Anthropic Aims to Claim Over $30 Trillion Annual Revenue Opportunity for Its AI to IPO Investors
Anthropic plans to assert to IPO investors that its AI serviceable annual revenue opportunity exceeds $30 trillion, calculated based on the total amount of human work that the model can theoretically accomplish, rather than traditional software subscription or cloud service spending markets.
Reportedly, the company is considering raising up to $100 billion at a valuation of around $2 trillion; if this financing materializes, it will surpass SpaceX's previous IPO financing record of about $86.2 billion. The aforementioned plans are still in the preparation and banker discussion stage and have not been submitted or priced for public offering terms.
The $30 trillion TAM is not a forecast of Anthropic's near-term revenue but rather a theoretical revenue ceiling if the company captures all work that can be done by AI. This calculation path differs from traditional bottom-up market size calculations: the latter counts enterprise software, cloud computing, or AI budgets, while the former incorporates the labor value of tasks that can be automated into the serviceable market.
Anthropic's annualized revenue run rate exceeded $65 billion by the end of July, growing more than 7 times from about $9 billion at the end of 2025; comparing $65 billion to $30 trillion, the company's current theoretical market penetration rate is about 0.22%. This revenue is a run rate rather than audited full fiscal year revenue, and different media have varying interpretations of its statistical timing and criteria.
The company's internal target points to a revenue range of $190 billion to $200 billion by 2028. Even if it reaches $200 billion, based on the $30 trillion theoretical TAM, the penetration rate would be about 0.67%; however, whether the revenue target is achieved depends on corporate procurement, model performance, inference costs, competitive landscape, regulatory requirements, and the proportion of AI savings that customers convert into vendor spending.
Anthropic's market size argument is said to be partly based on its Economic Index framework: this framework maps model usage to approximately 20,000 occupational tasks in the U.S. Department of Labor's O*NET database and estimates that the proportion of work that AI can theoretically handle in occupational categories such as computer, mathematics, and finance can exceed 80%, with some indicators approaching 94%. Tasks that can be assisted or completed by the model do not equate to the corresponding job wages being fully converted into Anthropic's revenue.
In market mechanisms, the $30 trillion narrative shifts investor attention from current revenue multiples to labor substitution space: if corporate buyers aim to reduce costs in customer service, coding, documentation, analysis, and backend processes, they will shift budgets from wages, outsourcing, and traditional software to model invocation, AI applications, and computing power; the beneficiaries are companies that control models, GPUs, cloud computing, and enterprise distribution channels, while the pressured parties are repetitive white-collar services and low-value outsourcing that can be standardized and split. Whether model suppliers can capture most of this value depends on whether they have customer relationships, workflow control, and ongoing pricing power. fortune+1
Source: Public Information
ABAB AI Insight
Anthropic's $30 trillion figure continues its own labor task research rather than extrapolating from software budgets. The Anthropic Economic Index previously used Claude.ai dialogue samples and the O*NET occupational task database to attempt to measure AI usage distribution across different professions; this framework focuses on task exposure rather than the proportion of jobs being replaced. Expanding task feasibility into TAM translates "potential productivity space" into "potential vendor revenue space." There remains a gap between this and corporate process transformation, customer bargaining, competitive profit sharing, and regulatory costs.
The capital path is already very clear: Anthropic aims to raise large-scale financing to purchase training and inference computing power, integrate the Claude model into enterprise workflows, and then attempt to convert a portion of wages, consulting fees, outsourcing costs, and software budgets into model usage fees. Unlike Salesforce charging per seat or AWS charging based on computing resources, Anthropic's end narrative is to charge based on the value of work that is replaced or enhanced. The upper limit of this model is high, but it also means that customers will demand to renegotiate prices based on saved labor costs, task completion quality, and business outcomes, rather than permanently accepting high-margin generic token charges.
SpaceX previously proposed about $28.5 trillion as the "largest serviceable market in human history" in its IPO documents, most of which is related to AI and space infrastructure; Uber also supported its long-term narrative with approximately $6 trillion in the global mobility market in its 2019 IPO materials. The similarity lies in explaining high valuations with giant TAMs, while the difference is that Anthropic has already achieved a $65 billion annualized revenue run rate and has a near-term anchor of $190 billion to $200 billion by 2028. The real capital market question is not whether the $30 trillion is mathematically valid, but how much revenue, margin, and free cash flow this company can retain after competition and customer bargaining.
Essentially, it is about the transfer of pricing power: in the past, companies broke down knowledge labor into wages, outsourcing contracts, SaaS licenses, and professional service fees; AI vendors attempt to re-aggregate these expenditures into model inference and automated workflow revenue. Pricing power does not automatically arise because a model can complete tasks; it only stabilizes as revenue for model companies when they control task entry points, data context, result accountability, and switching costs. Otherwise, the gains will be shared among customers, open-source models, cloud platforms, and vertical applications.
ABAB News · Cognitive Law
Task automation does not equate to value being collected by a single entity
Market upper limits tell stories, cash flow upper limits set prices
Profits from substituting labor depend on who controls the work entry