Anthropic Files S-1 in Nasdaq Push, Valuation May Be Cut from $1.5 Trillion to $1.11–1.48 Trillion

[[**Anthropic files S-1 in dash to Nasdaq; prospectus reveals unstable profitability, nearly half of revenue must be shared with cloud channels, founders lock up 50.1% voting power, market valuation may be cut from $1.5–2 trillion to $1.11–1.48 trillion.**

**Key points**

**1. Quarterly profit turns positive but sustainability is doubtful**

In 2026Q2, GAAP operating profit turned positive at $974 million, and on an adjusted basis Q3 remained profitable; however, 80% of revenue comes from API calls, making it highly dependent on new model iterations, and take-or-pay compute bills will peak after 2027, so profit may turn negative again.

**2. Cloud channel revenue sharing lowers actual revenue**

In the first half of 2026, about 48% of revenue was settled through AWS and Google Cloud marketplaces; in 2025 channel fees accounted for 16% of cloud sales, with a net impact on overall revenue of about 7.5%, meaning an ARR-based valuation needs a 0.92x haircut.

**3. Founders' lock on control creates a governance discount**

Seven co-founders hold Class F shares through an LLC, locking up 50.1% of voting power; after the IPO, external shareholders' voting power is capped at 49.9%; referencing the 15%–20% discount seen in later-stage dual-class structures, a 15% governance discount is applied.

**4. With three gates stacked, valuation is clearly cut**

Profitability at half-open gets a 0.95x haircut, channel revenue sharing a 0.92x haircut, and governance a 0.85x discount; multiplying the three leaves only about 74% of the valuation weight, corresponding to $1.11–1.48 trillion; if profit turns negative, the pessimistic range is $1.05–1.4 trillion. Original 2026-10-27 20:44 Anthropic's $1.5 trillion valuation is also in doubt Zhang Beibei Quick Take

On the evening of October 26, Anthropic submitted an S-1 prospectus to the SEC, applying to list on Nasdaq under the ticker ANT.

For an AI company with $16.3 billion in half-year revenue and 1,300% year-over-year growth, Q2 single-quarter GAAP operating profit has already turned positive ($974 million), and Q3 remains profitable on an adjusted basis. The primary market had long given it a $2 trillion valuation expectation; however, amid the recent Q3 sequential growth slowdown from 144% to about 45%, some investors have lowered expectations to $1.5 trillion.

The number of shares offered, offering price and target valuation in the prospectus are still blank; the price will be filled in only after the roadshow bookbuilding is completed.

What the prospectus will really directly change is the information structure on which pricing is based. Variables that could be folded into a broad narrative during the private stage—such as nearly half of revenue being settled through cloud channels, and roughly $600 billion (based on disclosed contracts summed) in take-or-pay compute commitments—are now broken out item by item and placed in a public document carrying legal liability.

Institutional investors participating in the roadshow will reprice after these variables are laid out.

So how much is Anthropic actually worth?

Three questions, three price tiers

The market previously gave a private valuation of up to $1.5–2 trillion; after the prospectus lays its cards on the table, that price may no longer hold.

To judge what it is actually worth, one must first understand three questions: Is it really making money? Does all the money it makes belong to it? After buying, can you participate in decision-making?

Treat these three questions as three valuation gates; each gate passed opens a price tier, and each gate failed cuts the price:

(Chart production: Huxiu Miaotou)

If Anthropic were an ordinary company with stable profits, all the money in its pocket, and shareholders calling the shots, it would be worth $1.5–2 trillion; but it has passed only half of the three gates, so its true valuation may be closer to $1.11–1.48 trillion.

It is like buying a milk tea shop. If the owner told you the shop makes money steadily, all the money goes into your pocket, and you make the big decisions, you would be willing to pay 2 trillion; but what he actually says is that it only just became profitable in the past three months, the platform takes 20%, there are still ten years of tea-leaf bills owed, and I call the shots at the shop—then you would apply a discount and not pay that much.

Anthropic's 2026Q2 single-quarter GAAP operating profit turned positive at $974 million, and at the earlier investor day it disclosed that Q3 would remain profitable on an adjusted basis. This is the first publicly disclosed profit-turned-positive report card among large-model companies. A big part of why the primary market dared to give $1.5–2 trillion is this number.

But a single quarter of positive operating profit does not mean Anthropic already has the ability to be stably profitable.

Looking through the S-1 prospectus submitted on October 26, one can see that Anthropic's current profit improvement mainly benefits from revenue growing faster than compute costs. Whether this can continue still faces two tests.

(1) Revenue growth is highly dependent on new models.

The prospectus contains an easily overlooked structure: Anthropic's revenue consists of API calls and subscriptions, with API accounting for 80%; enterprise customers pay based on actual usage and use whichever model is stronger; subscription revenue, though more stable, accounts for a relatively small share and cannot support overall growth.

In this situation, the company must keep releasing new models like running a relay race. Because once iteration speed lags by half a beat, or a release falls short of expectations, API call volume will immediately migrate to competitors, and revenue may suddenly stall. This is also an important reason why more than 80% of its 2026Q3 revenue mainly comes from models released in the past 6 months.

In Q2 2026, quarterly GAAP operating profit turned positive at 974 million, and in Q3 it continued to be profitable on an adjusted basis—two consecutive quarters of making money. An important reason behind this: revenue growth (1368% year-over-year) temporarily outpaced the growth of locked-in compute costs (about 300%), a phased timing-window dividend.

But whether this can be sustained carries significant uncertainty.

(Chart production: Huxiu Miaotou)

Because compute is usually contracted 12–18 months in advance, compute costs such as lease fees and electricity are locked in first, while revenue can only be realized after models are released.

And starting in the second half of 2026, large compute contracts will successively enter their billing periods: Google Cloud in April, AWS in May, and Broadcom equipment leasing have all begun to incur charges. This money is not paid all at once. Google and AWS are minimum commitments that must be paid whether used or not; Nscale and Lambda bill upon delivery acceptance; Space XAI will not enter its monthly-fee peak until capacity is delivered at the end of the year.

In other words, the 8.945 billion in compute spending in the first half of 2026 is only the "down payment" on future bills in the hundreds of billions.

The real bulk may come after 2027.

What is more troublesome is the mismatch. Most of the aforementioned compute commitments are rigid take-or-pay expenditures, while 80% of the company's revenue comes from pay-as-you-go API calls, with no long-term locked-price contracts. The cost curve only goes up, while revenue fluctuates with model iterations and call volume. When the 2027 compute bills enter their monthly-payment peak and revenue growth slows, GAAP profit will most likely be squeezed and may even turn negative again.

(Chart production: Huxiu Miaotou)

Therefore, the result of the first gate: half open. It really is making money, but not yet stably.

So the first gate is not fully passed, and the valuation, on a base of 1.5–2 trillion, must first take a cut.

If a 5% discount is applied, the valuation may be revised down to 1.425–1.9 trillion USD; but if profit turns negative in the future and the gate closes, then the discount must be revised further downward; if a 10% discount is applied, the valuation may be revised down to 1.35–1.8 trillion USD.

It is like a milk tea shop selling 100,000 cups this month with 300,000 in gross sales—sounds impressive; but 150,000 of that is sold through delivery platforms, which take a commission; in addition, to lock in low-priced tea leaves, the owner signed a rigid contract three years ago requiring 50,000 in payments every month until 2030. So the 300,000 in gross sales and the money the owner actually pockets are completely different things.

Anthropic's 16.3 billion in revenue is the same story.

The prospectus discloses that in 2025 its sales completed through AWS and the Google Cloud marketplace were about 2.16 billion USD, accounting for 47% of full-year revenue; in the first half of 2026 this proportion rose further to about 48%. But Anthropic records revenue on a gross basis, booking the channel share owed to cloud vendors as sales expense.

In this case, for market investors who value the company on ARR (annualized revenue), the previous pricing was clearly too high.

Therefore, the result of the second gate: not passed.

For reference, this channel fee was 351 million USD in 2025, accounting for 16% of the 2.16 billion USD in cloud vendor sales for the period. In this case, the net impact on overall revenue is ≈ 16% × 47% (the cloud vendor sales share for the period) ≈ 7.5%, so revenue and ARR need to be discounted by about 8%.

Correspondingly, the valuation may be further revised down on the basis of the first cut: ① first gate half open: 1.31–1.75 trillion USD; ② in the pessimistic case, first gate closed: 1.24–1.65 trillion USD.

You spent 1 million to buy 30% of a milk tea shop; logically, if the shop replaces refrigerators, raises wages, or even sells a branch, you should have a vote. But the owner says: "I alone decide everything in this shop; you just collect dividends and don't stick your nose into anything else." Then you certainly would not be willing to pay 1 million; you might only be willing to pay 800,000, because say also has a price.

Anthropic is even more extreme: seven co-founders hold one share of Class F stock through an LLC (limited liability company), locking in 50.1% of voting rights, while you as a shareholder do not even have the voting right to remove the CEO. Because with the shares issued in the IPO, voting rights are diluted to a ceiling of 49.9%, and a majority can never be formed.

Research by scholars such as Cremers on the valuation life cycle of dual-class companies found that there is a valuation premium in the early stage after listing, but a discount gradually emerges after 6–9 years and later stabilizes at the 15%–20% level;

The CFA Institute's corporate governance white paper lists non-veto minority shares as a second-tier discount factor but does not provide a unified quantitative standard for the discount.

The OECD Principles of Corporate Governance explicitly treat dual-class shares as a valuation adjustment item.

We take 15% as the discount range for the third layer.

In this case, after another cut on the basis of the second gate, the valuation may be further revised down to: ① first gate half open: 1.11–1.48 trillion USD; ② in the pessimistic case, first gate closed: 1.05–1.4 trillion USD.

(1) Is it really making money? Yes, but only phase by phase; whether it can be sustained carries significant uncertainty.

(2) Does all the money it earns belong to it? No; after deducting the cloud vendors' channel share, an 8% discount is needed.

(3) After buying, can you participate in decision-making? No, so apply another 15% discount.

Multiplying the three answers together gives Anthropic's possible true pricing.

Starting from the 1.5–2 trillion when all gates are passed, first cut the money-printer premium (half open −5%), then remove the channel share (−8%), and finally deduct the governance discount (−15%); after multiplying the three cuts, only about 74% of the valuation weight remains.

So, when the three realities of unsustainable profitability, removal of channel share, and governance discount are placed on the table, the valuation will most likely be revised down to 1.11–1.48 trillion USD.

This may not be the endgame.

The prospectus itself warns that revenue growth may decline and losses may continue. If adjusted profit turns negative again in the future, this pricing may continue to be revalued downward to 1.05–1.4 trillion USD.

Therefore, if the market still prices it on an ARR basis, it will most likely systematically overvalue this company, and IPO subscription requires caution.

Article title: Anthropic's 1.5 trillion USD valuation is also in doubt

Has published 246 articles on Huxiu. This content may not be reproduced without permission. For authorization, please contact hezuo@huxiu.com. Everyone who is changing and wants to change the world is on the Huxiu APP[ ]