Anthropic’s public IPO filing reveals a company betting that AI coworkers will spread across knowledge work, supported by enormous commitments to computing capacity.
Its growth is already substantial, but realizing that vision depends on businesses learning to put increasingly autonomous AI to work.
Anthropic, the company behind Claude, has filed IPO paperwork with the SEC to offer shares to the public. The document, called an S-1, describes a future built around virtual collaborators: AI systems that perform complex work using a company’s data, software, and permissions. Anthropic has applied to list on Nasdaq as ANT, but has not disclosed a share price.
The filing reports $16.3 billion in first-half 2026 revenue, compared with $1.2 billion a year earlier. About $72.4 billion of its $80.7 billion net loss came from non-cash accounting charges, driven mainly by changes in the value of loans convertible into shares, not cash spent operating the company. Anthropic reported a $974 million operating profit in the second quarter, though an operating loss for the first half overall.
Alongside that growth are multiyear cloud commitments whose minimum payments generally apply regardless of usage, subject to limited exceptions.
SmarterX founder and CEO Paul Roetzer examined how that spending connects to Anthropic’s ambitions, on Episode 249 of The Artificial Intelligence Show.
$16.3 billion - Anthropic’s revenue in the first half of 2026
$8.9 billion - First-half compute and infrastructure expenses
$114.5 billion - Minimum Google commitment through 2033
$110 billion - Minimum Amazon commitment through 2036
$44.9 billion - Minimum Microsoft commitment through 2033
The next market is the rest of the company. Software development has driven much of Anthropic’s growth, but Roetzer sees a much larger expansion still ahead: “We haven’t really crossed over to where there’s massive demand on marketing and sales and operations and HR and finance and legal and all the other domains,” he says.The filing’s vision depends on those teams using AI for substantial, ongoing work.
Coding becomes a capability people use without seeing it. Claude Cowork extends the approach behind Claude Code to tasks such as reading files, navigating applications, and producing finished work. A professional requesting a presentation or report does not need to understand the code underneath.
“You never know it happens,” Roetzer says, referring to the coding going on in the background.
Useful AI coworkers need organizational context. Roetzer emphasizes the need to provide AI agents, systems that take multiple steps to complete a task, with customer records, financial systems, company knowledge, and information about how decisions get made. Connecting those resources helps AI work more accurately. But getting enterprises ready to do this remains a major constraint on the market Anthropic envisions.
“Most companies still have no idea what to do with this stuff, especially when we start getting into the agents and truly autonomous agents and agent swarms or groups of teams of agents.”
— Paul Roetzer, founder and CEO of SmarterX, Episode 249 of The Artificial Intelligence Show
For business leaders, the filing offers a preview of the capabilities Anthropic wants organizations to use. Anthropic is building toward AI that can take on open-ended assignments for hours, days, or weeks. Making that useful requires deciding which work to delegate and which information an AI coworker needs.
Start with a defined workflow and the people who understand it. Identify the necessary systems, permissions, context, and review responsibilities. That preparation connects the promise of an AI coworker to an actual business process and gives a team a way to assess whether it can rely on the results.
Enterprise adoption must support computing commitments made years ahead. Beyond the cloud minimums, Anthropic has about $161.2 billion in five-year Broadcom equipment-lease payments, excluding interest.
Some contracts allow more flexibility. Its up-to-$120.5 billion SpaceXAI capacity agreements through 2029 let either party cancel with 90 days notice after initial lock-in periods ending in early 2027. How quickly businesses adopt AI coworkers will help determine whether the capacity supports growth or becomes a financial burden.
Only 25% of respondents say their organization has reached the Scaling phase of AI adoption, according to the 2026 State of AI for Business Report. Scaling means achieving widespread adoption while consistently improving efficiency and performance. That finding gives context to the enterprise challenge Roetzer describes: buying access to capable AI is one step in building an organization that can use it effectively.
Based on more than 2,100 responses across roles, functions, and industries, the report examines adoption, training, governance, and organizational readiness. Use it to compare your company’s progress with peers and identify the foundations that need attention as AI takes on more work. Read the full report →