# Anthropic account plan

> Anthropic builds the Claude AI models and apps for enterprises; $33,242 million revenue in 9M 2026, Nasdaq IPO as ANT in November 2026.

## At a glance

Anthropic, PBC is a Delaware public benefit corporation founded in 2021 and headquartered in San Francisco that develops Claude, a family of frontier AI models sold to businesses through models, the Claude Platform and applications such as Claude Code and Claude Cowork. Revenue was $4,585 million in FY2025 and $33,242 million in the nine months to September 30, 2026, with about 5,300 employees at that date. It is still loss-making on a GAAP basis (operating loss of $1,260 million in 9M 2026) but reported Adjusted Operating Income of $4,081 million. It priced its Nasdaq IPO under the ticker ANT on November 18, 2026; the Long-Term Benefit Trust and a Founder LLC of the seven co-founders control the vote, and Amazon and Google hold large non-voting stakes.

## Business units

- **Consumption-based revenue (pay-for-intelligence)** (79% of revenue in 9M 2026 ($26.3 billion)): Customers pay for usage metered in tokens and priced by the model chosen, mostly when building applications on Claude or using first-party applications. A growing number of enterprises sign annual or multi-year contracts with minimum spend commitments. It grew from $288 million in 2024 to $3.8 billion in 2025.
- **Subscription-based revenue** (21% of revenue in 9M 2026 ($6.9 billion)): Applications sold on subscription to enterprises and prosumers, generally for 12 months or less, with a path to buy more usage. It grew from $98 million in 2024 to $789 million in 2025, and prosumers act as a top-of-funnel engine for enterprise sales.
- **Claude models**: Four model classes, Mythos (including Fable), Opus, Sonnet and Haiku, priced along a capability-cost-speed curve. Model versions released in the six months to September 30, 2026 accounted for over 80% of Q3 2026 revenue.
- **Claude Platform**: Developer building blocks: tool use, computer use, memory, the Agent SDK, Claude Managed Agents, Agent Skills and the Model Context Protocol (MCP). Over 50% of customers with $100,000+ TTM revenue used the platform and at least one first-party application in Q3 2026.
- **First-party applications (Claude Code, Claude Cowork, vertical offers)**: Claude Code is the coding agent; Claude Cowork, the knowledge-work agent launched as a research preview in January 2026, is being merged with Claude.ai from September 2026. Domain versions cover Financial Services, Healthcare, Life Sciences, Legal, Education and Government, alongside Claude for Microsoft 365 and Claude in Chrome.

## Strategic priorities

- **Enterprise adoption of Claude**: Anthropic organises research, roadmap, go-to-market and Applied AI teams around enterprise use cases. It had more than 11,900 customers above $100,000 TTM revenue at September 30, 2026, up from more than 950 a year earlier, and a Net Revenue Retention of about 900%.
- **Agentic applications for knowledge work**: Claude Code and Claude Cowork are positioned as Virtual Collaborators that do long-running work alongside employees; Cowork's adoption outpaced Claude Code's at the equivalent point, and Cowork is being merged with Claude.ai.
- **Securing and using compute efficiently**: The growth plan is to secure compute beyond existing commitments, expand its own data-center footprint and improve efficiency across AWS Trainium, Google TPUs, NVIDIA GPUs and AMD GPUs. Revenue grew 1,127% year over year in Q3 2026 while compute and infrastructure expenses grew 382%.
- **Distribution through clouds and partners**: Claude is sold direct and on AWS, Google Cloud and Azure, with Anthropic keeping the customer contract; systems integrators (Accenture, Deloitte, KPMG, PwC), data platforms (Salesforce, Snowflake) and Ode with Anthropic for mid-sized companies extend reach.
- **New markets beyond software**: Anthropic intends to extend Claude into healthcare and life sciences, robotics and physical work, and to pursue selective acquisitions, investments and partnerships.
- **Safety alongside capability**: The company says safety and capability must advance together; it held back a Mythos-class model from broad release over cyber-warfare risk and disabled Fable 5 and Mythos 5 during a June 2026 export restriction.

## Latest results

Revenue for the nine months to September 30, 2026 was $33,242 million, up 1,206% from $2,545 million a year earlier, and Q3 2026 revenue reached $16,956 million against $11,555 million in Q2 2026. Q2 2026 was the first quarter with operating income ($974 million), but Q3 2026 returned an operating loss of $567 million, driven by $5.0 billion of short-term compute and $2.0 billion of non-cash donation-matching charges. Adjusted Operating Income was $4,081 million for 9M 2026, a 12% Adjusted Operating Margin, against a loss of $3,133 million in 9M 2025. Management expects margins to fluctuate as new compute is delivered and expects the revenue growth rate to decline over time. In May 2026, before the IPO, the company said run-rate revenue had crossed $47 billion.

## Investments and deals

- **Compute commitments**: About $900.9 billion of purchase and lease commitments, substantially all compute and infrastructure: about $451.6 billion before December 31, 2030 and about $70.7 billion due in the 12 months after September 30, 2026.
- **Anthropic Direct**: A growing share of compute comes from leasing chips directly and hosting them in dedicated facilities, alongside cloud agreements; data-center leases also cover Australia and Canada.
- **IPO and concurrent private placement**: 980,000,000 Class A shares at $83.00 plus a $26.0 billion private placement with AMD, Broadcom, Google, Micron, Microsoft, NVIDIA and TSMC Partners; net proceeds estimated at about $104.8 billion.
- **Series H funding and capacity deals**: $65 billion Series H at a $965 billion post-money valuation in May 2026, alongside up to five gigawatts of capacity with Amazon, five gigawatts of TPU capacity with Google and Broadcom, and GPU capacity at SpaceX's Colossus 1 and Colossus 2.
- **Anthropic Public Benefit Fund**: A wholly owned subsidiary received 846,840,960 Class N shares (5% of fully diluted shares as of June 30, 2026) to fund public benefit initiatives.

## Challenges and risks

- **Compute cost and forecasting risk**: Compute is bought 12 to 18 months ahead; short-term compute costs significantly more and is expected to compress margins. Dario Amodei describes overbuying by a year of growth as a path to bankruptcy.
- **Model behaviour and misuse**: The prospectus lists models behaving in unintended, misaligned or autonomous ways, and misuse by malicious actors, as leading risks with liability and reputational exposure.
- **Regulation and export controls**: In June 2026 a U.S. worldwide export restriction forced Anthropic to disable Fable 5 and Mythos 5 for all customers until it was lifted; similar controls could recur.
- **Copyright litigation**: FY2025 included the $1.5 billion Bartz settlement, approved in July 2026, and suits by music publishers continue.
- **Competition and concentration**: It competes with OpenAI, Google and open-weight models, and depends on a limited number of cloud providers and suppliers that are also investors and customers.

## What leadership says

> "we are doing everything we can to make Anthropic's revenue grow 20 or 30x a year instead of 10x a year" — Dario Amodei, Chief Executive Officer and Co-Founder, Dwarkesh Podcast, February 13, 2026

> "We're an enterprise business, therefore we can rely more on revenue. It's less fickle than consumer. We have better margins" — Dario Amodei, Chief Executive Officer and Co-Founder, Dwarkesh Podcast, February 13, 2026, on how much compute to buy

> "the majority of our revenue comes from businesses" — Daniela Amodei, Co-Founder and President, Stanford GSB View From the Top, June 29, 2026, on safety and revenue

> "will help us serve the historic demand we are experiencing, stay at the research frontier, and bring Claude to more of the places where work happens" — Krishna Rao, Chief Financial Officer, Series H announcement, May 28, 2026


## Where a supplier can help

- **Data-center and infrastructure supply**: Tied to the compute commitments and Anthropic Direct: suppliers of data-center capacity, power, networking, memory and facility operations fit a company securing compute across several chip families and its own sites.
- **Enterprise deployment partners**: Tied to enterprise adoption and the partner channel: integrators and consultancies that roll Claude Code or Cowork out across large customers address the procurement, security and compliance steps that slow diffusion.
- **Vertical data and connectors**: Tied to the domain offers (financial services, healthcare, life sciences, legal, government): data providers and software vendors whose content or connectors make Claude useful in a regulated vertical.
- **Safety, security and compliance tooling**: Tied to the model-behaviour, misuse and export-control challenges: evaluation, monitoring, security and regulatory-compliance services that support safe release of frontier models.
- **Hiring and scaling the organisation**: Tied to headcount growth to about 5,300 employees and international expansion: recruiting, workplace and HR services for a fast-growing research-heavy workforce.

_Tamtam's reading of the priorities and challenges above, not the company's own words._

## Key documents

### [Final IPO prospectus (Form 424B4) — Anthropic, PBC, Nasdaq listing as ANT](https://www.sec.gov/Archives/edgar/data/2130357/000162828026074944/anthropicpbc-424b4.htm)

Registration statement · 2026-11-18

- Anthropic, PBC (Delaware public benefit corporation, founded 2021, HQ San Francisco) priced its Nasdaq IPO (ticker "ANT") on November 18, 2026: 980,000,000 Class A shares at $83.00, $81,340,000,000 gross, plus a $26.0 billion concurrent private placement of Class N stock with AMD, Broadcom, Google, Micron, Microsoft, NVIDIA and TSMC Partners; net proceeds to the company estimated at about $104.8 billion.
- Revenue went from $41 million (2023) to $386 million (2024) and $4,585 million (FY2025), then $33,242 million for the nine months to September 30, 2026 (up 1,206%); Q3 2026 alone was $16,956 million versus $11,555 million in Q2 2026.
- Still loss-making on a GAAP basis: operating loss $8,060 million in FY2025 and $1,260 million in 9M 2026; Adjusted Operating Income of $4,081 million in 9M 2026 (12% Adjusted Operating Margin). Q2 2026 was the first quarter with operating income ($974 million); Q3 2026 swung back to a $567 million operating loss on $5.0 billion of short-term compute.
- One operating segment. Revenue is 79% consumption-based ("pay-for-intelligence", metered in tokens) and 21% subscription-based (9M 2026); United States $23,470 million and International $9,772 million of 9M 2026 revenue.
- Enterprise-led: more than 11,900 customers above $100,000 TTM revenue (from more than 950 a year earlier), more than 1,900 above $1 million, about 250 above $10 million; Net Revenue Retention about 900%; no customer above 10% of revenue.
- Products: Claude model classes Mythos (including Fable), Opus, Sonnet and Haiku; Claude Platform (Agent SDK, Claude Managed Agents, Agent Skills, MCP); Claude Code; Claude Cowork (being merged with Claude.ai from September 2026); vertical offers (Financial Services, Healthcare, Life Sciences, Legal, Education, Government); sold direct and through AWS, Google Cloud and Azure.
- Compute is the dominant cost and commitment: about $900.9 billion of purchase and lease commitments, $451.6 billion of it before end-2030 and $70.7 billion due in the 12 months after September 30, 2026; a shift toward "Anthropic Direct" (leased chips in dedicated facilities). Liquidity of $146.6 billion at September 30, 2026 before the IPO.
- About 5,300 employees at September 30, 2026 (about 940 at end-2024); TTM revenue per employee over $6.6 million. Over 90% of codebase contributions in September 2026 were written by Claude.
- Governance: the Long-Term Benefit Trust elects a majority of the board via Class T stock; a Founder LLC of the seven co-founders holds Class F stock with 50.1% of voting power; Amazon and Google hold large non-voting Class N stakes. Controlled company under Nasdaq rules.
- Key risks named: misaligned or autonomous model behaviour, staying at the frontier, securing compute and chips (TSMC concentration), customer and channel concentration, regulation and export controls (a June 2026 worldwide export restriction on Fable 5 and Mythos 5), copyright litigation (Bartz settlement of $1.5 billion), and the novel governance structure.

### [Stanford GSB View From the Top: Daniela Amodei says curiosity is underrated](https://www.gsb.stanford.edu/insights/daniela-amodei-says-curiosity-underrated)

Podcast · Daniela Amodei, Co-Founder and President · 2026-06-29

- Daniela Amodei, co-founder and President of Anthropic, in a Stanford GSB View From the Top conversation (podcast episode published June 29, 2026), on why Anthropic exists, safety versus revenue, jobs, adoption and the AI bubble question.
- Anthropic was founded as a public benefit corporation by seven people who left OpenAI; she says they were "running towards something, versus running away from something".
- Safety means "radical responsibility": preventing chemical, biological and cyber misuse, plus user wellness, child safety and election integrity; she cites the decision not to release a Mythos-class model widely because of cyber-warfare risk (Project Glasswing).
- "the majority of our revenue comes from businesses", and businesses do not want unsafe models, so safety and revenue have mostly been aligned; the tension now is about time.
- On jobs: Anthropic's economic index shows AI mostly complementing work so far, except in customer service; software developers will "still exist" but write less code.
- On adoption: usage is skewed by education, gender, wealth and geography; developing countries are more optimistic about AI; "the gun just went off" on the race.
- On the bubble: the risk she takes seriously is capex, since compute is bought far in advance; "we could absolutely be wrong".
- Regulation: favours "sensible regulation" built with regulators; no ads in Claude because conversations are personal.
- Cites an 81,000-person qualitative study of AI use run by Anthropic.

### [Anthropic raises $65B Series H at $965B valuation](https://www.anthropic.com/news/series-h)

Report · 2026-05-28

- Anthropic raised $65 billion in Series H funding at a $965 billion post-money valuation (announced May 28, 2026), led by Altimeter Capital, Dragoneer, Greenoaks and Sequoia Capital.
- Run-rate revenue crossed $47 billion earlier in May 2026, after continued enterprise adoption since the Series G in February.
- Co-leads: Capital Group, Coatue, D1 Capital Partners, GIC, ICONIQ and XN; the round includes $15 billion of previously committed hyperscaler investments, $5 billion of it from Amazon.
- Memory and chip makers Micron, Samsung and SK hynix joined as strategic infrastructure partners.
- New compute: up to five gigawatts with Amazon, five gigawatts of next-generation TPU capacity with Google and Broadcom, and GPU capacity at SpaceX's Colossus 1 and Colossus 2; AWS "remains our primary cloud provider and training partner".
- Uses of funds: safety and interpretability research, compute to meet demand for Claude, and scaling products and partnerships.
- CFO Krishna Rao: the funding "will help us serve the historic demand we are experiencing".

### [Dwarkesh Podcast: Dario Amodei on scaling, compute and Anthropic's economics (transcript)](https://singjupost.com/anthropic-ceo-dario-amodeis-interview-on-dwarkesh-podcast-transcript/)

Podcast · Dario Amodei, Chief Executive Officer and Co-Founder · 2026-02-13

- Dario Amodei, CEO and co-founder, on the Dwarkesh Podcast (published February 13, 2026; transcript on The Singju Post), a long conversation on scaling, timelines, Anthropic's economics, compute buying, Claude Code, regulation and China.
- Revenue grew about 10x a year: "2023 it was 0 to $100 million, 2024 it was $100 million to a billion. 2025 it was a billion to $9 or $10 billion", and Anthropic "added another few billion to revenue in January" 2026.
- Puts 90% odds on a "country of geniuses in a data center" within 10 years, with a hunch of one to two or one to three years; expects end-to-end software engineering by models in one or two years.
- Diffusion is fast but "not infinitely fast": enterprises adopt Claude Code faster than usual technology but still go through legal, security and compliance.
- Compute buying is the central bet: overbuying by a year of growth "you go bankrupt"; Anthropic is buying "an amount that's comparable" to the biggest players but not trillions a year; as an enterprise business it relies on less fickle revenue and better margins.
- Profitability depends on demand forecasting errors; a 2028 target cited to investors was "our attempt to do the best we can with investors".
- Expects a few-player industry like cloud, with more differentiated models; believes the API business model is durable and pay-for-results pricing may come.
- Claude Code began as an internal tool ("Claude CLI") and was launched after fast internal adoption.
- Policy: opposed a 10-year moratorium on state AI laws without a federal standard; supports transparency rules first and chip export controls to China.
- Spends a third to 40% of his time on culture; runs a roughly 2,500-person company and speaks to all staff every two weeks.

## Questions about Anthropic

### What does Anthropic do?

Anthropic develops Claude, a family of frontier AI models, and sells it to businesses and prosumers through models, the Claude Platform and applications such as Claude Code and Claude Cowork. It is a public benefit corporation focused on safe development of advanced AI.

### What is Anthropic's revenue?

Revenue was $4,585 million in FY2025 and $33,242 million in the nine months to September 30, 2026; Q3 2026 alone was $16,956 million.

### How many employees does Anthropic have?

About 5,300 at September 30, 2026, up from about 940 at the end of 2024, a majority in research, engineering and product development.

### Who leads Anthropic?

Dario Amodei is CEO and co-founder, Daniela Amodei is President, co-founder and Chairperson, and Krishna Rao is CFO.

### Is Anthropic a public company?

Yes. It priced its IPO on the Nasdaq Global Select Market under the ticker ANT on November 18, 2026, at $83.00 per Class A share.

_Written by Tamtam from the company's own published documents listed above, read in full. Figures are as those documents state them, for the periods they name. (2026-11-19)_
