OpenAI's round came in under half of its target, and the price barely moved
OpenAI raised about $11bn against its $30bn target and cut the ask to $20bn at a $1.25tn valuation. A smaller cheque, not a lower price, is weak demand.
Vlad Makarovreviewed and published
OpenAI went out asking for $30 billion at a $1.4 trillion valuation. By the end of last week it had a smaller round and a smaller price, and the two did not shrink by anything like the same amount. The Wall Street Journal reported that the company had secured about $11 billion of that $30 billion target. The Financial Times reported that the target itself had been cut by $10 billion, to about $20 billion, at $1.25 trillion instead of $1.4 trillion. Those are not the same claim, and neither has been audited. They do agree on direction.
Two accounts of one shortfall
The Journal's version, as summarised by THE DECODER, is a raise that is running at less than half of plan: roughly $11 billion in commitments so far, a valuation that landed at $1.25 trillion rather than the $1.4 trillion it was aiming for, and a gap attributed to fading interest from large investors during a technology sell-off. The Financial Times' version, relayed by Bloomberg, is a company that has stopped pretending: the ask is now about $20 billion, down from $30 billion a month ago, because some investors reduced the commitments they had already made.
Read them side by side and the shortfall is large under either reading, but for different reasons and at different points in the process. One describes money that did not arrive. The other describes a number that was revised downward before the money was counted. Anyone quoting a single percentage of target achieved is picking one of those framings and discarding the other.
The backdrop both reports share is a market that repriced AI spending in the last week of October. The Nasdaq 100 fell 3.5% on Friday, the session after Alphabet's results prompted the record one-day loss of market value we covered in our piece on the capex sell-off. Weak quarterly results at Google, Amazon and Intel did the rest. A private company raising capital into that tape has one obvious lever, and it is not price.
The price moved 11%. The cheque moved a third.
On the reported figures, the valuation fell from $1.4 trillion to $1.25 trillion, about 11%. The round fell from $30 billion to $20 billion, about 33%. Those two numbers are the whole story, because in a public market they would be the same number: a stock that cannot find buyers at $140 finds them at $125, and the quantity on offer never changes.
| Figure | September plan | Reported 31 October |
|---|---|---|
| Valuation | $1.4tn | $1.25tn |
| Round size | $30bn | $20bn |
| Implied dilution | ~2.1% | ~1.6% |
A private round does not work that way. The price is negotiated with the same people who are writing the cheque, so a demand problem cannot express itself as a markdown without everyone in the room agreeing to be marked down. It expresses itself as a smaller cheque instead. Selling $30 billion of stock at $1.4 trillion is about 2.1% of the company on these post-money numbers; selling $20 billion at $1.25 trillion is about 1.6%. The smaller round is, conveniently, the less dilutive one, which is how a shortfall gets described as discipline.
The tell is not the 11%. It is that nobody tried to fill the round by cutting the price to $1.3 trillion, or $1.2 trillion, and taking the $30 billion. A company that believed it could raise $30 billion at $1.4 trillion would not have halved the ask four weeks later.
Sticky marks and who benefits from them
Private valuations are not printed every day, which is precisely why they are useful to the company that holds one. A round that closes small leaves $1.25 trillion standing as the reference point for the next raise, for employee option strikes, for the secondary sales that staff use to turn paper into cash, and for any conversation about an IPO. A round that closes at $1.1 trillion would reset all of those at once.
That asymmetry explains the shape of the news rather than any single number in it. Bloomberg had already reported a week earlier that OpenAI was negotiating at $1.25 trillion, so the price was not the surprise. The size was. What leaked out on 31 October was not a valuation event but a demand event, and demand events in private markets surface as arithmetic about how much money showed up.
The buyer of last resort changed
The most consequential line in the Journal's account is not a number. It is that OpenAI now plans to raise up to $20 billion from family offices and high-net-worth individuals. The Financial Times account has a different composition: a group of investors from the United Arab Emirates, including state-backed MGX, expected to be the largest participant, with SoftBank, Sequoia, Coatue, Founders Fund and Andreessen Horowitz also planning to take part.
Whether those are the same $20 billion is not something either report settles, and it would be a mistake to assume they are. What can be said is that the marginal buyer has changed character. Sovereign-scale strategics write cheques in the billions and accept a decade of illiquidity as the cost of a strategic position. Family offices and wealthy individuals write smaller tickets, ask earlier about liquidity, and have their own reasons for wanting exposure that are not obviously correlated with a compute build-out. Neither group is a worse investor. They are a different kind of money with a different time horizon, and a round assembled from them takes longer to close and is easier to unwind.
That fits the rest of the week's news, in which the constraint was not enthusiasm for AI but the willingness to fund it at the price being asked.
The number that makes the shortfall legible
In March, OpenAI raised $122 billion. Against that baseline, a $20 billion round is roughly a sixth and $11 billion is roughly a ninth, seven months later, from the same company, in a market that spent the interval bidding up everything adjacent to it. The comparison is not a like-for-like test of demand, because the March round had different participants and different terms. It is a scale check, and it is the reason this week's numbers read as a reversal rather than a rounding error.
Competition for the same dollars is now explicit. Anthropic is preparing to go public next month at a valuation of at least $1.5 trillion, per the FT, on third-quarter revenue of $16.8 billion with growth slowing. An institutional investor allocating to AI in November has a liquid alternative for the first time at the top of the market.
What the arithmetic does not say
None of this is audited, and almost none of it is confirmed. OpenAI, Sequoia, Founders Fund and SoftBank declined to comment to the FT; MGX, Coatue and Andreessen Horowitz did not immediately respond. Every figure here traces to people described as familiar with the matter, relayed through two newspapers on two continents within a day of each other.
The round is also still open. The Journal's "so far" is doing real work: a commitment is not a wire, and a round that stands at $11 billion on Friday can close higher. The reported plan for an IPO early next year is likewise a report, not a filing. What would settle the question is a confirmed close, a filing, or an IPO price, and none of those exists yet. Nor is a smaller round a verdict on the company: OpenAI has raised more private capital than any startup in history, and $20 billion is still $20 billion.
What it is, is a data point about price discovery in a market that has spent two years avoiding it. The valuation moved 11% and the cheque moved 33%, because that is the only way a private round can report weak demand without admitting to a lower price. The next time someone quotes $1.25 trillion as a mark, the number to ask about is the one that came in beside it.
