Anthropic
Frontier AI safety lab — the Claude model family (Opus, Sonnet, Haiku), Claude.ai, the Claude API and Claude Code — built on a Constitutional AI safety stack with multi-cloud capacity anchored by a $100B / 10-year Amazon Web Services commitment and continuing Google Cloud TPU access.
The Business
Anthropic is a frontier AI safety lab and the maker of the Claude model family — Claude Opus (frontier reasoning), Claude Sonnet (balanced production workload) and Claude Haiku (cost-optimised inference) — distributed through four main surfaces: Claude.ai (consumer AI assistant on web, mobile and desktop, with Free, Pro at $20/month and Max tiers), the Claude API (developer and enterprise inference, OpenAI-compatible at the endpoint level), Claude on AWS Bedrock (the lead enterprise distribution channel), and Claude Code (the developer-coding agent). The Constitutional AI methodology and the mechanistic-interpretability research programme anchor the company’s safety-first positioning. Governance sits under a Public Benefit Corporation structure with a Long-Term Benefit Trust empowered to elect a majority of board seats over time — Anthropic’s distinctive safety-governance architecture, separate from the OpenAI Foundation / PBC structure finalised in October 2025.
Customers and Distribution
Anthropic disclosed 300,000+ business customers across its direct API and AWS Bedrock channels in the Claude Partner Network announcement, with approximately 80% of revenue generated by enterprise API usage. AWS separately disclosed more than 100,000 businesses running Claude on Amazon Bedrock in the April 2026 expanded-strategic-collaboration announcement — the largest single distribution channel for Claude into the enterprise. Distribution sits across four channels: the Claude API direct (developers and enterprises), AWS Bedrock (the prime placement and the volume channel), Google Vertex AI (a continuing distribution leg alongside the original Google Cloud TPU relationship), and Microsoft 365 Copilot (Claude as a frontier-model option since late 2025). Channel partners include Cloudvisor as an authorised reseller for startups (May 2026) and the broader Claude Partner Network with $100M of Anthropic investment in channel partner development. Anthropic crossed approximately $30B annualised run-rate revenue in April 2026, ahead of OpenAI’s $20B end-2025 disclosure at the comparable disclosure point, with roughly ten-fold annual growth in each of the preceding three years per the Series G announcement.
Model Strategy
Anthropic is a frontier-capability play with a safety-first positioning: Claude Opus and Claude Sonnet are designed to hold leaderboard-leading capability on coding and long-context reasoning, with the Constitutional AI safety stack and the mechanistic-interpretability research programme as the brand differentiator the company most often cites in enterprise procurement. The infrastructure backstop is the deepest single-hyperscaler commitment in the FMP cohort: up to $100B / 10-years of AWS spend for up to 5GW of capacity, announced April 2026, anchored on the AWS Trainium / Inferentia accelerator stack alongside continuing NVIDIA capacity through AWS. Google Cloud TPU access is retained as the supplier-diversity leg — a relationship Anthropic has held continuously since the original Google strategic investment in 2024. The Claude API is OpenAI-compatible at the endpoint level, which both benefits Anthropic (developer mindshare and easy substitution into rival-API customer stacks) and exposes it (API customers can substitute at low switching cost). The Model Context Protocol (MCP), Anthropic-led and open-source, is the canonical agent-tooling ecosystem play; it creates ecosystem stickiness without itself locking customers in.
At A Glance
The Numbers
Annualised revenue
API tokens (monthly)
Headcount (FTE)
Funding to date
Leadership Team
Governance sits under a Public Benefit Corporation structure with a Long-Term Benefit Trust empowered to elect a majority of board seats over time — Anthropic’s distinctive safety-governance architecture. Senior research-team continuity has held through the 2024-2026 window with no senior departures of the magnitude seen at OpenAI (Mira Murati, Bob McGrew, Barret Zoph in September 2024) or Meta (Yann LeCun, November 2025). Other named senior figures publicly active in 2026 include Jan Leike (alignment research lead, joined from OpenAI in 2024) and Krishna Rao (CFO).
IM Framework Scoring
IM’s structured assessment of Anthropic’s competitive position. The summary below is the headline; expand “Show the full analyst-grade analysis” near the bottom for the per-dimension reasoning and evidence. Methodology →
Funding History
| Date | Round | Raised | Post-money | Lead investor(s) |
|---|---|---|---|---|
| Feb 2026 | Series G | $30B | $380B | GIC and Coatue (co-led with D.E. Shaw Ventures, Dragoneer, Founders Fund, ICONIQ, MGX) |
| Sep 2025 | Series F | $13B | $183B | ICONIQ |
| Mar 2025 | Series E | $3.5B | $61.5B | Lightspeed Venture Partners |
| Nov 2024 | Amazon strategic | $4B (Amazon) | — | Amazon (NASDAQ: AMZN) |
| Sep 2023 / Mar 2024 | Amazon strategic | $4B (Amazon) | — | Amazon (NASDAQ: AMZN) |
| Feb 2024 | Google strategic | ~$2B (Google) | — | Google (NASDAQ: GOOGL) |
| 2021-2023 | Seed through Series C | ~$1.5B | — | Jaan Tallinn, Spark Capital, Google, Salesforce Ventures, others |
Anthropic raised a $30B Series G at a $380B post-money valuation co-led by GIC and Coatue in 2026 — the second-largest private AI round on the public record after OpenAI’s $122B / $852B March 2026 raise. Amazon has separately committed up to $20B of additional investment in 2026 on top of the prior $8B Amazon strategic stack (two $4B tranches in Sep 2023 / Mar 2024 and Nov 2024) for a cumulative Amazon position of up to ~$33B. Google has invested approximately $2B as a separate strategic partner. Anthropic has pledged $100B / 10-years of AWS spend for up to 5GW of capacity announced in April 2026 — the largest single hyperscaler-customer commitment on the public record. Cumulative external equity is now well above $70B excluding the AWS spend commitment, which is framework-spend rather than equity. Round dates and lead-investor identities are taken from Anthropic’s own news pages and corroborated by named-author CNBC and TechCrunch coverage.
Competitive Landscape
Anthropic’s competitive set sits in three concentric rings: frontier closed-model labs running on Western hyperscaler infrastructure (OpenAI, Google), hyperscaler-attached AI portfolios that are simultaneously partners and competitors (Microsoft, Meta), and the Chinese frontier labs that reset the cost-of-inference curve. Anthropic is unusual in the set because it is enterprise-leaning on brand — Claude is the model most enterprises pair against OpenAI in 2026 RFPs — while sitting on a single dominant hyperscaler relationship (AWS, up to $100B / 10-years) with Google Cloud TPU access retained as the diversification leg.
| Competitor | Positioning | Distribution edge | Threat profile |
|---|---|---|---|
| GPT-5 / o-series (OpenAI) |
Frontier consumer-AI default brand; ChatGPT and the OpenAI API are the primary closed-model competitors to Claude on capability and the only rival API at the same per-token-pricing band. | Direct ChatGPT distribution (900M+ weekly active users on the March 2026 funding-round disclosure), Microsoft Foundry / Azure OpenAI partnership, AWS Bedrock placement (live May 2026), and a broad developer API. | High — ChatGPT is the consumer-AI default brand and the largest single substitute for Claude.ai; OpenAI is also the brand Anthropic is most often paired against in enterprise 2026 RFPs. |
| Gemini 2.5 / 3.x (Google (NASDAQ: GOOGL)) |
Frontier multimodal family with the largest distribution surface in the set — Workspace, Android, Search, YouTube, Vertex AI — backed by self-funded $180-190B 2026 capex. | 900M-MAU Gemini app, Workspace install base, Android default placement, plus Vertex AI for developers and enterprises — where Claude is also available. | High and asymmetric — Google is a Claude distribution partner via Vertex AI and an early Anthropic investor (~$2B), but Gemini competes head-to-head on every enterprise inference workload. |
| Copilot / Azure AI (Microsoft (NASDAQ: MSFT)) |
Productivity-suite AI on Microsoft 365 footprint; OpenAI-powered with growing in-house Phi / MAI model contribution — and a Claude option inside Microsoft 365 Copilot since late 2025. | Enterprise Office 365 footprint; Foundry developer platform; Azure OpenAI Service. | Medium-high — Microsoft channels Claude into the largest productivity-suite footprint in the world but is OpenAI’s largest customer and an emergent model-layer competitor. |
| Llama / Meta AI (Meta (NASDAQ: META)) |
Open-weight frontier models plus consumer Meta AI assistant embedded in WhatsApp, Instagram, Facebook. | Consumer reach across Meta’s family of apps; open-weight enterprise distribution via Hugging Face and direct download. | Medium — less head-to-head than OpenAI or Google on enterprise reasoning workloads, but open-weight pressure on Claude API pricing and a consumer-distribution lever Anthropic does not have. |
| DeepSeek / Qwen / Z.ai | Chinese frontier labs with credible reasoning and coding capability at materially lower training and inference cost; open-weight by default. | API plus open-weight distribution; constrained in US enterprise procurement by geopolitical posture but widely used in EU/APAC developer ecosystems — including for Claude-API-compatible workloads. | Medium-high — reset the price-of-frontier-inference curve and the open-weight quality floor; structural cost pressure on Claude API margin, particularly on the coding workloads where Claude leads today. |
Pricing benchmark: Claude Opus 4 and Claude Sonnet 4 on the Claude API and on AWS Bedrock are priced per-token in line with OpenAI GPT-5 mainline and Google Gemini 2.5 Pro — within roughly a 2x band depending on input/output mix and context window. Claude.ai consumer pricing (Free / Pro $20 / Max $100-200) tracks the ChatGPT consumer band. The competitive frame is capability cadence and distribution, not headline per-token price.
Potential Risks
The case for Anthropic at IM Framework 8.44 rests on annualised run-rate revenue ahead of OpenAI at the comparable point ($30B vs. $25B in April 2026), the Series G capital base at $380B post-money, the AWS distribution channel through Bedrock to 100,000+ businesses, and Claude’s enterprise-default status on coding and long-context reasoning. The case against splits into five risks of differing magnitude — with copyright/regulatory exposure the most active (the $1.5B Bartz settlement is the largest AI-copyright payment on record) and AWS supplier concentration the most structural.
Regulatory and copyright exposure — the $1.5B Bartz settlement is the largest AI-copyright payment on record
The Bartz v. Anthropic class action settled at approximately $1.5B in 2025 for the use of an estimated ~500,000 pirated books in training data, at roughly $3,000 per book — the largest AI-copyright settlement on the public record. The final approval hearing was scheduled for April 2026 with author distributions by June 2026. Anthropic separately signed the EU GPAI Code of Practice in July 2025, signalling a cleaner go-forward compliance posture under the EU AI Act’s general-purpose-AI obligations. The sub-rubric score on D4c regulatory exposure was lowered from 6 to 5 on this evidence; the settlement establishes precedent that AI companies face material settlement exposure on pirated training data, and other frontier labs (OpenAI, Meta) face active litigation on adjacent theories.
AWS supplier concentration vs. multi-cloud diversification
Anthropic’s April 2026 commitment of up to $100B / 10-years of AWS spend for up to 5GW of capacity is the deepest single-hyperscaler commitment in the FMP cohort. The bull case is operating leverage — Claude on Bedrock is the lead distribution channel into AWS’s enterprise install base, and the AWS Trainium / Inferentia accelerator stack offers a cost structure that NVIDIA-only rivals cannot match. The bear case is single-supplier entanglement: the AWS commitment is materially larger than Anthropic’s continuing Google Cloud TPU exposure, and a strategic shift inside AWS (Bedrock prioritisation, custom-silicon roadmap timing, capacity allocation across Anthropic vs. AWS’s own Nova family) would propagate at a scale no other Anthropic supplier-shift could.
Customer portability at the API layer is real
The Claude API is OpenAI-compatible at the endpoint level — the de facto standard across rival frontier APIs — so API-only customers can substitute Claude for GPT-5 or Gemini 2.5 Pro with low switching cost. The Model Context Protocol (MCP, Anthropic-led and open-source) creates ecosystem stickiness around the agent-tooling layer but does not itself lock customers in. AWS Bedrock placement shares AWS-stack switching cost — a positive for retention — but the sub-rubric score on D1c portability was held at 7 on this evidence: substitution risk is real and structural, and the strategic bet relies on continued capability lead rather than lock-in.
Capital intensity vs. cash-flow profile
Anthropic crossed $30B annualised run-rate revenue in April 2026 — ahead of OpenAI’s $20B end-2025 disclosure (Friar/CNBC) and an estimated ten-fold annual growth rate over each of the preceding three years per the Series G announcement — on a $380B post-money valuation. The cumulative compute commitment (up to $100B AWS over 10 years, plus continuing Google Cloud TPU access) is structurally large, and Anthropic does not separately disclose burn or cash-flow positions. The capital base behind the Series G is enormous, but so is the implied investment trajectory; the disruption composite at 9.65 is essentially priced for continued frontier-capability cadence.
Frontier-capability cadence is contested every quarter
Anthropic’s strategic bet is enterprise-default frontier-model status sustained by Claude Opus and Claude Sonnet release cadence, with Claude Code as the lead developer surface. The leaderboard position changes monthly — OpenAI, Google and Meta have all retaken the top slot at points across the past 18 months, and Chinese frontier labs (DeepSeek, Z.ai, Alibaba Qwen) are credible at the frontier on coding and reasoning where Claude leads today. A sustained capability slip against GPT-5 or Gemini 3.x would compress the pricing power that the Claude API and the Claude.ai Max tier currently enjoy.
Recent IM Coverage
- Eight Futures for Agentic AI (#IM105) May 2026.
- Foundation Model Providers — Category Report May 2026.
Show recent press coverage of Anthropic
- Apr 2026 — Anthropic to spend $100B with Amazon Web Services over 10 years for up to 5GW of capacity.
- Apr 2026 — AWS and Anthropic deepen strategic collaboration — 100,000+ businesses now running Claude on Bedrock.
- 2026 — Anthropic Series G at $380B post-money valuation — GIC and Coatue co-lead $30B round.
- May 2026 — Cloudvisor named authorised Claude reseller for startups.
- 2026 — Anthropic launches Claude Partner Network with $100M investment in channel partners.
- 2025 — Amazon doubles its bet on Anthropic with an additional $4B investment.
- Jul 2025 — Anthropic signs EU General-Purpose AI Code of Practice.
- Mar 2025 — Anthropic raises $3.5B Series E at $61.5B valuation, led by Lightspeed.
- 2025 — Bartz v. Anthropic class settlement: $1.5B for ~500,000 pirated books in training data — the largest AI-copyright settlement on record.
Curated feed of named-source coverage — Anthropic’s own news pages, AWS / Amazon Newsroom announcements, named-author tech and business press, and named regulatory and litigation primary sources. Excludes paywalled article bodies of The Information, WSJ, FT and Bloomberg (headline + free-snippet only), wire-aggregator reposts and unsourced AI round-up pieces.
Show the source register for the figures on this page
IM operates a primary-source-where-possible discipline. The figures above come from:
- Revenue: Anthropic is private and does not file. Annualised run-rate revenue crossed $30B in April 2026 per the Anthropic Series G announcement, up from approximately $9B at end of 2025 and roughly ten-fold annual growth in each of the preceding three years. The figure is ahead of OpenAI’s $20B end-2025 disclosure at the comparable disclosure point. We label this “annualised run-rate revenue” rather than “GAAP revenue” because Anthropic does not publish GAAP financials.
- Usage — Claude on Bedrock customer count: AWS disclosed more than 100,000 businesses running Claude on Amazon Bedrock at the April 2026 expanded-strategic-collaboration announcement per the Amazon Newsroom post. Anthropic separately discloses 300,000+ business customers across its direct API and Bedrock channels in the Claude Partner Network announcement, with approximately 80% of revenue generated by enterprise API usage. We decline-to-publish a precise Claude.ai consumer monthly-active-user figure pending a primary disclosure.
- Headcount (basis-disclosure note): Anthropic does not publish precise headcount. We decline-to-publish a precise figure pending a primary disclosure and reference only that the company has expanded its research, product and go-to-market organisation materially through 2025-2026 alongside revenue growth. Anthropic’s Series G announcement references continued team-build but does not publish a headcount number.
- Funding to date: Cumulative external equity well above $70B through the Series G, excluding the $100B / 10-year AWS spend pledge which is framework-spend rather than equity. References: Anthropic Series G $30B / $380B post-money; Amazon Newsroom up-to-$20B 2026 expansion on top of the prior $8B Amazon stack (two $4B tranches in Sep 2023 / Mar 2024 and Nov 2024); Series E $3.5B at $61.5B post-money, March 2025; Google strategic ~$2B (February 2024).
Methodology & Disclaimer
For metric definitions, source-tier hierarchy, and decline-to-publish rules, see the tracker methodology. Confidence dots (• green / • amber / • red) follow the same convention as the AI Tracker.
Spotted a figure you believe is wrong? Send corrections to info@informationmatters.net.
Information Matters Framework scores are the considered opinion of the IM team — human and AI — applied to publicly-available evidence under a disclosed methodology. They are not statements of fact about the companies scored and they are not investment advice.
