The race to go public first is about to end. Anthropic, the company behind Claude, is expected to hit Wall Street within weeks in a listing that could erase the record SpaceX set in June. But before anyone starts measuring the biggest debut in market history, there is a number that has to be believed first: roughly 200 billion dollars in 2028 revenue. That forecast is the load-bearing wall under the entire valuation. And here is the uncomfortable part — nobody outside the company has actually seen it.

What the Anthropic IPO Actually Is

Anthropic confidentially submitted its draft Form S-1 to the SEC on June 1, 2026. That is the only step the company has directly confirmed. The October listing window and the valuation talk come from people connected to the deal, not from a public prospectus. But the direction is unmistakable: Euronews reported on August 26, citing AFP, that the company is expected to go public within weeks, and Bloomberg reports it aims to raise more than the 75 billion SpaceX secured in June.

The financial foundation is real, if young. On May 28, Anthropic closed a 65 billion Series H round at a post-money valuation of 965 billion, led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital. At the time, the company said its annualized revenue had crossed 47 billion. Three months later, coverage puts the run rate above 65 billion — driven almost entirely by enterprise adoption and Claude Code, the developer tool that made coding one of the few AI categories customers actually pay serious money for. The company has roughly 5,000 employees and was founded in 2021 by former OpenAI executives Dario and Daniela Amodei, who left believing their old employer was not taking AI risk seriously enough.

The 2 Trillion Question: What Investors Are Really Pricing

Here is where the story stops being about a company and becomes about a bet. The Financial Times reported on August 13 that six Anthropic investors expect the listing to value the company above 2 trillion dollars. Their case rests on an extraordinary forecast: annualized revenue reaching between 100 and 120 billion by the end of 2026. At the midpoint, a 2 trillion valuation works out to roughly 18 times annualized revenue. One optimistic investor reportedly applied a 30-times multiple to an 800 percent growth estimate and came out near 3 trillion.

Reuters then added the detail that actually matters. Two people familiar with Anthropic's financials told the wire that the IPO valuation hinges on a 2028 revenue projection of roughly 190 to 200 billion — about four times the May run rate. For context on what that multiple means, LSEG data cited by US News shows Palantir trading near 53 times expected 2026 revenue, with SpaceX and Cloudflare around 41.6 times. Anthropic's backers are effectively asking the market to value a five-year-old company like the most expensive software businesses on the planet, before any audited financial statements are public.

The SpaceX Benchmark — and Why It Cuts Both Ways

SpaceX is the obvious yardstick. Its June debut raised more than 75 billion at a 135 dollar offering price, after absorbing Elon Musk's xAI, and briefly looked like it would fly into orbit. Then it came back to earth. The shares now sit near the offering price. That is a useful reminder: even a genuinely historic AI-adjacent listing can produce a round trip for late buyers.

Anthropic plans to pitch an even bigger tent. The Wall Street Journal reports the company will tell investors its total addressable market exceeds 30 trillion dollars — the value of all work AI could eventually automate — surpassing the 28.5 trillion estimate SpaceX used. A 30 trillion TAM is not revenue, and the WSJ is careful to say so. It is a narrative device. It exists to justify infrastructure spending and a projected 200 billion revenue target that, again, has not been filed anywhere the public can read.

The Political Elephant: A Company at War With Its Own Government

There is a risk factor in this story that most coverage mentions in passing and no roadshow can avoid. In March, the Trump administration terminated government contracts with Anthropic and designated the company a supply-chain risk after it refused to give the military unfettered access to its AI models. Anthropic called the move unconstitutional retaliation and sued the federal government. The White House has also objected to Dario Amodei's repeated warnings about AI risk and his calls to regulate the technology the way we regulate airlines or banks.

That dispute could take years to resolve, and it lands at an awkward moment. A company asking public investors for a record valuation is simultaneously telling the largest buyer on earth, its own government, that it will not build what the military asks for. That is a principled position. It is also a disclosed risk factor that no amount of TAM math makes smaller.

The Questions Nobody Has Answered Yet

The confidential filing means the draft prospectus is invisible. Until the public S-1 lands, here is what investors cannot check: where the 200 billion in 2028 revenue actually comes from and how much of it is concentrated in a handful of enterprise contracts; what gross margin looks like after the compute bill, because every Claude request burns infrastructure and frontier inference is still expensive; how the Amazon and Google relationships get disclosed, since both companies are simultaneously investors, suppliers, distribution partners, and competitors; whether the run rate survives OpenAI, Google, and Microsoft all cutting prices on coding agents; and what the government dispute does to the federal and defense pipeline. The last question is the quiet one: OpenAI is reportedly considering waiting until 2027. When your biggest rival chooses to let you go first, it is worth asking what they see.

What This Means: The AI Economy Is About to Get a Price Tag

This is the first real public test of whether frontier AI is a business or a heavily subsidized science project. If Anthropic pops, capital floods into every AI name waiting behind it. If it sinks, the private valuations that have been compounding for three years get re-rated in a hurry. For anyone running infrastructure — the servers, the data centers, the power contracts this sector consumes — the stakes are concrete. The outcome of this listing decides how much compute gets bought next year, and at what price. It also sets the template for OpenAI's own debut, whenever that comes.

What Comes Next

Watch for the public S-1. That is the next major information event — audited revenue, actual losses, customer concentration, risk factors, and a real explanation of how the company makes money. The October window is plausible but not guaranteed; SEC review rounds and market conditions can move it. If the listing lands, it is the biggest debut on record. If it slips, the AI funding cycle gets its first genuine pause since the boom began. Either way, the 2 trillion story gets tested the moment the numbers become public. Until then, treat the forecast the way you would treat any unreleased benchmark: it is a claim, not a result.

— Allan Ali, Sylt.ing