Anthropic Asks Wall Street to Believe in AGI

The S-1 filing for Anthropic's Nasdaq IPO, under the ticker "ANT", revealed not just the company's financial figures but an ambitious narrative that goes far beyond a simple chatbot: the company wants investors to believe that artificial general intelligence (AGI) is not merely a technical goal, but the foundation of a business potentially larger than any software platform in history.

The central contradiction

In the first half of 2026, Anthropic recorded revenue of $16.3 billion — a 14-fold increase from the same period the year before. Yet it reported a net loss of $80.7 billion. This colossal loss is largely related to a one-time accounting charge, not the company's operational cash flow. What really matters are the operational numbers: running rate revenue exceeded $47 billion in early 2026, and incremental margins are running near 90%.

But the S-1 reveals a structural tension that already concerns Wall Street analysts. Anthropic is contractually committed to SpaceX (which acquired xAI) to spend up to $120.5 billion on compute capacity through 2029 — about $1.25 billion per month, or roughly 30% of its estimated 2026 revenue. On the other side, revenue is usage-based, metered and continuously repriced, with no long-term contractual terms. If demand for inference tokens slows — or if prices fall as compute capacity expands, as happened in the 1999-2002 long-haul bandwidth bubble — this fixed commitment could become a stranded cost.

The business model behind Claude

Anthropic's market philosophy is distinct from its main competitor, OpenAI. While OpenAI built a massive consumer base through ChatGPT, with billions of daily prompts, Anthropic has always focused on the enterprise market, which represents about 80% of its business versus 40% for OpenAI. Since the beginning of 2026, Anthropic already had 43.8% of American companies using its services on the Ramp platform, versus 40.6% for OpenAI.

Anthropic's most strategic product is Claude Code, launched to the general public in May 2025. It already generates a running rate revenue of over $2.5 billion, with more than 4% of all GitHub public commits being written by Claude Code — a figure that doubled in a month. Enterprise use of Claude Code represents more than half of the product's total revenue, and enterprise subscriptions quadrupled since the start of 2026.

Anthropic's pricing strategy has always followed OpenAI's: when ChatGPT Enterprise launched at $60 per seat, Claude Enterprise was priced identically. When OpenAI reduced to $40 in Q1 2025, Anthropic followed. But that pattern broke in Q3 2025, when OpenAI migrated to a pooled credit model, and for the first time, Anthropic did not follow. Inverting the historical pattern, it was OpenAI that later adopted Anthropic's pricing model.

The $10 trillion bet

Joey Brookhart of SemiAnalysis argues in The Verge's podcast that Anthropic could truly be worth $10 trillion — a number that sounds impossible to conservative analysts, but makes sense if you believe in the revenue-per-compute-dollar trajectory. Since Q1 2025, revenue per dollar of compute has risen 5.2x. A customer who spent $1 last year now spends $9. And this is only with frontier models, not including the volume of smaller and intermediate models powering large-scale inference workloads.

Brookhart compares Anthropic's narrative to Amazon's in the early days: "you can say 'this is a good business model, but don't expect those margins to continue, even though long-term margins can be quite a bit higher'". The comparison is deliberate — Dario Amodei and the Anthropic team are selling the same market vision that Jeff Bezos sold Amazon investors in 1997: aggressive growth, margins that expand organically with scale, and a total addressable market (TAM) vision that encompasses literally "revolutionizing every area of human endeavor," as the S-1 states.

S-1 data supports this narrative in some respects. Among companies spending over $100,000 annually with Anthropic, every one of the Fortune 10 is included, over 50% of the Fortune 500, and over 30% of the Global 2000. Within the Fortune 500, the 10 largest software and IT services companies, the 10 largest internet companies, nine of the 10 largest pharmaceutical companies, the 10 largest banks, nine of the 10 largest semiconductor companies, and eight of the 10 largest healthcare companies are represented. The enterprise penetration is real.

The risk nobody wants to discuss

Dr. Chan Ahn, CEO of Tessera PE and former Goldman Sachs and JPMorgan executive, identified in Benzinga the most delicate structural risk in Anthropic's business: revenue quality. In a public market, investors do not tolerate volatile revenue and fixed long-term cost commitments. If AI inference demand slows — which is plausible given that prices per million tokens fell 24% in the blended price between OpenAI and Anthropic in recent weeks — Anthropic would have $15 billion per year in fixed costs with SpaceX that do not adjust, while revenue falls.

The regulatory stress test already happened: in June, U.S. export controls temporarily took Anthropic's two leading models offline for 19 days, slowing revenue growth before it recovered. This exposed the fragility of a usage-based revenue model in contrast to fixed contractual costs.

The question is whether Wall Street will buy Anthropic's AGI story as a long-term investment or if it will price in the risks on the first day of trading. Anthropic is asking investors to believe not just in a product, but in an entirely new category of business — and the stock market was not built to fund beliefs without proof.

Sources: The Verge, Ramp, Benzinga / Yahoo Finance

✓ Independent sources cross-checked and verified before publishing