The commitment letters are being gathered now, according to people familiar with the matter, and the terms carry a deadline as much as a number: at least 30 billion dollars of new capital, a valuation near 1.4 trillion dollars, and a close timed for before Anthropic’s shares begin trading in the middle of November. Reuters reported the fundraising on October 19, and Bloomberg published a similar account the same day.
SoftBank Group and MGX, the Abu Dhabi fund, are in talks to take part, the people said. Neither company confirmed the discussions. OpenAI declined to comment on the round.
The sequence is the story. A private company raising money is routine. A private company raising money against a rival’s listing date is not.
Anthropic’s initial public offering is expected to price in the first half of November, and once it does, every large investor in the sector will hold a public multiple to argue with. OpenAI’s round is an attempt to set its own mark before that happens.
The figures shown to investors explain the timing. Annualized revenue reached about 50 billion dollars in the third quarter, up nearly 80 percent from a year earlier, and the company has told backers it expects to leave the year at a 70 billion dollar run rate. Anthropic’s annualized revenue has moved past 65 billion dollars on its own trajectory, which puts the gap between the two at something an investor can measure in quarters rather than years.
Both are run rates, not audited revenue, and the distinction matters when the numbers are used to justify valuations in the trillions. A run rate multiplies the most recent month by twelve. It flatters a business growing quickly, punishes one with a soft month, and says nothing about what customers paid across a full year.
The prior round is the reference point everyone uses. In March, OpenAI raised 122 billion dollars at a post-money valuation of 852 billion. Seven months later it is asking for a quarter of that sum at a valuation a little over 60 percent higher. The step-up is smaller than the one that preceded it, which is either discipline or the arithmetic of scale, depending on which investor is talking.
Anthropic’s backers have already done part of that arithmetic in public. Expectations for its IPO valuation have been cut from 2 trillion dollars to roughly 1.5 trillion, according to people involved in the discussions. A number that moves down before the shares are sold is a number that has been tested.
The capital has an obvious destination. Both companies have committed to computing capacity years in advance, and those commitments are paid whether or not the models trained on the machines earn back their cost. A financing round at this size is not primarily a bet on next quarter’s revenue. It is a prepayment on a construction schedule that has already been signed.
The investor list says as much. Sovereign funds and strategic partners can hold an asset through a cycle in which a public shareholder cannot. SoftBank has been the largest backer of the company and carries its own obligations to the lenders who fund those positions.
MGX has a mandate to build computing capacity in Abu Dhabi and a reason to want OpenAI’s models running on it. Neither is buying a quarter. Both are buying a position in a supply chain.
What the round does not settle is the price of the equity after the lock-ups expire. Private marks are set in negotiations between a company and the people who already believe in it. Public marks are set by whoever shows up on a given morning, including the investors who bought at the top and the funds that have to report a valuation to their own clients every quarter.
The same pool of money is being asked to fund three things at once: model developers, the data centers they use, and the chips inside those data centers. That pool is not large. A handful of sovereign funds, a few technology-focused mutual funds, and the balance sheets of companies that sell hardware to all of the above. Each new round at a higher price makes the next one harder to fill, which is why the timing matters more than the total.
There is also the matter of what Anthropic’s filing will disclose. A prospectus carries audited revenue, customer concentration, contract terms and the cost of the computing it has bought. Those are the numbers private investors have been estimating from the outside for two years. Once they are on paper, the estimates stop being estimates.
OpenAI has survived every previous comparison with its rival by growing faster than the comparison implied. The 50 billion dollar run rate in the third quarter and the 70 billion dollar exit it has guided toward are the strongest version of that argument it has produced. Whether the argument holds at 1.4 trillion dollars is a question the public market will answer first, for a different company, a few days earlier.
Two prices will be set within weeks of each other, for two businesses that sell similar things to similar customers. The first is a negotiation between a company and the people who already own a piece of it. The second is a quotation from anyone with a brokerage account. Investors will spend the next year deciding which one was right, and the companies will spend it raising money on the assumption that it was the first.
