Last fall, Anthropic was playing second fiddle to OpenAI. It had a lower valuation, while OpenAI was still drawing much of the attention as the first mover in the generative AI boom. But the dynamic shifted in late November, when Anthropic released Claude Opus 4.5, which gave the company’s Claude Code coding agent a new brain and helped elevate it to the status of “AI killer app.” Arguably, that was the moment that set Anthropic on its path toward a forthcoming initial public offering (IPO).
Developers had been using Claude Code to build software for much of 2025, but the tool had shown more promise than truly game-changing results. Opus 4.5 gave Claude Code the intelligence to build an app or feature from end to end, based only on plain-language planning and guidance prompts from the user.
Opus 4.5 enabled longer-running agents and better planning and execution workflows. Claude Code got better at discussing a project with an engineer-user, presenting a plan, incorporating feedback, and then carrying out a focused set of multistep tasks to complete a software build. Anthropic sweetened the deal by imposing fewer usage caps, which makes a big difference for software engineers who spend their days deploying multiple agents to build different parts of a project.
Just as important, Anthropic changed the way software engineers access Claude Code. Instead of using the tool through their machine’s command line interface, they could now select the Code tab in the Claude desktop app, which includes its own integrated terminal. Under that tab, the company also brought together many of the resources developers normally use, including a file editor for viewing and editing code, code-change review windows, parallel coding sessions for different tasks, and other tools. Anthropic turned Claude Code from a terminal/chat tool into something closer to a full desktop coding environment.
“Claude Code is excellent at fixing all those small bugs from old projects. . . . ” tweeted Pieter Levels, an influential Dutch entrepreneur and self-taught developer. “[B]efore I’d not have the time to fix these kinds of projects . . . but now it takes me an hour to do this and it works again!”
Claude’s beachhead
The new unified interface in the Claude desktop app invited software engineers to do their nonsoftware development work in the same place. With easy access to Claude chat and CoWork, engineers could use Claude for more general tasks, including conducting research, accessing company data, composing emails, and styling presentations.
The interface also accommodates workers who code only occasionally, perhaps to slap together a prototype app to present to colleagues, but who benefit from doing much of their day-to-day work with help from Claude chat and CoWork’s data connections, skills, and workflows.
Acceleration
Whatever the motivation, enterprises are indeed choosing Anthropic, and the company’s accelerating enterprise momentum can be traced back to Claude Opus 4.5. At the end of 2025, Anthropic reported just $9 billion in annualized revenue run rate. (ARR extrapolates a recent month’s revenue out to a full year.) That number grew rapidly in 2026 as Claude Code proliferated. The company reported $14 billion in ARR in February and $30 billion in April. A May Reuters report pegged the company’s ARR at $47 billion. For context, OpenAI said at the end of March that it was generating about $2 billion per month in revenue, which works out to ARR of $24 billion.
Anthropic reportedly said that more than 1,000 companies are now paying over $1 million per year for Claude, and that this number has more than doubled since February 2026. The company has announced major deployments with PwC, Allianz, Snowflake, Accenture, Deloitte, and IBM. Anthropic expects to make an operating profit of $559 million during the quarter ending in June, on revenue of $10.9 billion, according to The Wall Street Journal.
While Anthropic does not release raw user numbers for Claude Code, survey data suggests the tool has plenty of room to grow. An early-April JetBrains survey of 10,000 software developers globally found that 29% used GitHub Copilot, while only 18% used Claude Code, tied with Cursor.
Stage is set
On Monday, Anthropic submitted a confidential draft registration statement, or S-1, to the Securities and Exchange Commission (SEC). Under U.S. rules, the SEC can review IPO-seeking companies privately, allowing a company to delay making detailed financials public until later, closer to the actual listing.
Anthropic’s could end up being the second of three big artificial-intelligence-related IPOs in 2026. SpaceX has already filed, and OpenAI is expected to file later in the year. Both Anthropic and OpenAI have a chance to IPO at trillion-dollar-plus valuations, putting them among the largest tech IPOs in history. Anthropic just last week raised $65 billion in new financing at a valuation of $965 billion, including the new money.
Experts are not worried about whether enterprise software engineering groups will adopt Claude Code. They believe the demand is real and growing. The more serious threat, for both Anthropic and its customers, is the cost of using the tool. Part of Claude Code’s appeal is that it develops a deep understanding of code bases and complex coding problems. That requires the agent to reason across large context windows, which means using a lot of tokens, the chunks of text, data, and code it processes.
Corporate users, including Uber, are already running up big bills as engineers max out on tokens. Anthropic, meanwhile, is trying to find additional computing resources to process all the tokens being generated by Claude Code users. The company is even paying Elon Musk $1.25 billion a month for the use of the xAI/SpaceX Colossus 1 data center in Memphis.
When Anthropic’s prospectus shows up, likely sometime this summer, we’ll finally learn more about the real profitability of selling Claude Code.