AI-related companies have issued or priced more than $300 billion of stock in 2026, according to a count that leaves Anthropic’s initial public offering out. The figure was published by The Information on Nov. 8 and covers everything else the sector sold. Another $100 billion of issuance is queued behind it.
The number describes how the buildout is being financed. Chips, land, power contracts and salaries are cash costs. Equity is how a company turns a private valuation into cash without a lender’s consent or a rating committee’s opinion, and at these prices it is the cheapest currency most of them have.
It is also a floor. The total counts what was sold, not what was wanted. Companies that filed and then waited are not in it, and neither is any debt raised against the same assets.
Two things are happening at once. Public investors are being asked to fund a capital cycle whose returns arrive years out, and private owners are being handed a way to convert paper gains into dollars while the multiple that produced those gains still holds.
The buyers are not bottomless. Each quarter the market absorbs a finite amount of new paper, and index funds take it only at the weight the index assigns. Active managers have to choose, and choosing means comparing an AI issuer against a software company with a renewal rate and a chipmaker with a cycle.
A convertible bond sits between the two. It pays a coupon and converts into stock later, which lets an issuer raise money at a low rate and defer the dilution. It is also a supply of shares dated to a future price, so an equity total understates what the market will eventually have to absorb.
The $100 billion in the queue is the more informative number. A queue is a set of decisions that can be postponed, and issuers postpone when the price they are offered falls below the price they believe they are worth. Watch the calendar, not the total: a full calendar is a statement that sellers and buyers agree on price.
There is a signal inside the mix. A company selling stock at a high multiple is telling the market it thinks its own paper is expensive relative to cash. Management teams that expected the multiple to expand would borrow or wait. The volume says a large group of them has decided this is a good moment to be a seller.
That decision has a cost attached. A public issuer reports every quarter, discloses customer concentration and defends its accounting to people paid to find problems. It also starts a lockup clock, and the first two quarters after expiry will be read as a verdict on insiders rather than on the business.
The competitive logic pushes the same way. Whoever prices first sets a reference for everyone behind them. A strong debut lifts the followers and validates the category; a weak one hands the skeptics a number built from the leader’s own disclosure, and the next issuer spends its roadshow explaining the first one’s chart.
So the calendar is a confidence instrument. One large deal that breaks, either at pricing or in its first month of trading, costs more than that issuer. It reprices the queue, because every banker on it is marking the same comparables.
That is the mechanism by which a window closes. Not a judgment about artificial intelligence, and usually not a change in any company’s revenue. A single failed offering, a widening discount on a follow-on, a lockup expiry that meets more selling than the syndicate expected: any of those can move the clearing price enough that the next issuer chooses to wait.
There is another side. Capital raised now buys capacity that will be sold for years, and a company that funds a multi-year asset with equity rather than short-dated debt has taken less risk than the headline suggests. The dilution is real; so is the absence of a maturity wall.
The risk sits in the mismatch. Equity has no maturity, but it has a price that resets daily, and employees paid in stock watch it. A falling multiple raises the cost of retaining the researchers who build the product, which is a slower and more expensive problem than a failed deal.
What would change the picture is a run of offerings that price at or above the last private mark and trade well afterwards. That would tell the market the demand is deeper than one quarter’s savings, and the queue would clear.
Until then, $300 billion is the year’s score and $100 billion is the year’s open question. The first number is history. The second one is a decision that has not been made yet.
