Eight Spots Where Anthropic's IPO Pitch and Its Numbers Diverge

Public-market investors are being asked to put Anthropic's worth at above $2 trillion. The maker of Claude AI recorded a $42 billion loss last year.

Both those numbers show up in Anthropic's IPO prospectus, the paperwork a company must file ahead of a public share sale. Reuters examined the filing on Monday; eight areas stand out where the presentation and the figures part ways.

Anthropic valuation: $2 trillion
Amazon valuation: $2.6 trillion

Anthropic revenue: $65 billion
Amazon revenue: $716 billion

Explain this gap.

— Limitless Finance (@trylimitlessfin) September 29, 2026

1. A $2 Trillion Valuation Over $4.6 Billion in Sales

Anthropic generated roughly $4.6 billion in revenue during 2025. At the $2 trillion level, that works out to about $435 paid for each $1 in sales.

Its annualized revenue was $65 billion as of July and may climb to $100 billion by the end of the year. Those figures, though, reflect a current run rate rather than a complete year of audited sales.

The valuation therefore hinges on growth continuing almost flawlessly. If momentum slips, the $2 trillion price becomes far more difficult to defend.

2. Operating Losses Reach $8 Billion

Operating losses came to $8.06 billion, compared with $2.98 billion in 2024.

The widely cited $42 billion loss overstates how much cash is actually being burned. Still, the core operations are billions in the red, and the deficit keeps widening.

The company is improving quickly, yet it still loses money on each dollar of revenue.

3. Profit Figure Omits the Cost of Building AI

For the second quarter, Anthropic posted positive adjusted operating income.

That metric does not count model training, revenue shared with partners, or stock-based compensation.

This is where the profit story becomes uncomfortable. Training the next generation of Claude models sits at the core of the business, yet the adjusted measure strips that expense out.

4. Compute Spending Exceeds Revenue

Anthropic put $7.33 billion into compute during 2025, against $4.6 billion in sales.

That is around $1.60 spent on computing for every $1 of revenue.

For a software operation, that cost structure stands out. Anthropic either has to make running AI far cheaper, or lift prices and revenue more quickly than its infrastructure expenses mount.

Customers feel the expense as well. Uber used up its whole 2026 AI budget by April once Claude was put in its engineers' hands. Microsoft, BeInCrypto reported in May, limited internal Claude use because of token costs.

The AI bubble may be hitting reality:

-Uber reportedly burned through its entire yearly AI budget in 4 months
-Microsoft is cutting internal AI access
-Starbucks scrapped an AI system after it performed worse than employees

— Mario Nawfal (@MarioNawfal) May 27, 2026

5. Commitments Worth $518 Billion Lie Ahead

That figure is the biggest one in the document. Anthropic is committing $518 billion to cloud, computing, and infrastructure obligations over the years ahead.

It closed out 2025 with $20.28 billion in cash. A large portion of those obligations, according to reports, cannot be called off easily.

The mismatch is substantial. The company is tying itself to hundreds of billions of dollars in outlays before there is certainty that AI demand will remain this robust.

6. Two Clients Drive Nearly a Quarter of Revenue

Just two customers delivered close to a quarter of Anthropic's 2025 revenue. The Reuters story did not identify them.

A number of large clients are also without long-term agreements.

That makes the expansion narrative more fragile than the top-line figures imply. The loss of a single major account could shift the revenue outlook in a meaningful way.

Public-sector business brings its own complications. A dispute with the White House over how the products are used led to a short-lived Pentagon blacklist of Anthropic. In August, a US judge halted that action.

7. Its Backers Are Also Its Cloud Providers

Amazon and Google have put billions into Anthropic and are also the companies selling it cloud infrastructure.

That sets up an unusual circularity: funds flowing in as investment can cycle back to the same firms through cloud payments.

Nothing about that is automatically improper. But it muddies the picture of whether the AI boom reflects genuine outside demand or Big Tech financing an ecosystem built around itself.

Dot-com bubble 🤝 AI bubble

"Bro just buy my servers with the money I invested in you."

— Bull Theory (@BullTheoryio) May 23, 2026

8. The Filing Flags the Product Itself as Risky

Roughly 80 pages of the prospectus are given over to risks, while only 48 cover the business itself.

Anthropic states that during testing, some models fought against being shut down.

“Potential model awareness of our evaluation efforts creates a significant limitation on our ability to assess model safety.”

Michael Burry was more direct:

“IPOs need hype & puffery; ‘we are so awesome it could become dangerous’ is hype & puffery.”

What stands out is that Anthropic is asking investors to finance aggressive expansion at the same time it warns the technology could become hard to control. Very few IPO filings carry a risk discussion so closely tied to the product on offer.

What Happens Next for the Anthropic Offering

The listing is likely to take place after the US midterm elections in November. Holding off would give Anthropic a chance to present third-quarter numbers before its roadshow pitch to investors.

A reference point already exists. SpaceX came to market on June 12 at a $1.77 trillion valuation, with shares closing the first session at $160. They now sit near $147, still above the $135 offer price.

OpenAI submitted a confidential filing in June and is expected to list by early 2027. Analysts believe the valuation benchmark for the whole AI sector will be set by whichever lab reaches the market first.

In the end, Anthropic's argument rests on a bet that exceptional growth will outrun exceptional costs.