Compute Is Revenue: Why AI Infrastructure Demand Is Still Accelerating
Ten days ago I wrote that the world is structurally short compute. Jensen Huang just made the case in plain language: compute is revenue, and demand is accelerating with the buildout still running at full steam.
The receipts are already visible. The quarter came in at roughly $96 billion, data-center revenue was up 117%, and the next leg of infrastructure spending is being pulled forward rather than postponed. This is not merely a story about one company; it is a supply-and-demand problem across the entire AI stack.
The underlying mechanism is familiar. Jevons paradox says that when a resource becomes more efficient, the cost per unit of work falls and demand for the resource can expand. Watt made steam power more efficient, and Britain burned more coal. More capable and cheaper compute can create more workloads, not fewer.
That is why the important question is not whether AI servers become more efficient. It is whether efficiency expands the market faster than supply can respond. Right now, the evidence still points to an infrastructure cycle where compute remains the scarce inputand the revenue opportunity follows the scarcity.
