Anthropic S-1/Aamended IPO prospectus
MixedControlled company
Sells Claude AI models and coding agents; revenue grew thirteenfold Y/Y in 9M 2026
Filed 2026-11-17 06:26 ET · Plans to list on Nasdaq
How this filing reads
Each score is our read of the filing's tone, not a rating. The bar runs from −1 (bad) to +1 (good). The tick marks the typical filing on Goblin, and the shaded band is where the middle half of filings score.How we read filings
- HealthLiquidity is $132B against $70.7B due within a year; $900.9B commitments dwarf cash generation.
- MomentumNine-month revenue rose from $2.55B to $33.2B as operating loss narrowed from $5.22B to $1.26B.
- OutlookManagement expects slower growth and higher compute spending, with temporary capacity sourcing pressuring margins.
A registration statement filed before the IPO. The offering price, if shown, is a proposed range; later amendments or the final prospectus may change the terms.
Changes since the S-1/A filed 2026-11-10
Amazon and Google will together own 28.4% of total capital, clarifying the scale of Anthropic’s ownership concentration among cloud partners.
The filing adds ownership, equity-plan reserve and resale details without adding financial periods; final IPO pricing remains pending within the indicated $70–$78 range.
All 3 changes
- New Cloud-partner ownership
Amazon and Google will own 14.7% and 13.7% of total capital, respectively; Amazon could reach 48.2% of Class N on full conversion without the ownership cap. - Changed Equity-plan share reserves
Additional reserve mechanics and amount unspecified. → The 2026 Plan reserves 1.92B shares, plus up to 3.40B as 2021 Plan shares become available or awards return. - New Class N resale conversion
Class N shares automatically convert one-for-one into Class A upon registered offerings, open-market sales or certain transfers to unaffiliated third parties.
Executive Summary
Anthropic sells Claude AI models and applications to developers, businesses and individual subscribers, earning usage fees and subscriptions. It is expanding from coding into agents that perform broader knowledge work, combining frontier-model research, enterprise controls and cloud distribution. Revenue increased thirteenfold to $33.2B in 9M 2026, while operating loss narrowed to $(1.26)B; management nevertheless expects growth to slow and compute spending to rise. The predominantly primary, unpriced IPO implies a fully diluted valuation of $1.26T–$1.40T and gives IPO buyers 4.58% ownership, requiring durable enterprise expansion and economical utilization of an exceptionally large infrastructure build-out.
Business Model and Growth Drivers
Consumption generated 79% of 9M 2026 revenue; subscriptions supplied the balance. Customers generating at least $100,000 over the trailing year exceeded 11,900, versus more than 950 a year earlier. This cohort contributed $24.5B of revenue, versus $1.9B. Net revenue retention was ≈900%: current-quarter revenue relative to the same customers’ year-earlier quarter, excluding individual fixed-price subscriptions. That extraordinary expansion reflects early adoption, not a mature recurring-revenue base.
Claude Platform serves developers, Claude Code handles software development, and Claude Cowork extends agents into knowledge work. Agents can plan, use tools and repeat actions across longer tasks. Cowork adoption exceeded Code’s at the equivalent post-preview stage; the company began merging Cowork and Claude.ai in September 2026. Management emphasizes better capabilities at lower task costs: Opus 5.5 reduced customer cost per task by ≈40% versus Opus 5. Internally, Claude wrote more than 90% of September codebase contributions, illustrating both potential productivity and dependence on its own technology.
AWS, Google Cloud and Microsoft Azure broaden distribution, but cloud marketplaces supplied 44% of revenue and the top five customers supplied ≈25%. These partners can simultaneously be suppliers, investors and competitors. “Anthropic Direct,” using leased chips and dedicated facilities, aims to secure capacity and improve economics against competing frontier and open-weight models.
The Offering
- Implied IPO gross proceeds are $57.8B–$64.4B at the unpriced $70.00–$78.00 range. Including the price-dependent placement, post-deal outstanding shares are 13.1B, implying $917B–$1.02T capitalization, 6.3% ownership for IPO buyers and 200–222 times FY2025 revenue. Fully diluted shares are 18.0B, implying 275–306 times revenue. These computations use the filing’s reconciliation, adjusting placement shares and treasury-stock option treatment for price; it includes awards, rollover notes, acquisition shares, matching obligations and public-benefit shares, but not unused grant reserves.
- The 825M shares comprise 801M primary and 23.6M secondary shares sold by seven co-founders. Goldman Sachs and Morgan Stanley lead; listing is approved on Nasdaq as ANT. A 30-day option covers 124M additional primary shares.
- Seven strategic investors committed $26.0B alongside the IPO: Google and NVIDIA $10.0B each, AMD $2.5B, Broadcom, Micron and Microsoft $1.0B each, and TSMC $500M. The placement depends on the IPO, but not vice versa. A Reed Hastings-associated fund and Christopher Liddell’s household indicated nonbinding interest totaling $303M.
- At the $74.00 midpoint, estimated company net proceeds are $84.6B including the placement. They fund compute, infrastructure, operations and personnel, with ≈$3.8B for employee-share tax withholding; secondary proceeds go to sellers.
- Each midpoint-priced share is backed by $8.21 of net tangible assets after the IPO, a $65.79 gap. Existing holders paid $11.50 on average.
- Lock-ups generally end at specified earnings windows or 180/365 days, with earlier staged releases. Up to 1.7B shares become eligible after FY2026 results and 3.0B at the Q1 2027/day-180 milestone. Placement shares are outside these lock-ups, creating additional supply potential.
Key Financial Highlights
$B unless stated; historical figures except “After IPO,” which assumes conversions, share settlements, the midpoint offering and placement, and related tax payments.
| Metric | FY2023 | FY2024 | FY2025 | 9M 2025 | 9M 2026 | After IPO | Investor interpretation |
|---|---|---|---|---|---|---|---|
| Revenue | 0.041 | 0.386 | 4.59 | 2.55 | 33.2 | — | Growth accelerated to 1,206% from FY2025’s 1,088%; Q3 moderated to 1,127%. |
| Operating loss | — | — | — | (5.22) | (1.26) | — | Compute efficiency improved, but expensive temporary capacity constrained margins. |
| Net loss | (1.5) | (8.3) | (42.0) | (25.1) | (104) | — | Financing revaluations dominate reported losses. |
| Operating cash flow | — | — | — | (2.02) | 6.57 | — | Large valuation losses did not consume cash. |
| FCF: operating cash flow less equipment purchases | — | — | — | (2.14) | 5.64 | — | Excludes leased-equipment prepayments and future commitments. |
| Cash and short-term investments | — | — | — | — | 132 | 213 | New capital substantially increases liquidity. |
| Convertible notes, fair value | — | — | 41.7 | — | 131 | 111 | Carrying value substantially exceeds principal. |
| Total liabilities | — | — | — | — | 147 | 127 | The equity deficit is not equivalent to borrowing. |
| Stockholders’ equity | — | — | (50.5) | — | (150) | 108 | Conversions and new capital reverse the deficit. |
| Tangible equity | — | — | (50.7) | — | (151) | 107 | Removes goodwill and acquired intangible assets. |
- Earnings bridge: $94.3B of noncash financing revaluation losses arose mainly because rising company valuation increased convertible-note obligations; partner-financing expense added $11.0B.
- Compute: $18.8B included $6.5B of expensive short-term capacity. Q3 returned to operating loss after Q2 profitability; management warns further temporary sourcing could pressure margins.
- Commitments: ≈$900.9B of purchase and lease commitments includes ≈$70.7B due within twelve months; another estimated $150B–$175B of variable commitments is excluded. These obligations dwarf current operating cash generation.
- Equity bridge: before the IPO, $158B of preferred stock sat outside equity, while accumulated losses totaled $156B. Conversions and settlement adjustments produce $23.4B of equity before new proceeds lift it further.
Governance and Control
- Founder LLC holds 50.1% of general voting power regardless of future issuance; the Long-Term Benefit Trust elects four of seven authorized directors. Class A has one vote per share; Class N is generally nonvoting. Public investors cannot redirect control proportionately to ownership.
- The public-benefit mandate permits safety and societal priorities to outweigh near-term returns. Anthropic uses the controlled-company exemption for its nominating committee.
- Outstanding options and RSUs, including subsequent grants, cover 2.95B shares, or 16% of fully diluted shares. Options alone had $119B intrinsic value; IPO effectiveness triggers ≈$14.2B of cumulative RSU expense. Donation matching and discretionary public-benefit issuances create further dilution.
Risk Assessment
- Infrastructure mismatch: long-duration, often usage-independent commitments face short-term customer demand, potentially forcing financing or underutilized capacity.
- Valuation: the fully diluted revenue multiple requires sustained exceptional growth and durable profitability despite management’s expected slowdown.
- Partner concentration: cloud partners control essential distribution and compute while competing with Claude.
- Safety and regulation: unauthorized model activity, copyright exposure, export restrictions and government-use disputes can delay products or restrict sales.
- Control and dilution: mission-led controllers, substantial awards and staged resale releases limit influence and threaten ownership value.
What to Watch
- Enterprise expansion: continued growth in the large-customer cohort and retention would validate broader workflow adoption; declining expansion would challenge durability.
- Product economics: lower task costs must support profitable usage; persistent emergency compute sourcing would undermine that claim.
- Infrastructure funding: operating cash generation must increasingly support scheduled commitments rather than depend on repeated financing.
- Share supply: later filings will show whether staged releases and new benefit or employee issuances materially expand tradable supply and dilute holders.
Longer version: the summary of each of the prospectus's 27 sections, written by an AI model from the same filing.
Get emails when companies you follow file: email digest settings.
The summary and scores is written by an AI model reading the filing.
Model 006 · prompt v230 · 2026-11-17 15:54:25Z