Anthropic
Usage-based API (pay-per-token) + Claude.ai consumer/Pro subscriptions + Claude Code agentic coding; sold direct and via AWS Bedrock / Google Vertex / Microsoft resale (revenue booked gross).
Priced primary rounds from Series C on are well-disclosed; 2021 Series A ($124M) and 2022 Series B ($580M, incl. FTX) are omitted because their post-money valuations were never disclosed, as is the Oct-2023 Google round (only a rumored $20-30B range).
The thesis on this name
State of AI Compute
Frontier AI safety lab behind the Claude model family and Claude Code; #1 most-valuable private AI startup, confidentially filed for a ~Oct-2026 IPO.
State of Frontier AI
Enterprise-first frontier lab that overtook OpenAI on enterprise AI spend share (34.4% vs 32.3%, Apr'26) and run-rate revenue (~$47B vs ~$25B, May'26), led by Claude Code's category-defining developer traction ($2.5B+ run-rate by Feb'26, fastest enterprise-software ramp on record). 8 of the Fortune 10 are customers; 500+ accounts spend >$1M/yr. The durability case is the enterprise lock-in + coding-agent moat, not consumer mindshare.
Earnings, margins, COGS & capex
Hyper-growth, pre-profit foundation-model lab. Disclosed ~$47B annualized run-rate by mid-May 2026 vs ~$9–10B end-2025, driven by enterprise API and Claude Code (>$2.5B run-rate, >half of enterprise spend). ~80% of revenue is business customers; >1,000 accounts spend >$1M/yr. Revenue from cloud resellers is booked GROSS (inflates top line vs net-reporting peers). Gross margin ~40% in 2025 (below prior target); burn falling; first operating profit guided Q2 2026; no audited financials (confidential S-1 only).
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~60¢ is cost of goods and ~38¢ operating expense, leaving ~2¢ of operating profit.
Revenue trend
Margins
improving from ~-94% (2024); cut vs prior target; ~77% targeted 2028
improving (-$5.6B 2024 → -$3B 2025); FCF+ ~2027
first operating profit guided Q2 2026 (press)
COGS structure
Dominated by inference + training compute (AWS Trainium, Google TPUs, NVIDIA GPUs) plus reseller payouts (gross-booked). Free users ~95% of consumer base make inference a cost drag; margin gated by chip cost/efficiency.
Capex
Compute is contractual, not owned: Amazon up to 5GW / ~$25B (Trainium2/3 online through 2026); Google·Broadcom up to 1M TPUs / ~$40B, >1GW online 2026. Demand ~1GW 2026 → 3GW+ 2027.
Latest earnings
N/A (private, no consensus)
Internal projection ~$70B revenue by 2028 (TechCrunch, Nov 2025); gross margin to ~77% by 2028; first operating profit Q2 2026 per press
- Annualized run-rate
- ~$47B (mid-May 2026)
- Post-money valuation
- $965B (Series H)
- Series H raise
- $65B
- Business-customer share of revenue
- ~80%
- Claude Code run-rate
- >$2.5B
- $1M+/yr accounts
- >1,000
Growth drivers
- Claude Code agentic coding — >$2.5B run-rate, >half of enterprise spend, customers incl. Netflix/Spotify/KPMG/L'Oréal/Salesforce
- Enterprise API via direct + AWS Bedrock (100k+ customers) / Google Vertex
- High-value accounts: >1,000 spending >$1M/yr (2x in <2 months); >$100k accounts up 7x YoY
- Claude.ai Pro/consumer subscriptions; use-case diversification beyond coding
- Frontier model cadence (Claude Opus/Sonnet) sustaining quality lead in code + agents
Bull & bear
The fastest-scaling enterprise software franchise ever, leading the highest-value AI use case (agentic coding) with improving margins and a lower burn than its arch-rival — about to IPO into the AI supercycle.
- Run-rate compounding faster than any company in history; ~$70B revenue plausible by 2028
- Claude Code is a genuine category killer with >half of enterprise spend and Fortune-500 logos
- Enterprise mix + $1M+ account growth signals durable, expanding net revenue retention
- Capital efficiency edge: ~$2.10 revenue per $ of compute targeted by 2028 vs OpenAI ~$1.60
- Multi-cloud + custom-silicon roadmap (Trainium3/TPU) is a credible path to ~77% gross margin
- IPO + ~$132B raised gives balance-sheet firepower to outlast a capex war
A ~$965B valuation (~20x run-rate, far higher on net revenue) on a pre-profit, ~40%-gross-margin company whose top line is gross-inflated, concentrated in one use case, and structurally dependent on two giants who are also its rivals.
- Headline run-rate flatters reality: gross reseller booking + run-rate (not trailing) revenue overstate the base
- ~40% gross margin (cut from target) on a commoditizing token market — margin may not reach 77%
- Concentration: Claude Code / coding is the engine; a competitive or pricing shock hits the core
- Amazon + Google are investor, supplier, AND competitor — strategic dependence cuts both ways
- OpenAI is better-funded ($122B round) and larger by some revenue measures; arms race is brutal
- Valuation prices flawless execution; any growth deceleration or AI-capex pullback re-rates hard
- No audited financials — profitability and margin claims are projections until the public S-1
What it is worth
Private mark anchored to Series H ($965B post-money, $65B raise, ~Apr 21/announced May 28 2026); cross-checked vs OpenAI ($852B) and run-rate multiple. Reverse sanity: $965B / ~$47B run-rate ≈ 20x run-rate (far higher on trailing/net revenue), implying years of ~flawless hyper-growth + margin expansion to ~77% by 2028.
$400–600B re-rate
if token prices commoditize, gross margin stalls near ~40%, Claude Code growth slows, or an AI-capex / sentiment drawdown compresses multiples across the complex.
~$965B–$1T (last-round mark holding into an Oct-2026 IPO): growth decelerates from parabolic but stays exceptional; margin improves gradually; market keeps Anthropic #1 private AI co.
≥$1.2T at/after IPO if run-rate sustains toward $70B+ by 2028, gross margin marches to ~77%, and the AI-capex bid holds — agentic coding compounds and OpenAI fails to displace.
No audited financials (confidential S-1 only). Headline run-rate is gross-booked and run-rated, so the true revenue base is lower — adjust the multiple upward accordingly. Private mark; not a public-market price.
SWOT
Strengths
- Fastest revenue ramp of any company on record (~$9B → ~$47B run-rate in ~5 months)
- Clear category lead in agentic coding (Claude Code) — the highest-monetizing AI use case
- Enterprise-heavy mix (~80% business revenue; >1,000 $1M+ accounts) = stickier than consumer
- Multi-cloud compute (AWS Trainium + Google TPU + NVIDIA) reduces single-vendor lock-in
- Brand as the 'safety-first / trustworthy' frontier lab — resonates with regulated enterprises
- Lower historical burn than OpenAI (~$3B 2025 vs OpenAI ~$14B 2026E) — more capital-efficient growth
Weaknesses
- Still pre-profit with only ~40% gross margin (cut from prior target); margin gated by compute cost
- Revenue concentrated in Claude Code / coding — single dominant use case
- Top-line flattered by GROSS reseller booking — net revenue is lower than headline
- Deep dependence on Amazon + Google, who are simultaneously investors, suppliers, and competitors
- No audited financials public; key profitability claims are projections, not verified results
- Talent + compute cost inflation in an arms race with far larger-capitalized rivals
Opportunities
- IPO (~Oct 2026) unlocks public capital to fund multi-GW compute buildout
- Agentic AI / autonomous software engineering TAM still early — Claude Code can expand seat + usage
- International + vertical expansion (currently US-skewed ~25% of traffic)
- Enterprise platform / tooling layer (MCP, agents) deepening lock-in beyond raw API
- Margin expansion as custom silicon (Trainium3, TPU) lowers inference cost per token
Threats
- OpenAI ($852B, ~$25B revenue, $122B megaround w/ Amazon/Nvidia/SoftBank) — best-funded direct rival
- Google (Gemini) and Amazon — partners that can in-source / favor own models
- Commoditization / price war on tokens compressing margins
- Frontier-model regulation, copyright litigation, safety/liability exposure
- Compute supply + power constraints capping the 3GW+ 2027 ramp
- A market/AI-capex drawdown re-pricing the entire AI-infrastructure complex
Moats, dependencies & bottlenecks
Moats
Best-in-class for code/agents today, but model leadership is contestable each cycle.
100k+ Bedrock customers, >1,000 $1M+ accounts; integrations/agents raise stickiness.
Trust matters to regulated buyers; intangible and imitable.
Multi-GW Amazon+Google deals are real scarcity, but rented from rivals.
Frontier talent is mobile and aggressively poached.
Dependencies
Investor + cloud/training supplier (Trainium) + Bedrock distribution + competitor Up to 5GW / ~$25B; ~$8B+ prior investment; primary cloud. Also ships own models.
Investor + TPU compute supplier + Vertex distribution + competitor (Gemini) Up to ~$40B / up to 1M TPUs; structural reliance on a direct rival's silicon.
GPUs remain part of the multi-chip stack; supply/price exposure.
Custom-silicon / TPU co-design partner Named in the Google·Broadcom multi-GW TPU deal.
Funding for multi-GW capex Buildout needs continued mega-financing; sensitive to AI-sentiment cycles.
Advantages
- #1 in agentic coding with the highest-monetizing AI use case
- Most capital-efficient hyper-scaler of revenue (vs OpenAI burn)
- Deep enterprise penetration and high-value-account growth
- Multi-cloud, multi-silicon flexibility
- Safety/trust brand with regulated buyers
Weaknesses
- Pre-profit with ~40% gross margin
- Gross-booked top line overstates net revenue
- Use-case concentration in coding
- Strategic dependence on Amazon + Google
- No audited public financials yet
Bottlenecks
- Compute supply + data-center power for the 3GW+ 2027 ramp
- Gross margin / inference cost per token
- Frontier-talent acquisition and retention
- Capital intensity vs path to FCF positivity (~2027)
- Model-quality cadence vs OpenAI/Google each release cycle
Top signals & trends
Top signals
Public capital + liquidity event; validates scale, but exposes audited numbers.
Unprecedented compounding; watch for run-rate vs trailing gap.
Margin path harder than guided; compute cost pressure real.
Market crowns Anthropic #1 private AI co.
Land-and-expand working at the high end.
Trends
Directly expands Claude Code's seat + usage TAM.
Enables scale and margin, but ties Anthropic to rivals' chips and huge spend.
Pressures gross margin on the API business.
Compliance + legal overhang for frontier labs.
Shifts spend from experiments to production budgets — favors enterprise-heavy mix.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Primary cloud + training silicon; up to 5GW / ~$25B; also investor.
Up to 1M TPUs / ~$40B; also investor.
GPUs across the multi-chip training/inference stack; ~1GW Grace Blackwell / Vera Rubin committed (Nov 2025) — also a ~$10B investor.
Custom TPU/accelerator co-design in the Google·Broadcom deal.
Azure compute — $30B commitment + up to 1GW (Nov 2025); also a ~$5B investor. Claude now runs on all three hyperscalers.
In talks (Jul 2026) to fabricate Anthropic's first custom AI chip on Samsung's 2nm foundry — early-stage, specs undefined; ex-OpenAI chip lead hired.
Leases all of xAI's Colossus 1 (Memphis — 220k+ Nvidia GPUs, >300MW) at ~$1.25B/month through 2029 (~$15B/yr); disclosed in SpaceX's pre-IPO S-1 (May 2026). Interest in orbital compute — despite xAI/Grok being a Claude rival.
In talks (reported 17 Jul 2026): Anthropic would lease up to $10B of compute from Meta over ~2yr — monthly payments, early-exit either side, unsigned. Would make Anthropic the anchor tenant of Meta's nascent Meta Compute line.
Named Claude Code / enterprise customer.
Named Claude Code customer.
Named Claude Code customer.
Named enterprise customer (non-US listing — analysis only, not a recommendation).
Named professional-services Claude Code customer.
Claude is a leading selectable model in the Cursor AI IDE / agent (one of several providers).
Claude is the primary model powering Replit Agent (code generation).
Notion AI Agents/Teammates are Claude-powered (multi-model).
Supports Claude in its model-agnostic enterprise assistant/agent platform.
Runs Claude within a multi-model legal-AI architecture (deepening Anthropic partnership).
Closest rival; ~$852B valuation, ~$25B revenue, $122B megaround (Amazon/Nvidia/SoftBank). ChatGPT + Codex compete head-on.
Investor + TPU supplier that also ships a directly competing frontier model + Vertex platform.
Distributes Anthropic on some surfaces but anchors the OpenAI/Copilot stack against it.
Open-weight models pressure pricing and the low end of the API market.
Investor + supplier that also fields its own foundation models on Bedrock.
Well-funded frontier entrant competing for talent, compute, and mindshare.