ASML says customers are already ordering chip tools for 2029 and 2030, easing AI spending fears

In a boom people over-order, so orders alone don't prove much about real demand.

Are these 2029 and 2030 orders backed by OpenAI and Anthropic earnings? Both are still losing money, so I'm not sure what's paying for it.

I'd push back on the stock not moving. ASML is up almost 70% this year and sits near historic highs, so the market clearly bought the story.

I don't think running out of steam is the failure mode here. These machines make chips that hit the market in about five years, so ASML will probably be the last to feel an AI bust.

Orders don't mean billings or profits though. If the bubble pops, none of that revenue shows up.

Even after a pop, ASML's machines stay useful, since people always want faster chips. But the order book would dry up and take a couple years to recover, so the echo lasts years.

Buying machines you don't use costs you maybe several hundred million dollars, and some of that comes back as resale value. Not buying in time and being unable to produce could cost tens of billions.

The title confuses me. How do orders for machines delivered years out predict AI spending?

Companies need years to earn back the cost of those machines, so they wouldn't order this many without expecting a return.

Backlogs are speculative, like airplane orders. You lock in a slot now and can sell it, push it back, or cancel later.

Long lead time slots for a finite resource are basically out-of-the-money call options on future demand. Firms are saying demand is plausible, not how likely it is.

Record chip gear orders in February 2000 also looked like confidence, and then the bust came. A peak only looks like a peak afterward.

This looks like a lagging indicator to me. These orders come early in a build-out, so they would only drop after the bust starts.

How many of these orders carry a zero or near zero cancellation fee? That's the real question.