Jensen Huang just printed $60 billion in profit in three months, and he says next quarter's going to be bigger.
The Summary
- Nvidia posted $96.2 billion in Q2 revenue, up $10 billion from last quarter, with guidance of $108 billion for Q3
- Data center revenue hit $89 billion, more than doubling year-over-year as AI infrastructure spending accelerates
- Profit more than doubled to $59.7 billion, a margin that would make oil barons jealous
- CEO Jensen Huang says demand is accelerating, meaning the compute arms race isn't slowing down
The Signal
Nvidia is about to join Amazon, Apple, and Alphabet as the only companies to crack $100 billion in quarterly revenue. But there's a difference. Those companies took decades to build their revenue engines across millions of customers. Nvidia did it in less than three years, selling pickaxes to prospectors in the agent gold rush.
The data center business alone hit $89 billion, more than double last year's number. That's not a typo. Companies are spending on GPU infrastructure like it's oxygen. Every major lab, every cloud provider, every enterprise trying to build agents is bidding for the same compute. The constraint isn't demand. It's how fast TSMC can print chips.
"Profit more than doubled to $59.7 billion in a single quarter."
What's remarkable isn't just the scale. It's the acceleration. Revenue jumped $10 billion quarter over quarter, and Huang is saying demand is speeding up, not plateauing. This contradicts the narrative that AI spending would moderate as companies figured out what actually works. Instead, the useful applications are multiplying fast enough that compute demand is outrunning supply again.
Here's the signal for the agent economy:
- Training runs are getting bigger, not smaller, as labs chase emergence at scale
- Inference costs are dropping per query, but query volume is exploding
- Every agent that works creates demand for ten more
The margin story is almost absurd. A 62% profit margin on $96 billion in revenue means Nvidia is printing money faster than most countries. That kind of margin doesn't come from competitive markets. It comes from monopoly positions during infrastructure buildouts. The last time we saw margins like this was Standard Oil.
The Implication
If Nvidia's guidance holds and demand keeps accelerating, we're watching the formation of the most important infrastructure layer in the agent economy. Every company building agents is a Nvidia customer. Every breakthrough in reasoning or multimodal capability increases GPU demand. The consumer gaming business is now a rounding error in Nvidia's empire.
For builders, this means two things. One, compute costs aren't coming down fast enough to change your unit economics this year. Plan accordingly. Two, if Huang's right about accelerating demand, the companies that figure out inference efficiency will have a massive moat. The winners in Web4 won't just be the ones with the best models. They'll be the ones who can deliver intelligence at a price point that doesn't require venture funding to sustain.