The Under-the-Radar Valuation Battle Between OpenAI and Anthropic: One's Fundraising Headline Has Become the Other's IPO Bottom Line

In the current global capital market, every financial move by OpenAI and Anthropic is no longer an isolated corporate action, but a top-tier game of mutual reference and mutual anchoring.

At the end of September 2026, media reports including Bloomberg stated that after postponing its IPO, OpenAI plans to raise at least 300 billion USD with a pre-money valuation of approximately 1.4 trillion USD. Shortly after in early October, another source claimed that OpenAI provided investors with guidance on its annualized revenue run rate (commonly referred to as ARR by the media) by the end of the year; among them, OpenAI's own net-caliber figure was close to 500 billion USD by the end of September, and after being adjusted by media and investors to a caliber comparable to Anthropic's gross method, it is estimated that the figure could reach 700 billion USD or more by the end of the year.

This set of data not only defines OpenAI's own market position, but also instantly transmits huge pressure to its largest competitor, Anthropic. In mid-October 2026, Anthropic held an investor day in San Francisco, where participating investors discussed an IPO valuation of about 1.5 trillion USD, a roughly 25% downward revision from the previous market expectation of about 2 trillion USD. Meanwhile, its Q3 revenue was announced at 16.8 billion USD, a quarter-on-quarter increase of about 45%.

Piecing these fragments together, an extremely cold game theory structure emerges: the financing headline of one company is becoming the business bottom line of another.

Valuation Anchoring: Why is 1.5 Trillion Anthropic's Bottom Line?

In the logic of venture capital, absolute valuation often lacks a reference frame, and the pricing of top competitors is the best ruler.

OpenAI's financing at a private valuation of about 1.4 trillion USD has actually drawn a passing line for the entire asset class of "frontier large model laboratories". For Anthropic, this is an extremely embarrassing but unavoidable anchor point. If Anthropic chooses to go public at a valuation of 1.4 trillion or lower, the narrative it has shaped in the capital market over the past two years of "being able to stand as a rival to OpenAI" will instantly collapse.

Over the past two years, Anthropic has had impressive performances where it overtook OpenAI in terms of annualized revenue run rate. The rise of Claude 3.5 Sonnet and Claude Code once made the market believe that it has stronger explosive power in vertical fields such as commercial implementation and coding. If such a company, which is considered to lead in growth rate, ends up with a public market pricing lower than its competitor's private financing, investors will immediately vote with their feet.

Therefore, 1.5 trillion to 1.6 trillion USD has become the lower limit of IPO that Anthropic must firmly hold on to in investor discussions. This figure is not precisely calculated from Anthropic's own financial model, but to a large extent "forced out" by OpenAI's quotation. It barely maintains Anthropic's superficial dignity of "valuation not lower than OpenAI", but also turns this IPO into a do-or-die battle with no retreat.

The Rashomon of ARR Caliber: How 700 Billion Forced Out the Trillion-Level Narrative?

If valuation is the facade, then the annualized revenue run rate is the foundation that supports the valuation. On this indicator, the two companies are playing an extremely exquisite "accounting Rashomon".

One point needs to be clarified first: OpenAI's "700 billion USD" is not its official guidance based on its own net caliber, but an adjusted estimate made by the media and investors to align with Anthropic's gross method caliber. OpenAI's own net caliber figure was close to 500 billion USD by the end of September 2026. Prior to that, the annualized run rates disclosed by the two companies according to their respective calibers were roughly: OpenAI about 500 billion USD (net), Anthropic more than 650 billion USD (gross, data as of the end of July 2026; simple annualization based on the single-quarter revenue of 16.8 billion USD in Q3 is about 670 billion USD).

Looking at the numbers alone, Anthropic seems to have the upper hand. But the problem lies in the statistical caliber.

According to industry analysis, Anthropic's revenue calculation method includes the total gross sales generated through channels such as Amazon AWS and Google Cloud, and the share of cloud vendors is listed as expenses; while OpenAI is more conservative, only counting the net income shared from partners such as Microsoft. This means that under the same actual business scale, Anthropic's book revenue figure is naturally "fatter", while OpenAI's figure is naturally "thinner". This difference is based on accounting standards - as the principal of the transaction, Anthropic is responsible for pricing and delivery, so it includes all channel sales in full, which is not a mere digital game.

When the media used the 500 billion of the "net caliber" to deduce "700 billion after adjustment", it actually announced to the market that OpenAI's real net income scale is about to step onto a new level. This constitutes a fatal reverse pressure on Anthropic.

The following is an analytical deduction, not a confirmed fact: If Anthropic wants to provide performance guidance in its prospectus or roadshow, in order to prove to investors that "our real scale and growth rate are still leading", its revenue guidance based on the gross method caliber may need to be raised to the level of hundreds of billions of USD - only figures of this magnitude can, after excluding channel shares and adjusting for the difference between gross/net calibers, overshadow OpenAI's adjusted 700 billion in narrative. It should be noted that there is no evidence in public reports that Anthropic has officially given a "1 trillion USD ARR guidance"; analysts once estimated that its year-end annualized run rate could reach about 1.1 trillion USD, but this is a third-party estimate, not corporate guidance.

This is the direct embodiment of game theory in commercial reality: you shout 500 billion in net caliber and 700 billion after adjustment, and I must use a trillion-level narrative in gross caliber to hedge. The annualized revenue run rate, which was originally an indicator to measure business health, has now become a leverage for the game between the two sides, and the expansion speed of figures far exceeds the delivery speed of real commercial value.

The Cold Eye of the Public Market and the Carnival of the Private Market

This trillion-level valuation hidden war takes place under an extremely fragmented macro background.

In the private market, OpenAI and Anthropic are still sought after by top capitals. According to public reports, sovereign funds and venture capital institutions such as Thrive Capital, SoftBank, and MGX are still willing to issue huge checks to model companies that have not yet achieved stable profitability. Figures of 1.4 trillion and 1.5 trillion are unimaginable in any traditional industry.

But in the public market, investors' fanaticism for AI is rapidly ebbing. The recent sharp correction of tech stocks is largely due to the underlying debt pressure of computing power infrastructure. According to market observations, the rising debt scale of computing power leasing companies such as CoreWeave and the widening CDS spreads expose the vulnerability of high-leverage expansion; Microsoft and Meta have also begun to adjust the pace of capital expenditure, and the market's doubts about "when AI investment can be converted into real profits" are growing louder and louder.

At this critical juncture, Anthropic is pushed to the front line of the IPO. It not only has to face the high valuation anchor set by OpenAI, but also face the extremely critical scrutiny of the public market. The quarter-on-quarter revenue growth rate in Q3 slowed down from more than doubling in the previous quarter to about 45%, which has already disappointed some investors. If under this growth rate, it still has to carry the valuation of 1.5 trillion USD and the trillion-level revenue narrative, Anthropic's IPO roadshow is destined to be full of gunpowder.

Anthropic's Hidden Cards and Structural Weaknesses

In this game, Anthropic is not without hidden cards.

Its moat in B-end enterprise services and AI programming fields is still deep. Claude's advantages in complex logical reasoning and long text processing give it high customer stickiness in professional service fields such as finance and law. The call volume and customer retention rate of its enterprise API are the core assets supporting its high valuation. In addition, the strategic investment and computing power binding from the two cloud giants Amazon and Google provide it with a relatively stable "ammunition depot".

But its weaknesses are equally obvious. The first is the heavy pressure of computing power cost. According to its prospectus disclosure, Anthropic has committed to spending about 518 billion USD in the next ten years on contractual obligations related to cloud computing, computing power and infrastructure, of which about 80% are non-cancellable, with an average annual amount of about 51.8 billion USD. As its revenue moves towards the trillion mark, the high reasoning cost will continue to erode profits, making it difficult to achieve large-scale positive cash flow in the short term.

The second is the dependence of customer structure. A large part of Anthropic's revenue comes from the channel distribution of Amazon AWS and Google Cloud; the prospectus shows that about 47% of sales in 2025 flowed through these two channels, and about 16 cents were taken away for every 1 USD. Although this deep binding brings revenue, it also means that it lacks sufficient say in channel pricing and profit sharing. When the gross caliber is stripped away, the real net retained profit may be far lower than the outside world's expectation.

Who Can Laugh Last?

The cruelest part of this game is that the bullets have already been fired, and no one can easily leave the table.

OpenAI postponed its IPO and chose to continue to maintain popularity with sky-high financing and revenue guidance in the private market, which is actually a strategy of "exchanging time for space". It uses private market valuation to put a heavy yoke on Anthropic, which is about to go public.

And Anthropic must prove in the public market that it is worth 1.5 trillion USD. If its IPO performance falls short of expectations, it will not only be its own defeat, but also break the foundation of OpenAI's 1.4 trillion valuation in reverse. Conversely, if Anthropic is successfully listed and performs strongly, it will become the anchor of the entire AI sector.

At the trillion-level card table, there is no permanent winner, only rising chips. One company's financing headline has become the other's IPO bottom line; and this bottom line will eventually be paid by millions of investors in the public market. This grand narrative about artificial general intelligence is evolving into an extremely sophisticated digital game. Whoever can stabilize the position in the maneuver of caliber and growth rate can truly get the ticket to the next era.

This article is from the WeChat official account "Tech Business", author: Tech Business, published with authorization from 36Kr.