Five years old. Zero public filings. A chatbot called Claude. And investors are now telling the Financial Times they want to take this company public at two trillion dollars, which would make it the largest initial public offering in the history of capital markets. The FT reported on August 13 that half a dozen Anthropic backers confirmed a $2 trillion target for an October listing, with one investor going further and calling $3 trillion "the incredibly low end." That is not the kind of thing a person says quietly about a company that did not exist before 2021.

Where the $2 Trillion Number Actually Comes From

The valuation is grounded in one thing: a revenue trajectory with no real precedent in enterprise software. When Anthropic filed confidentially for an IPO at a $965 billion valuation in June, that figure already seemed audacious for a private company. Two and a half months later, investors are modeling something more than double that. Their basis is the annualized revenue projection: between $100 billion and $120 billion by the close of 2026. Anthropic reported $47 billion in annualized revenue as recently as May. Getting from $47 billion to $100 billion in seven months would represent a run rate acceleration that has no direct comparison in the history of software companies.

The Wall Street Journal and CNBC both confirmed in May that Anthropic projected $10.9 billion in actual quarterly revenue for Q2 2026, which would have been the company's first-ever operating profit at roughly $559 million. Those projections were shared with investors as part of an ongoing fundraising process, and came with a significant caveat: profitability in Q2 was partly a function of discounted compute costs during the initial ramp-up of Anthropic's Colossus data center deal with SpaceX, costs that escalate through the second half of 2026. The honest framing is that Q2 operating profit was real but probably not durable, and investors pricing the company at 20 times forward revenue are betting on the growth line staying steep even as compute costs normalize.

The Math Behind the Multiples

At $100 billion in annualized revenue and a $2 trillion valuation, Anthropic trades at roughly 20 times sales. That is expensive by any conventional software benchmark but not entirely detached from reality for a company growing at this pace. For context, investors noted that Palantir and cloud provider Nebius have traded at approximately 55 times revenue this year among publicly listed AI-adjacent businesses. One investor told the FT that at 30 times revenue on $100 billion, the valuation reaches $3 trillion on its own arithmetic. What makes that framing interesting rather than just self-serving is that comparable public companies have traded at those multiples while growing far more slowly than Anthropic.

The harder question is what $100 billion in revenue actually means for a company whose single largest cost line is AI compute at hyperscale. Anthropic's compute obligations are expanding rapidly alongside its revenue. The May Q2 profit figure excluded stock-based compensation and benefited from a temporarily discounted infrastructure deal. A company spending $15 billion annually on compute at full rate, growing toward $120 billion in annualized revenue, carries a fundamentally different profit profile than the Q2 snapshot suggests. Fortune put the underlying challenge plainly: to sustain a $2 trillion valuation, Anthropic would need to produce Amazon-scale earnings from a business that has barely turned a corner on profitability.

The IPO Context Nobody is Pricing Carefully

The October timing matters because it arrives in a specific market context. SpaceX debuted at $1.77 trillion in June, briefly making its listing the largest in history. A $2 trillion Anthropic IPO would surpass that immediately, and the sheer size of the deal would require institutional buyers to absorb a company valued at more than twice the current market cap of Berkshire Hathaway at the moment of first trade. Bloomberg separately reported this week that Anthropic is also in talks to acquire the AI startup Decart AI for roughly $6 billion, a transaction that would add complexity to the IPO process just as the S-1 documentation is presumably nearing completion.

Beyond the mechanics, Anthropic's regulatory exposure is real and unresolved. The company hired former California Supreme Court Justice Tino Cuéllar as its first Chief Global Affairs Officer in August, specifically to manage what amounts to a simultaneous policy battle on multiple fronts: a standoff with the Pentagon over model access, ongoing EU regulatory scrutiny, and the emerging question of whether frontier AI companies will face mandatory safety certification requirements before a new model can be released. A company preparing a $2 trillion public offering while its most capable models are entangled in government approval processes is asking institutional investors to underwrite both the growth story and the regulatory risk in the same transaction.

What the Revenue Projections Don't Fully Explain

One detail in the investor commentary deserves more attention than it has received. Investors citing "800 percent annual growth" as the basis for a 30 times revenue multiple are conflating two different things: the growth rate over a period of explosive adoption, and the sustainable forward growth rate into a larger base. A company growing from $5 billion to $47 billion in annualized revenue in seventeen months is not doing so because its underlying product is compounding at that rate indefinitely. It is doing so because enterprise AI adoption crossed a threshold, because Anthropic's Claude Code product captured a developer category with unusual speed, and because a handful of very large compute lease agreements contributed to the revenue figure in ways that do not automatically recur.

This is precisely the kind of accounting opacity that inflated Big Tech earnings this past quarter, when paper gains on Anthropic and OpenAI stakes made S&P 500 earnings growth look 19 percentage points larger than underlying business performance actually justified. The same mechanism runs in reverse here: Anthropic's headline revenue trajectory looks cleaner than the underlying compute cost structure suggests it is.

What Happens If October Works, and If It Doesn't

If Anthropic prices at $2 trillion in October and trades well in its first weeks, it resets the entire valuation conversation for frontier AI. OpenAI, which has been developing its own IPO with Goldman Sachs and Morgan Stanley and was last valued at $852 billion, would effectively receive an independent market test of what public investors are willing to pay for this category. An Anthropic success at $2 trillion makes OpenAI's IPO roadshow dramatically easier to conduct at a meaningful premium to its last private valuation.

If Anthropic fails to hold its listing price, the reverse is also true. A stumble on day one, or a first earnings report that shows Q3 profitability deteriorated as compute costs normalized, would hand skeptics of AI valuations a landmark data point at the worst possible moment. The FT report noted that investors themselves acknowledged the valuation requires Anthropic to grow into a company with no close comparable. An investor quoted saying $3 trillion is "the incredibly low side" is either confident to the point of certainty or performing confidence for an audience. At the moment, those are difficult to tell apart, and the S-1, whenever it becomes public, will be the first document that actually has to answer for the difference.