A $2 trillion number landed, and everyone reached for a calculator

Here's the deal: on August 13, the Financial Times reported that Anthropic's investors expect the company to go public in October at a valuation of $2 trillion or higher. If it lands, that's the largest stock market debut ever — bigger than SpaceX's $1.77 trillion listing from just two months ago.

To get a feel for the size: Anthropic's last recognized private valuation was $965 billion, set in its Series H round in May. Three months later, investors want to more than double it. At $2 trillion, the company would be worth more than Amazon. A company barely five years old.

Then, a day later on August 14, Fortune did the arithmetic. Apply the Nasdaq 100's average multiples — 34x trailing earnings, 25x forward — and the annual profit Anthropic needs to justify $2 trillion comes to $59–79 billion. Amazon's net income in its most recent fiscal year was $77.7 billion. Which means Anthropic would have to be earning roughly what Amazon earns, on listing day, for the number to be "normal."

It isn't earning that. This story is about how wide that gap is, and why investors are calling the number anyway.

What Anthropic has actually become

The origin story — founded in 2021 by people who left OpenAI — barely matters at this scale anymore. In 2026 alone, venture firms, sovereign wealth funds, and institutional investors have poured close to $100 billion into Anthropic. As private capital directed at a single company, that's without precedent.

The revenue curve goes some way toward explaining the money. Annualized run-rate was roughly $9 billion at the end of 2025. By mid-May 2026 it was $47 billion. Salesforce CEO Marc Benioff has publicly put the figure at $74.1 billion. Investors expect $100–120 billion by year-end — more than ten times where the year started.

Drop to quarterly numbers and the shape is the same. Per estimates Fortune cites, Q2 2026 revenue was around $10.9 billion, more than double the prior quarter. That's also the quarter in which the company reportedly came close to its first operating profit, though what happened at the net income line isn't confirmed.

The comparison worth holding onto: OpenAI's annualized run-rate over the same stretch is estimated at $41.3 billion. So the two companies sit in a shape that rhymes with an earlier generation of cloud. One has overwhelming consumer brand strength (ChatGPT); the other is dug in deeper on the enterprise side. Anthropic's revenue mix skewing toward enterprise and API is being used as a strength in the listing narrative — corporate contracts are more predictable and churn less than individual subscriptions.

The mechanics are already moving. The target is a Nasdaq listing, with Goldman Sachs, JPMorgan, and Morgan Stanley leading an offering expected to raise more than $60 billion. OpenAI's own IPO is expected in 2027, which means Anthropic gets graded by the public market first.

So is $2 trillion a real number?

Let's line up everything on the table.

Item Figure
Target IPO valuation $2 trillion or more (October 2026)
Last private valuation $965 billion (Series H, May 2026)
Previous record IPO SpaceX, $1.77 trillion (June 2026)
Private capital raised in 2026 ~$100 billion
Run-rate trajectory $9B (end 2025) → $47B (May 2026) → $74.1B (estimate)
Year-end run-rate target $100–120 billion (investor expectation)
Q2 2026 revenue ~$10.9 billion (more than 2x prior quarter)
Annual profit needed for $2T $59–79 billion
Amazon net income, latest FY $77.7 billion
Expected offering size $60 billion+
Lead banks Goldman Sachs, JPMorgan, Morgan Stanley

Read down the column and the revenue growth is beyond dispute. The problem is the last few rows: the distance between the profit required and the profit that exists.

Two arguments try to close that distance. The first is that growth justifies the multiple. Nasdaq 100 averages describe mature companies, and applying them to something growing 10x a year is unfair on its face. High-growth companies conventionally get valued on revenue multiples instead. At a $120 billion year-end run-rate, $2 trillion is about 17x revenue — steep, but not unheard of for high-growth software.

The second is that profits are imminent. Avery Marquez of Renaissance Capital put it this way: "Reaching near operating profitability will at least be something that in my mind makes this very large valuation maybe not seem so crazy." In other words, you're not buying today's earnings — you're buying the size of earnings at the moment they start.

The counterargument is substantial. The biggest piece is compute cost. Anthropic is set to spend roughly $19 billion on compute this year, and that's a cost that scales with revenue rather than staying flat. The classic software advantage — revenue grows, cost of goods barely moves — doesn't hold here. Every model response burns GPU time.

Evan Schlossman of Neostellar Capital flagged compute access structure as the thing investors need to examine: whether it's owned or leased is a core valuation variable. Anthropic holds compute partnerships with Amazon, Google, Broadcom, and SpaceX, and the terms of those agreements determine future margins. If you're renting someone else's infrastructure, your partner's leverage grows right alongside your revenue.

There's a competitive factor stacked on top. Low-cost Chinese models are pricing aggressively at comparable capability, and Grok 4.6, released August 12, serves a benchmark score two points off Claude Opus 5 at a quarter of the price. Growing revenue tenfold while expanding margins, in a market where the quality premium is compressing, is not an easy combination.

Who gains from this listing

Anthropic gains compute capital, above all. Raising $60 billion-plus secures years of training and inference spend. Given that frontier lab competition is, at bottom, a capital-raising competition, whoever reaches the public markets first gains a structural edge. Private markets have a ceiling; public markets are far deeper.

Amazon has already booked an accounting win. $53.4 billion of Amazon's Q2 2026 net income came from gains on its Anthropic stake. A listing marks that gain to an actual market price, turning it into something realizable. It cuts both ways, though — once Anthropic trades, Amazon's earnings inherit its volatility.

Early investors and employees get liquidity. In a company that has absorbed close to $100 billion in private capital, a listing is the only real liquidity event available. Lockups mean actual selling is a story for a few months later.

OpenAI sits in an awkward spot. If Anthropic lists first and trades well, the bar for OpenAI's 2027 listing rises. If Anthropic struggles post-listing, valuation expectations across AI come down. Either way it isn't a spectator.

IPO buyers face the hardest call. Beautiful revenue curve, no profits yet, and a multiple that can't be explained with mature-company arithmetic. The historical pattern for that profile is extreme volatility over the first six to twelve months.

We've had this valuation argument before — the results split

There's plenty of precedent for how markets price enormous revenue growth with no profits attached.

The archetypal success is Amazon. It went public in 1997 and spent years not making money, fielding "when will you earn something" at every turn. Instead of profits it bought infrastructure and market share, and eventually extracted earnings from a second business entirely — cloud. Anthropic's narrative leans on that model: everything goes into compute now, and compute becomes the moat. Fortune's decision to hold up Amazon's $77.7 billion net income as the benchmark was aimed squarely at that story.

The failure pattern is just as clear. Many high-growth software companies that listed around 2021 traded at 20–40x revenue, then watched those multiples fall by four-fifths when rates rose and growth decelerated. The lesson is that a revenue multiple is a function of confidence in the growth rate. Multiples collapse on small growth misses, so the valuation drawdown runs far deeper than the revenue shortfall.

One more recent precedent worth holding: SpaceX. It listed at $1.77 trillion two months ago, and promptly used its own shares as currency to close a $60 billion acquisition (Cursor, effective August 14). A listing isn't only a fundraise — it's also a printing press for acquisition currency. Anthropic could easily do the same, and reports of acquisition talks in the $6 billion range have been circulating around it recently.

How the rivals counter

OpenAI counters with revenue velocity. Bloomberg confirmed on August 14 that its annualized run-rate topped $40 billion, roughly double where it ended 2025. Its consumer brand recognition remains unmatched. But with a listing expected in 2027, it has to chase whatever benchmark Anthropic sets in the public market in the meantime.

Google is exempt from the argument entirely. It's already public, and it trains and serves on its own TPUs, which puts its compute cost structure in a different category. It proved scale on August 11 by announcing Gemini crossed a billion monthly users. The most uncomfortable question in Anthropic's IPO story is exactly this: Google makes its compute, you rent yours.

Meta is playing a different board with open weights. Zuckerberg's 6,500-word essay on August 10 justified open model distribution with the argument that superintelligence shouldn't concentrate in a few hands. If that strategy works, the price premium on frontier models comes under pressure by design.

The cheap-model camp is already answering with price sheets. Grok 4.6, plus a set of Chinese open-weight models, are posting top-tier benchmarks at much lower prices. Justifying $2 trillion requires revenue to keep compounding — and in a market with intensifying price competition, that means shipping far more volume.

So what actually changes

If you're an individual Claude user, nothing today. But public companies have to prove numbers every quarter, so expect plan structures and usage limits to be adjusted more often than they are now. Product policy at listed companies generally moves in the direction of margin.

If your company runs on Claude, there's an upside. A public company files financials, which makes vendor risk assessment dramatically easier. "Will this company exist in three years" becomes a question with quarterly evidence attached. For procurement, that's a real benefit.

If you're an investor, a few line items matter more than the headline. Whether compute is owned or leased and on what terms; what share of revenue sits in enterprise contracts and how long those contracts run; and whether gross margin improves as revenue scales. Those three are the actual substance of the $2 trillion argument.

For the industry, this listing is the first public-market grade. Until now, frontier lab value has been set by a small number of investors in private rounds. From October, a market prices it every day. Where that price goes determines how much capital flows into this industry over the next several years. Anthropic's IPO isn't only Anthropic's event.

🥄 Three Things You're Probably Wondering

— Is $2 trillion locked in? No. It's an investor expectation; the actual offering price depends on October conditions and demand. What's out there is a target and an observation, not a decision. Valuations getting trimmed on the way from private to public is routine.

— Does this mean we're in an AI bubble? Too early to call flatly. Forbes ran a piece the same day headlined around exactly that question, which tells you the industry is split. Revenue really did grow tenfold, and the cost of producing that revenue really is enormous. Both things are true simultaneously.

— Can I buy the stock? In principle, once it lists. It's a Nasdaq listing, so non-US investors need international brokerage access, and buying in the open market after listing will be the realistic path rather than allocation in the offering. Worth keeping in mind that large IPOs have historically been very volatile in their early months.

References

Numbers and criteria are as of announcement and may change. Investment calls are yours to make!