AI's third constraint is capital, and the test is whether financial engineering bridges to revenue or widens the bubble

The AI buildout just added a third constraint, and the test is whether it holds. Compute came first, power second, and now capital arrives as its own bottleneck. The mechanism in play is financial engineering, and it sits inside the growth story rather than beside it. Nvidia's new long-duration funding vehicle lets investors lock in for the years AI revenue needs to mature. Google is already tapping equity alongside, naming the pattern as a market move rather than a one-company story.

Thompson's framing puts the question plainly: "If AI is as valuable as it seems, then it should pay for itself, but that hasn't happened yet." The clock here runs on years, not quarters. The test is whether AI earnings catch the spend curve before the funding structure matures.

The same machinery that bridges to sustainable revenue also widens a bubble's blast radius. Long-duration capital papers over the gap between today's outlay and tomorrow's income, buying time at the cost of heavier obligations if the bet misses. Margin pressure at Nvidia, and at the hyperscalers building beside it, is what makes the trade attractive now; whether the bet pays is what determines whether the same trade was a bridge or a detonator.

The constraint stack is the new story. Each leg extends the runway while raising the cost of a miss.