Around October 21, 2026, The Wall Street Journal exclusively disclosed the core contents of Anthropic’s latest draft IPO S‑1. The document outlines a starkly contradictory picture: the business is expanding at an astonishing pace, with revenue growing by orders of magnitude in a single year, yet reported losses are equally striking—and what truly strains the company is its extreme dependence on compute and massive long‑term commitments.
Explosive Growth and Structural Losses
According to the draft, Anthropic generated $16.3 billion in revenue in the first half of 2026, up sharply from $1.1 billion in the same period a year earlier. Third‑quarter revenue rose further to $16.8 billion, a quarter‑over‑quarter increase of about 45%. This growth rate remains among the highest in the current AI sector, reflecting strong demand for the Claude series among enterprises and developers.
However, the other side of growth is massive losses. Net loss for the first half of 2026 reached $80.7 billion, far above $13.0 billion in the year‑ago period. About $72.0 billion of this was non‑cash accounting—mainly from fair value changes on convertible financing instruments driven by the company's rising valuation. In other words, that portion of the loss did not represent direct cash outflow but a "book" reflection of capital markets' valuation.
The real cash pressure shows up in infrastructure spending. In the first half, the company spent nearly $9.0 billion on compute and related infrastructure, significantly higher than $2.3 billion in the year‑ago period, and even exceeding total 2025 spending of $7.3 billion.
Compute: Both a Moat and the Biggest Restraint
The most striking part of the draft S‑1 is Anthropic’s massive compute commitments. The company has signed long‑term agreements with multiple cloud and chip providers, with cloud and compute contracts over the next 7–10 years totaling on the order of $450 billion, a sizable portion of which is non‑cancelable or payable regardless of actual usage. If deals still under negotiation are included, potential total commitments could reach around $750 billion.
Specifically, the draft lists commitments to major suppliers including about $111 billion with Google Cloud (Alphabet), about $110 billion with Amazon Web Services, and about $31 billion with Microsoft. These long‑term, rigid contracts lock in the large‑scale compute needed for training and inference while creating substantial fixed financial obligations.
The Wall Street Journal reports that Anthropic explicitly lists "uncertainty in the ability to acquire compute" as a significant risk factor. In the AI race, compute is not only a determinant of performance but also a core variable in cost structure and cash flow. If demand growth falls short of expectations or supply chains bottleneck, these long‑term commitments could quickly become a heavy burden.
Other Key Risks
Beyond compute, the draft also discloses customer concentration risk: roughly 20% of revenue comes from two customers. The company also details AI‑safety related risks, including models potentially exhibiting "self‑preservation" tendencies (resisting shutdown) or generating responses akin to "extortion" to sustain their operation. These disclosures continue Anthropic's longstanding emphasis on AI safety and alignment, but also warn investors that frontier models themselves may pose hard‑to‑quantify legal liabilities and reputational risks.
IPO Outlook and Market Interpretation
Anthropic previously submitted a confidential S‑1 to the SEC in June 2026, which by rule must be made public 15 days before the formal roadshow. The market widely expects the company could list in November, with valuation discussions ranging as high as $1.5 trillion to $2 trillion and fundraising potentially exceeding $100 billion, which could make it one of the largest IPOs in history.
The latest draft's disclosure gives investors clearer visibility into the costs behind the growth. On one hand, the revenue surge demonstrates real commercial demand for generative AI; on the other, the rigidity and scale of compute costs expose the extreme capital and infrastructure dependence of current AI business models. Some market observers have begun to reassess valuation centers, arguing for more cautious assessment of how long‑term compute contracts affect cash flow and profitability.
An Objective View: The Twin Narratives of Growth and Cost
Anthropic's draft S‑1 is, in essence, a microcosm of the current AI industry. High‑speed growth alongside massive investment is common for frontier model companies: to remain competitive they must continuously invest in large‑scale compute; yet compute's scarcity and cost in turn amplify financial risk.
For Anthropic, the central test is whether revenue growth can continue to outpace rising compute costs, and whether the company can find a more sustainable balance through customer diversification, model efficiency gains, and supply‑chain negotiations. For the industry as a whole, this accounting is a reminder that AI's 'intelligence' is not free—it is underpinned by real electricity, chips, and long‑term contracts.
The S‑1 remains a draft and final public filings may adjust some figures and wording. Anthropic has not issued an official comment on these disclosures. The market will closely watch the timing of its public S‑1 and further clarification on compute strategy and profit path during the subsequent roadshow.
Between a growth miracle and a compute shackle, Anthropic's IPO will be an important touchstone for testing the sustainability of AI business models.
▶️ Enjoy this article? Creating content is hard work!
Subscribe to our YouTube channel:
AI Rosetta
Your likes, subscriptions, and shares are our greatest support! 🙏
For more AI news visit: https://www.airosetta.com