Nobody Has Read Anthropic's IPO Filing. Wall Street Is Already Pricing It at $2 Trillion.

The S-1 went in on June 1. It is confidential, the share count is not set, the price is not set, and no financials have been disclosed. Here is what is actually known, what is estimated, and the one precedent from this year that tells you what happens if you buy on day one.

HeyTheo Research · Tuesday, September 29, 2026

Quick Read

• There is no fine print yet. Anthropic's draft S-1 was submitted confidentially on June 1. The company's own notice says the number of shares and the price have not been set, and no financial statements are public.

• Revenue growth is the real story. The annualised run rate went from $9 billion at the end of 2025 to $65 billion by the end of July 2026.

• It out-earns its more famous rival. OpenAI's most recent reported quarterly revenue annualises near $23 billion, roughly a third of Anthropic's July run rate, and OpenAI has said it is not listing this year.

• The $2 trillion figure is a forecast on a forecast. It leans on 2028 revenue of $190 billion to $200 billion, which works out to about 31 times the current run rate.

• This year already ran the experiment. SpaceX priced at $135 in June, closed its first day at $161, and trades below that now. Allocation made money. The pop did not.

A company that did not exist eight years ago is being discussed at a valuation larger than the GDP of Italy.

That would be remarkable on its own. What makes it stranger is that almost nobody arguing about the number has seen the document it is supposed to be based on.

Anthropic submitted a draft registration statement on Form S-1 to the SEC on June 1, 2026. It was submitted confidentially, which is a normal and legal route for large private companies. The company's public notice about it was four paragraphs long and said, in plain terms, that the number of shares to be offered and the price have not been set, and that any offering depends on market conditions.

That is the entire public record of the filing. There is no revenue table, no cost line, no risk factors section, no share structure, no lock-up terms. The fine print does not exist in public yet, and it will not until the company files an amended registration statement, which must come at least fifteen days before any roadshow.

So the honest starting point for an investor is this: everything below is either a company disclosure from a funding round, a reported figure from journalists with sources, or a projection. We have labelled which is which, because the gap between those three categories is where people lose money on new listings.

What is actually known

Source: Anthropic investor notice (June 1, 2026), Bloomberg, CNBC, Reuters. Accessed Sep 29, 2026.

The revenue trajectory is the part that deserves more attention than the valuation headline, because it is the only part with a clean series behind it.

End of 2025: $9 billion annualised. February 2026: $14 billion. March: $19 billion. April: $30 billion. Mid-May: $47 billion. End of July: $65 billion.

That is a business adding roughly $9 billion of annualised revenue a month through the middle of this year. There is no comfortable precedent for it in enterprise software. The closest reference points are not software companies at all.

The comparison that surprises people

Ask most people which AI company is bigger and they will say OpenAI. It has the consumer brand, the app on everyone's phone, and the name that became a verb.

On revenue, the picture is different.

Source: Bloomberg, CNBC, Forbes, company statements. Run-rate definitions differ between companies and are not directly comparable.

That last line in the source note is not a formality. Run rate is not an accounting standard. Companies choose what to annualise and from which month. Two firms can both report a run rate honestly and still not be measuring the same thing. Treat the gap as directionally real and precisely unreliable.

Two structural points sit behind the revenue difference. Anthropic's business skews toward enterprise and developer contracts rather than consumer subscriptions, which produces larger, stickier, and more concentrated revenue. And it has two hyperscaler shareholders, Amazon at roughly 21% and Alphabet at roughly 15%, who are also infrastructure suppliers and distribution channels.

That second point cuts both ways, and it is the part the eventual risk factors section will have to address. A shareholder who is also your largest supplier and a major route to your customers is a strategic advantage and a concentration risk in the same paragraph.

Where the $2 trillion comes from

A number that large deserves to be taken apart rather than quoted.

The reported reasoning runs roughly like this. Assume Anthropic reaches $190 billion to $200 billion of annual revenue in 2028. Apply a multiple of about ten times that forward revenue. That produces something near $2 trillion.

Two things follow from writing it out.

First, at $2 trillion the company would be trading at about 31 times its current run rate. That is not a stretch for a company growing this fast, and it is not obviously wrong. It is simply a claim about 2028 wearing the clothes of a claim about today.

Second, the entire structure rests on revenue roughly tripling from here within about two years, while the compute to serve it arrives on schedule. Anthropic's own framing of its projections, as reported, is a construction schedule tied to contracted compute rather than a demand forecast. That is a more disciplined way to think about it than most AI projections, and it swaps one risk for another. Demand risk becomes delivery risk, and delivery risk in this industry now means power, permits and data centre timelines, which is exactly what tripped Oracle's New Mexico campus and which we wrote about in what AI's buildout actually costs.

The precedent nobody should ignore

You do not have to speculate about what happens when a company of this scale lists, because 2026 already ran the test.

SpaceX priced its IPO at $135 and began trading on June 12. It closed its first day at $161, up 19%, in a debut widely described as a record. As of the September 25 close it traded at $148.36.

Work through what that means for two different investors.

Somebody who received an allocation at $135 is up about 9.9%. Somebody who bought at the first-day close of $161, which is what most retail investors can actually do, is down about 7.9%. Same company, same period, opposite outcomes, and the only difference is which side of the pop they were on.

That is the single most useful thing to carry into any large listing this year, including this one. The pop is not the opportunity. The pop is frequently the exit for people who got in earlier.

What would actually change the picture

Three documents, in order.

The amended registration statement is the first, and it is the one that turns this from a story into an analysable security. It will carry audited financials, the share structure, the risk factors, the use of proceeds and the lock-up terms. Until it lands, any valuation view is a view about a forecast.

The pricing range comes second, typically in the roadshow. The gap between the range and the reported $2 trillion chatter will tell you how much of the chatter was banker anchoring.

The lock-up expiry is third and is the one retail investors most often forget. Insiders and early investors in a company that re-rated from $623 million to $965 billion have a great deal of paper profit. What they do when they are permitted to sell is a known future supply event with a known date.

The rules HeyTheo tracks

• Basket: group the listed AI infrastructure and hyperscaler names and watch them as one unit around any large AI listing, because a mega-IPO pulls capital from the same pool.

• Triggers: flag the amended S-1 filing, the pricing range announcement, the first day of trading and the lock-up expiry. Those four dates carry more information than any analyst note before them.

• Money flow: watch whether big money rotates into the sector ahead of a listing or funds the new position by selling existing AI exposure.

• Ask Theo: pull what is disclosed versus what is reported on any pre-IPO name, and the post-listing record of comparable large debuts.

• Check the rule behind any trigger before acting on it. You trade through your own broker; HeyTheo helps you decide.

The Allocation Question

Strip away the number and one question decides how this goes for an individual investor: are you buying at the price the insiders set, or the price the first day produces?

Almost no retail investor gets the first. Nearly all of them get the second. SpaceX already showed what the distance between those two costs, and it did so in a debut everyone called a triumph.

Nothing here argues Anthropic is a bad business. The revenue series is extraordinary, the first positive adjusted operating income arrived earlier than most expected, and its two largest outside shareholders are the companies best placed to judge whether the technology works.

It argues something narrower. A valuation being discussed before the financials are public is a negotiating position, not an analysis. The analysis becomes possible when the amended filing lands, and the sensible thing to do between now and then is read the run-rate series, note which numbers are disclosed and which are reported, and wait for the document.

There will be plenty of time. There always is.

FAQs

What is a confidential S-1 and why is it allowed?

A draft registration statement submitted to the SEC without being published. Companies use it to begin the review process and settle comments with regulators before revealing financials to competitors. It became widely available to large companies under rules extended in 2017. The filing becomes public when the company files an amended registration statement, which must happen at least fifteen days before it markets the offering to investors.

Does the confidential filing mean an IPO is definitely happening?

No. It gives the company the option. Anthropic's own notice says any offering depends on market conditions and other factors, and companies do withdraw. Prediction markets have put the odds of a listing this November near 57%, which is a market view rather than a company statement.

Why is Anthropic's revenue higher than OpenAI's if OpenAI is better known?

The mix differs. Anthropic's revenue skews toward enterprise and developer contracts, which carry larger contract values than consumer subscriptions. Consumer reach and revenue are different things. One caution: run rate is not a defined accounting term, so the two companies may not be measuring the same way.

What is the difference between a run rate and actual revenue?

A run rate annualises a recent period, usually a month, by multiplying it up. It is a snapshot travelling at speed. Real annual revenue is what a full audited year produced. For a fast-growing company the run rate will always look larger than the trailing year, which is why the amended filing's audited figures matter more than any run-rate headline.

Can an ordinary investor buy at the IPO price?

Rarely. Allocation at the offer price typically goes to institutions and to selected brokerage clients. Most individuals buy once trading opens, at whatever the market sets. This year's largest listing showed the difference clearly: allocation at $135 is up about 9.9%, while the first-day closing price of $161 is down about 7.9% as of the September 25 close.

Sources

Sources: Anthropic investor notice, Bloomberg, CNBC, Reuters, Forbes, NPR, StockAnalysis.

Disclaimer

Disclaimer: HeyTheo is a research and education platform, not an investment adviser or broker-dealer. Nothing here is advice to buy, sell, or hold any security. You trade through your own broker; HeyTheo helps you decide. Backtested results are hypothetical and do not guarantee future returns. References to governments, officials, or policies are for market context only and are not political endorsements. All investing involves risk, including loss of principal. Data is as of the dates noted.