What happened

On July 7, the US political outlet NOTUS reported on an internal Treasury document. In a confidential draft, staff analysts warned that if AI monetization falls short of expectations, the shock could resemble the dot-com bust of the early 2000s. And not just in equities. The draft pointed to data-center financing, private credit markets, chip manufacturing, and power utilities.

One caveat: this is not Treasury's official position. A spokesperson drew a firm line — the draft had not gone through review and does not represent the department's view. The official line is the opposite: AI will power America's new golden age.

The same week, JP Morgan called the recent pullback in chip stocks a buying opportunity, arguing that new fab capacity cannot come online before 2028. Morgan Stanley, somewhere in between, recommended trimming semiconductor exposure.

The market's eyes now turn to mid-July. ASML reports second-quarter results on July 15. TSMC follows on the 16th.

Why it matters

Let's be upfront about something. This news strikes directly at Signal Note's own frame. We have read semiconductors as the power grid of the AI economy, and hyperscaler investment — Google, Amazon, Microsoft, Meta, the giant data-center operators — as evidence of a civilization-scale transition. If the draft's warning is right, the lens itself needs regrinding. So we're not dodging it.

But look closely at this debate and something odd appears. The analysts who wrote the draft, the spokesperson who disowned it, JP Morgan, Morgan Stanley — every one of them is speaking in forecasts. If monetization fails. If supply stays tight through 2028. When even one building can't agree with itself, forecast versus forecast settles nothing. The louder voice merely appears to win.

We look somewhere else. Contract language.

Contract language means the specific phrases executives use on earnings calls. "Sold out." Prepayments — deposits customers put down for fear of missing out on supply. Long-term supply agreements. As long as those words stay alive, the supplier names the price. When "pricing pressure," "inventory correction," or "capacity expansion completed ahead of schedule" start appearing instead, the bottleneck is loosening.

Even in the dot-com bust, share prices were the last thing to fall. Contracts cooled first.

Apply that yardstick to this debate and the terrain looks different. Even the warning draft concedes that today's AI companies, unlike their dot-com predecessors, run on healthy cash flows. JP Morgan's optimism ultimately rests on the physical reality of supply contracts — "no new capacity before 2028." Both sides, without realizing it, are pointing at contract-side evidence.

Watching this debate, I jotted a line in my notes — the market cheers semiconductor growth, but it never quite trusts the speed or the substance. The draft's monetization warning questions the substance. Morgan Stanley's underweight questions the speed. And that doubt currently points at one signal: recent record earnings at memory makers rest on rising prices, not rising volumes. A boom driven by price alone is a classic late-cycle feature. Is this the top? I honestly don't know — but the language in the contracts isn't saying so yet.

That's why next week matters. ASML and TSMC sit at the deepest upstream of the AI supply chain. Whether contract language holds or wavers on those two earnings calls is, in our reading, the first verdict on who is right — the draft or JP Morgan. The referee of the bubble debate isn't a Wall Street report. It's the contracts upstream.

What to watch

For information only. Not investment advice.