The Journal’s View
Public markets can confirm that investor demand exists for Anthropic shares.
Only performance over several years can show whether commitments to compute on this scale produce a profit that lasts.
That tension is what the market is being asked to price: strong results from the companies that supply the infrastructure on one side, and an open question about the return on the capital of the companies that buy and deploy it on the other.
A covered book, or a firm first day, would be evidence about the shares.
It would show that buyers were willing to own them at the price then printed.
It would not show that the economics of artificial intelligence already work.
This edition’s cutoff is 18 November 2026, and the final price is still ahead.
Until that price is a print, even the demand for the shares is unfinished.
Reuters has put the scale of the obligation in view: nearly $900 billion of purchase and lease commitments through 2045, about half of them before 2030.

A supplier can report a good quarter because that spending is real.
The buyer of the compute still has to earn a return on it.
The Journal’s view is that those are different sentences, and that a successful offering does not collapse them into one.
The test that matters is therefore not the opening print alone.
It is whether revenue grows enough to justify the outlay, whether customers remain when prices change, and whether the obligations can be funded without a further raising of this size.
Demand for the shares can be settled in a day.
The economics cannot.
There is a serious objection, and it should be stated.
The firms that sell the chips and the cloud have reported real demand.
If the book is covered, buyers who have read more of the documents than this newspaper has were willing to own the shares.
Refusing a large ambition because it is large was also the mistake made about the chip complex before the sales arrived.
That objection does not close the question.
It says the suppliers’ results are evidence about the suppliers.
It does not convert them into evidence about the return earned by the company that must pay the commitments.
The Journal will not call the shares cheap or dear from a headline that has not yet been priced.
It will not treat the first trade as the proof the economics work.
Separate the demand for the stock from the durability of the profit.
Keep the currency, rate and commodity prints in their own columns.
A successful IPO does not prove the economics of AI work.

Chanan Zevin