize—if the apps don’t generate revenue, if the enterprise subscriptions don’t cover the compute costs—the $1 trillion annual capex evaporates. The tech giants will slash their infrastructure budgets. Nvidia’s order book will empty overnight.
We have seen this movie before. The Dot-Com bubble wasn’t popped by a lack of internet usage; it was popped by the realization that the infrastructure built to support it was wildly overpriced and premature. The internet eventually changed the world, but not before wiping out the investors who bought the narrative at the peak.
The AI revolution is real. The transformation is coming. But Nvidia’s $673 billion projection isn’t a victory lap for the industry. It’s the flashing red light on the dashboard of a market stretched to its absolute limit.
The AI revolution is coming. But first, we have to survive the correction.
FAQ
Q: Isn't Nvidia just supplying the picks and shovels for a gold rush?
A: Even pick-and-shovel suppliers go bankrupt when the miners run out of funding. Nvidia isn't selling to end-users; they're selling to tech giants gambling on speculative infrastructure.
Q: What happens if the downstream AI apps don't monetize?
A: The $1 trillion annual capex vanishes. Tech giants slash budgets, Nvidia's order book evaporates, and the entire sector faces a violent, Dot-Com-style correction.
Q: Can't Nvidia just pivot to software and services to survive a hardware crash?
A: Services won't save them from a hardware collapse. You can't service a rack of GPUs that nobody wants to buy because the end-user economics didn't pan out.