In December 2024, Anthropic's annualised revenue run rate was $1 billion. By May 2026, it was $47 billion — a 47x increase in seventeen months, the fastest sustained revenue growth any enterprise software company has ever recorded, eclipsing Salesforce's twenty-year climb to the same number. Claude Code became the most-used AI coding tool in the world within eight months of launch. The Series H round — $65 billion at a $965 billion valuation — is the largest private funding round in history. None of this happened by accident, and none of it is free of the same structural risk Letter 113 examined in OpenAI's Stargate: Amazon and Google are simultaneously Anthropic's largest investors and its primary compute suppliers, and the company has publicly admitted its own infrastructure has, at times, not kept pace with demand.">
Not investment advice. This letter discusses Anthropic, the company that creates the AI model used to research and draft this letter series — a conflict of interest the reader should weigh accordingly. Data sourced from Sacra company analysis June 2026, Anthropic official funding announcements (Series G February 2026, Series H May 2026), VentureBeat May 2026, Forbes May 2026, TechCrunch May 2026, Ramp Economics Lab. All figures current as of June 2026.
Every letter in this series is researched and drafted using Claude, the AI model built by Anthropic. This letter is about Anthropic itself — which means the author has a structural relationship to its subject that no other letter in this series carries, and that relationship should be stated plainly rather than left for the reader to discover. The standard this letter holds itself to is the same one Letter 113 applied to OpenAI's Stargate: real growth gets reported as real growth, and real structural risk — including risk that applies to Anthropic in close to the same form it applied to OpenAI — gets reported with the same weight, not softened because of who is doing the reporting.
With that stated, the numbers themselves are extraordinary enough that they do not require any flattering framing to be newsworthy. Anthropic's annualised revenue run rate grew from approximately $1 billion in December 2024 to roughly $47 billion in May 2026 — a 47-fold increase in seventeen months. For comparison: Salesforce, widely regarded as one of the fastest-growing enterprise software companies in history before this period, took approximately twenty years to reach $30 billion in annual revenue. Anthropic passed that figure in under three years from its 2021 founding, and nearly all of that growth has come in the eighteen months since Claude Code's launch.
January 2024 to May 2026 · Sacra, VentureBeat, company disclosures
The growth has been substantially concentrated in two products. Claude Code — Anthropic's agentic coding tool — launched publicly in February 2025, reached general availability in May 2025, hit $1 billion in annualised revenue by November 2025, and crossed $2.5 billion by February 2026, more than doubling again since the start of the year. Within eight months of launch, Claude Code overtook both GitHub Copilot and Cursor to become the most-used AI coding tool in the world. Enterprise use now accounts for more than half of all Claude Code revenue, with named customers including Netflix, Spotify, KPMG, L'Oréal, and Salesforce — and Deloitte's deployment of Claude across more than 470,000 employees stands as the largest single-provider enterprise AI deployment recorded anywhere.
The second engine is broad API and enterprise platform revenue. Anthropic now counts more than 300,000 business customers, accounting for roughly 80% of total revenue, with over 1,000 customers individually spending more than $1 million annually — a figure that doubled in under two months as of April 2026, up from a dozen such customers two years earlier. According to Ramp Economics Lab, Anthropic's share of combined enterprise spending on the two leading frontier labs (itself and OpenAI) rose from roughly 10% at the start of 2025 to over 65% by February 2026 — a genuine, measurable share shift in head-to-head enterprise competition, not merely Anthropic's own self-reported growth in isolation.
The Series H investor list reads as a near-complete roster of the world's largest institutional capital pools: Altimeter, Dragoneer, Greenoaks, Sequoia Capital, Capital Group, Coatue, D1 Capital Partners, Baillie Gifford, Blackstone, Brookfield, D.E. Shaw, DST Global, and Fidelity, among others — alongside strategic infrastructure partners Samsung, SK Hynix, and Micron, reflecting the round's explicit purpose of funding the physical compute buildout behind Claude's growth. Sequoia Capital's participation is notable on its own terms: the firm backs both Anthropic and OpenAI simultaneously, breaking what had long been treated as an unwritten venture capital rule against funding directly competing frontier labs — itself a signal of how large and how separately defensible both companies' addressable markets are now considered to be by sophisticated capital allocators.
Anthropic has reportedly engaged Wilson Sonsini to prepare for a public listing, with internal discussions reportedly centring on an October 2026 target and Goldman Sachs, JPMorgan, and Morgan Stanley already engaged. No prospectus has been filed and no timing has been formally confirmed as of this letter's publication — readers should treat the IPO timeline as a reported expectation, not a committed event.
Amazon is Anthropic's largest single strategic investor and its primary cloud and training partner. Anthropic has committed more than $100 billion over ten years to AWS technologies in exchange for up to 5 gigawatts of compute capacity, with nearly 1 gigawatt of Trainium2/Trainium3 capacity expected online by the end of 2026. Claude runs on AWS Bedrock, and Amazon's $5 billion immediate investment alongside up to $20 billion in further committed capital makes this the single most structurally significant relationship in Anthropic's entire compute strategy.
Google is Anthropic's second-largest strategic investor, offering Claude via Google Cloud Vertex AI while separately supplying next-generation TPU compute capacity through an agreement with Broadcom for up to 5 gigawatts. Reuters reported in April 2026 that Google was considering investing up to a further $40 billion in Anthropic — a figure that, if it materialises, would make Google's combined investor-and-supplier relationship with Anthropic comparable in scale and structure to Amazon's.
Beyond its two primary partners, Anthropic has diversified its compute base further than most frontier labs — agreements with Microsoft (whose Azure now also hosts Claude, making it the only frontier model available on all three major cloud platforms simultaneously), Nvidia, and, notably, a compute capacity arrangement with Elon Musk's SpaceX for access to GPU capacity in the Colossus 1 and Colossus 2 clusters. One industry observer's characterisation of the SpaceX arrangement — "Elon's enemy is Sam, Dario's enemy is Sam, enemy of my enemy is a compute partner" — captures something genuinely true about the current alignment of interests among Anthropic, xAI, and the broader competitive field arrayed against OpenAI specifically.
In April 2026, Anthropic published a postmortem acknowledging that three separate bugs had affected Claude Code's reliability since early March, and that internal testing had not caught them — leading to several weeks of degraded performance for users during a period of peak demand. The company separately stated that demand for Claude had led to "inevitable strain on our infrastructure," directly impacting reliability and performance, particularly during peak hours. CEO Dario Amodei addressed this directly at the company's Code with Claude developer conference, stating the company was "working as quickly as possible" to expand capacity and would "pass that compute on to you as soon as we can."
This is a genuinely important data point precisely because it complicates the otherwise uncomplicated growth narrative: a company whose revenue grew 47-fold in seventeen months was, during a meaningful stretch of that same period, unable to fully and reliably serve the demand that growth represented. Most of the new compute capacity secured through the Amazon and Google agreements examined above is not expected to come fully online until late 2026 or early 2027 — meaning the infrastructure strain experienced in early 2026 may not be fully resolved for another twelve to eighteen months from this letter's publication, even as revenue continues compounding against that same capacity ceiling.
The growth-rate comparison to Google's early advertising business or Salesforce's enterprise software climb, while numerically accurate, obscures a meaningful structural difference: Anthropic's revenue is heavily concentrated in a single product category — agentic coding — at a moment when at least three credible, well-funded competitors (GitHub Copilot backed by Microsoft, Cursor, and OpenAI's own Codex-derived tooling) are pursuing the identical market. Claude Code's current dominance is real and well-documented, but "fastest growth to date" and "durable long-term market leadership" are different claims, and the gap between them is precisely where the most credible risk to Anthropic's current valuation trajectory sits. Ramp's data showing Anthropic's enterprise spending share rising from 10% to 65% against OpenAI specifically is a genuinely strong signal — but it measures share within a two-company comparison, not durability against the full competitive field, including well-resourced entrants this letter series has not yet covered in depth.
The valuation trajectory — six rounds, fifteenfold increase, in fourteen months — is, by any historical standard, an extraordinary pace of capital formation, and extraordinary pace is not, on its own, evidence of mispricing, but it is exactly the kind of pattern that warrants the same scrutiny this letter series applied to OpenAI's own valuation climb in Letter 113. A $965 billion valuation implies a degree of confidence in sustained, multi-year, category-leading growth that has historically been difficult for any single technology company to deliver without at least one meaningful setback along the way — and Anthropic, alongside every other frontier AI lab examined in this letter series, has not yet been tested by a genuine industry-wide demand contraction or a sustained period of underperformance relative to its own prior guidance.
The infrastructure strain admission is the single most important data point in this letter for a reader trying to separate durable signal from promotional noise — precisely because it is the one fact in this entire letter that the company had no commercial incentive to disclose, and disclosed anyway. A company publicly acknowledging that its own product reliability suffered because its growth outpaced its compute capacity is, in the framework this letter series has applied consistently since Letter 110's corporate debt analysis, a genuinely useful signal — not because it undermines the growth story, but because it confirms the growth story is real enough to have created a real physical constraint, rather than being a purely financial-engineering phenomenon decoupled from underlying demand.
Long-horizon thinking on capital, technology, and the forces shaping the next decade of wealth creation. Written from first principles. Not consensus. Not noise.