Anthropic's Investors Priced It at $2 Trillion. Anthropic Hasn't.

TL;DR

The Financial Times reported early on August 13 that Anthropic shareholders are anticipating an October IPO at more than $2tn, a deal the paper says would eclipse SpaceX and become the largest IPO ever completed. Anthropic did not say that. Management has not set a target valuation, and the company has been in a quiet period since filing a confidential S-1 with the SEC on June 1. The $2tn is investors backing a price out of revenue that has not happened yet: $100bn to $120bn annualized by the end of 2026, against the $47bn run rate Anthropic actually disclosed in May.

What was actually reported

Three things in the FT piece are load-bearing, and only one of them is a fact about the present.

First, shareholders anticipate an October listing above $2tn. Second, investors interviewed by the paper estimate annualized revenue could reach $100bn to $120bn by the end of 2026, more than ten times its level at the start of the year. Third, one of those investors offered the arithmetic out loud: a company posting 800% annual growth should trade at at least 30x revenue, which on a $100bn base gets you to $3tn.

Note who is speaking. These are holders of a private position talking about what that position is worth. Anthropic itself is legally muzzled, having filed confidentially in June without disclosing the size or terms of the offering. It is an unusual kind of price discovery: the sellers can talk, the issuer cannot, and the number keeps going up.

The number is a forecast wearing a valuation's clothes

Here is the chain the $2tn hangs from, and every link past the first is an assumption.

Backing a valuation out of a forward multiple is like appraising a house from the extension the owner says they are going to build. The arithmetic is fine. The extension is the load-bearing part.

Which denominator you pick decides everything

Run the multiple against the numbers that actually exist and the picture changes hard. At $2tn on the $47bn run rate Anthropic disclosed in May, you are paying roughly 43x trailing revenue. At $2tn on the $120bn top end of the investor forecast, you are paying under 17x, which for a hypergrowth software business is not obviously insane.

Same price. Same company. The gap between those two numbers is entirely a bet on the next five months.

The secondary-market line matters too. Private trades have reportedly pushed the implied valuation to roughly $1.2tn since the May round closed at $965bn. That is real money changing hands at a real price, and it sits well short of $2tn.

What Anthropic has actually disclosed

Strip out the projections and the underlying business is still the fastest revenue ramp anyone has put on paper. Anthropic has published run-rate milestones alongside its own funding and partnership announcements, and they compound at a rate that makes the investor optimism at least legible.

Nine billion at the end of 2025. Fourteen by February. Thirty by April. Past $47bn in May, disclosed alongside the $65bn Series H that set the $965bn post-money and put Anthropic ahead of OpenAI on valuation for the first time. Nearly $100bn has been invested into the company across 2026 by the FT's count.

A lot of that curve is Claude Code, which is the part of this story you already live in. If you have been watching your own team's spend on agentic coding climb every quarter, you have been personally contributing to the denominator.

The record it would break, and the asterisk

SpaceX priced its IPO at $135 a share on June 11, raising about $75bn, the largest raise on record, at a valuation near $1.77tn. Shares closed their first session at $161, up 19%, briefly carrying the market cap above $2tn.

So read the "largest IPO ever" framing carefully. SpaceX set the record on capital raised. Anthropic at $2tn would top it on valuation, but the June filing was confidential and disclosed neither size nor terms, so nobody outside the deal knows how much stock is actually being sold. Two different records, one headline.

The other cautionary note is what happened after the confetti: SpaceX shares gave back ground in the weeks following the debut. A record first day is not a record year.

Why a builder should care

You do not have to care about equity markets to care about this. Three concrete consequences:

• Pricing becomes a quarterly line item. Once Anthropic reports to public shareholders, decisions about token prices, rate limits, and free tiers stop being product decisions in a vacuum and start being gross-margin decisions with an earnings call attached.
• You finally get real numbers. Today you are reasoning about Anthropic's viability from press-release run rates and leaked forecasts. An S-1 and 10-Qs mean audited revenue, real cost of revenue, and actual compute commitments. That is a genuine upgrade for anyone doing vendor risk on their stack.
• The $2tn assumes the moat holds. The forecast is a straight-line extrapolation of coding-agent demand. Cheaper open-weight Chinese models from the likes of Qwen and DeepSeek closing the gap is precisely the risk that eats the multiple, and it is the first thing skeptical investors reportedly raise.

The bear case is already on the record

Reporting two days earlier flagged three specific investor worries about this offering: cheaper Chinese systems compressing the frontier lead that was supposed to be six to eight months, friction with the Trump administration, and local pushback against building data centers in the US. None of those are exotic. All three are the sort of thing that turns a 30x multiple into a 15x multiple in a single bad quarter.

Meanwhile Anthropic's own homepage still leads with Opus 5 and a public benefit corporation mission statement. The trillion-dollar arithmetic is happening entirely in other people's spreadsheets.

Key Takeaways

• The FT reported on August 13 that Anthropic shareholders anticipate an October IPO above $2tn, which would eclipse SpaceX as the largest IPO ever completed.
• Anthropic has not set a target valuation and is in a quiet period after filing a confidential S-1 on June 1 that disclosed neither size nor terms.
• The $2tn rests on an investor forecast of $100bn to $120bn annualized revenue by end of 2026, versus the $47bn run rate Anthropic actually disclosed in May.
• At $2tn you pay about 43x the last disclosed revenue, or under 17x the forecast. Secondary trades currently imply roughly $1.2tn.
• Anthropic's published run rate went $9bn to $14bn to $30bn to past $47bn between December 2025 and May 2026, with Claude Code a major driver.
• For builders the real change is disclosure: post-IPO, Claude pricing and limits become decisions with an earnings call attached.

Sources: Financial Times, MarketScreener, Al Jazeera, Simon Willison, The Decoder, CNBC, CNBC