The Upgrade That Changed the Trajectory
For most of 2025, Anthropic was the credible alternative to OpenAI — well-regarded among researchers and safety-focused enterprises, but trailing in valuation and market attention. That changed in late November when the company released Claude Opus 4.5 and handed Claude Code a meaningfully better brain.
The upgrade wasn't incremental. Opus 4.5 gave Claude Code the ability to reason across large context windows, hold a planning conversation with an engineer, incorporate feedback, and then execute a focused sequence of tasks to complete a software build from scratch. Developers no longer had to babysit the tool through each step. They could describe what they wanted in plain language and come back to a working feature.
Anthropics also removed a friction point that mattered operationally: it loosened usage caps, which is significant for engineers running multiple agents simultaneously across different parts of a project.
From Terminal Tool to Desktop Environment
The interface change was as important as the model upgrade. Anthropic moved Claude Code out of the command line and into a dedicated tab in the Claude desktop app, bundling in a file editor, code-change review windows, and parallel coding sessions. It stopped being a chat-adjacent utility and started functioning as a primary development environment.
That consolidation had a secondary effect: it pulled engineers' non-coding work into the same interface. Research, email drafting, data access, presentation work — all of it became available through Claude chat and CoWork in the same window where the code was being written. The product became stickier for a broader slice of the enterprise workforce, not just senior engineers.
Revenue Numbers That Are Hard to Ignore
The business results followed the product logic. Anthropic ended 2025 with $9 billion in annualized revenue run rate. By February 2026 that figure had reached $14 billion. By April, $30 billion. A May Reuters report put it at $47 billion — nearly double OpenAI's reported $24 billion ARR at the end of March.
More than 1,000 companies are now paying over $1 million annually for Claude, a cohort that more than doubled in size between February and June 2026. Enterprise deployments include PwC, Allianz, Snowflake, Accenture, Deloitte, and IBM. The Wall Street Journal reported Anthropic expects $10.9 billion in revenue and $559 million in operating profit for the quarter ending June 2026.
The IPO Setup
Anthropics filed a confidential draft S-1 with the SEC in early June, beginning the formal IPO process. Under U.S. rules, the company can keep detailed financials private until closer to the actual listing. The company raised $65 billion in new financing last week at a valuation of $965 billion including the new capital — positioning it alongside OpenAI as a potential trillion-dollar IPO, among the largest tech listings in history.
The Cost Problem Underneath the Growth
The demand story is real. The cost story is the one to watch.
Claude Code's value proposition — deep understanding of complex codebases, long-context reasoning — is computationally expensive. Every useful session burns through tokens at scale. Corporate users including Uber are already hitting significant bills as engineers push usage limits.
To keep up with demand, Anthropic is paying $1.25 billion per month for access to xAI's Colossus 1 data center in Memphis. That's a substantial line item against a quarterly revenue figure of $10.9 billion, and it's the number that will draw the most scrutiny when the full prospectus becomes public — likely sometime this summer.
A JetBrains survey of 10,000 developers in early April found Claude Code at 18% adoption, tied with Cursor and trailing GitHub Copilot at 29%. There's room to grow. The question the S-1 will have to answer is whether Anthropic can grow into its compute costs faster than those costs grow into its margins.