Anthropic S-1/Aamended IPO prospectus

Executive Summary

Anthropic sells Claude artificial-intelligence models, developer tools and applications to organizations and professional users, earning usage fees and subscriptions. It is expanding from coding into multi-step knowledge work while securing dedicated computing capacity, which management considers the principal growth constraint. Revenue reached $33.2B in 9M2026 and operating losses narrowed sharply, although financing-related accounting charges overwhelmed that improvement. The unpriced IPO offers buyers ≈6.3% of the company at an implied $917B–$1.02T valuation: a price requiring sustained enterprise expansion, durable product advantages and productive use of enormous infrastructure commitments.

Business Model and Growth Drivers

Consumption-based revenue contributed $26.3B in 9M2026, versus $2.13B a year earlier; subscriptions contributed $6.94B. Growth therefore depends primarily on customers using Claude more, not merely signing contracts. Customers generating at least $0.100M of trailing-year revenue rose from more than 950 to more than 11,900. This enterprise-adoption proxy supplied 74% of revenue.

Claude Code plans, writes, tests and debugs software; Claude Cowork extends agentic workflows into broader knowledge work. Anthropic began combining Cowork with Claude.ai in 09/2026. The Claude Platform supplies tools, managed agents, orchestration and connectivity for customer-built workflows. Opus 5.5 reduced cost per task by ≈40% versus Opus 5 through pricing and efficiency improvements, encouraging usage. Management prioritizes coding, agents and safety rather than image or video generation.

Net revenue retention was ≈900%, comparing existing qualified customers’ latest three-month revenue with the corresponding period a year earlier; fixed-price consumer subscriptions are excluded. October Run-Rate Revenue reached ≈$81B, annualizing recent consumption and subscription fees—not reporting contracted revenue or forecasting sales.

AWS, Google Cloud and Microsoft Azure extend distribution but also supply infrastructure and financing. Cloud marketplaces generated 44% of 9M2026 revenue; the top five customers supplied ≈25%. “Anthropic Direct,” using leased chips and dedicated third-party-operated facilities, seeks better capacity access and economics while increasing execution obligations.

The Offering

• Implied base-offering calculations: 825M shares at $70.00–$78.00 raise $57.8B–$64.4B gross. Including the dollar-fixed placement, outstanding economic shares vary from 13.107B at the low price to 13.069B at the high: $917B–$1.02T market capitalization, 6.29%–6.31% IPO ownership and 200–222 times FY2025 revenue. Including disclosed options and unvested restricted stock units gives an award-diluted count of 15.955B–15.917B and capitalization of $1.12T–$1.24T; further note conversions, matching obligations and plan reserves add potential dilution.
• Anthropic sells 801M shares; existing holders sell 23.6M. Underwriters may buy another 124M company shares. Goldman Sachs and Morgan Stanley lead; Nasdaq listing is sought as ANT. The offering remains unpriced.
• Google, NVIDIA, AMD, Broadcom, Micron, Microsoft and TSMC agreed conditionally to a $26.0B nonvoting placement. The IPO does not depend on its completion. At $74.00, combined company net proceeds are ≈$84.6B, funding compute, infrastructure, operations and hiring, including ≈$3.8B for employee-share tax withholding. Secondary proceeds go to sellers.
• Each $74.00 share would be backed by $8.21 of net tangible assets after the IPO, a $65.79 gap; existing holders paid $11.50 per share on average.
• Lock-ups are staggered: groups covering ≈2.7B shares generally expire around Q1 2027 results or day 180; ≈8.0B shares and founders generally extend to Q3 results or day 365, with earlier releases. Up to 3.3B additional shares may become sale-eligible before the general 180-day expiry. Placement shares are restricted but not underwriter-locked.

Key Financial Highlights

Amounts in $B unless indicated; income and cash flows are historical.

After IPO assumes conversions, share settlements, the offering and placement at $74.00, expenses and tax withholding.

• Growth: 9M2026 revenue increased 1,206%, versus 1,088% in FY2025; Q3 growth eased to 1,127%. Management expects growth rates to decline.
• Operating economics: compute expense reached $18.8B, but fell to 57% of revenue from 172%. Q3 used $5.0B of expensive short-term compute; management expects near-term margin pressure.
• Accounting and capital: rising company valuations produced $94.3B of noncash financing remeasurement losses. Accumulated losses of $156B and $158B of preferred stock held outside permanent equity explain the deficit—not equivalent borrowing. Conversions and related adjustments create $23.4B of equity before offering proceeds; note carrying value falls $19.6B.
• Commitments: ≈$901B of purchase and lease obligations include ≈$70.7B due within twelve months and ≈$452B before year-end 2030. These dwarf current operating cash generation and are often payable regardless of usage.

Governance and Control

• Founder LLC’s Class F share fixes general voting control at 50.1%; the Long-Term Benefit Trust elects a board majority. Public Class A holders have one vote per share, while strategic Class N shares are generally nonvoting. Controlled-company relief permits a non-independent nominating committee.
• Public-benefit status allows mission-focused decisions to outweigh shareholder returns. The company contributed 847M shares to its Public Benefit Fund.
• The IPO triggers ≈$14.2B of restricted-share compensation expense. Options and unvested restricted stock units together represent ≈18% of the award-diluted count, with combined intrinsic/share value of ≈$206B at the midpoint; new plans permit further dilution.

Risk Assessment

• Valuation accounting: auditors identified convertible-note valuation as a critical audit matter; subjective assumptions can materially distort reported earnings.
• Capacity-demand mismatch: long-duration minimum spending faces largely consumption-based revenue, creating underutilization and financing risk.
• Product and partner dependence: Claude concentration and reliance on cloud partners that also compete could undermine pricing, distribution and access.
• Safety and legal constraints: testing incidents, misuse, copyright litigation and contested federal restrictions can force release delays or lost business.
• Valuation and share supply: the revenue multiple requires exceptional execution, while staggered releases can substantially expand tradable supply.

What to Watch

• Enterprise expansion: retention and the large-customer cohort must show broader, durable usage rather than temporary spending.
• Cowork adoption: subsequent disclosures should establish whether broader knowledge-work deployments repeat Claude Code’s commercial success or stall.
• Compute economics: watch whether dedicated capacity replaces costly short-term supply and supports improving margins, rather than idle commitments.
• Capital discipline: track infrastructure cash payments and financing against positive OCF; commitments funded productively support the case, repeated funding ahead of monetization weakens it.