OpenAI and Anthropic are presenting a version of profitability that excludes their largest expenses: model training and inference. Critics compare this to an airline ignoring the cost of its jets. This financial engineering aims to create a positive outlook for potential IPOs but masks their true cash burn rate.

Headlines pitting OpenAI against Anthropic on revenue are flawed. OpenAI is primarily a consumer subscription business with conservative revenue recognition, while Anthropic is an enterprise API business that recognizes "gross tonnage," creating fundamentally different financial pictures.

Unlike traditional SaaS, AI companies have significant variable costs for compute and tokens. This makes revenue a poor proxy for profitability, as their gross margins are fundamentally different from high-margin software businesses—a fact many investors miss.

Many AI startups multiply monthly consumption by 12 and label it Annual Recurring Revenue (ARR). True ARR is contracted and committed. This uncommitted "run rate" revenue is not durable and can disappear overnight if a competitor releases a better product, creating a misleading head fake for stakeholders.

Discrepancies in how private AI firms recognize top-line revenue versus partner commissions caused OpenAI's annualized revenue projections to swing by 20 billion dollars. This reporting mismatch triggered sharp sell-offs across major hardware and cloud infrastructure vendors, highlighting the public market's hyper-sensitivity to opaque private AI growth metrics.

The AI ecosystem has a systemic revenue recognition problem. A single compute token's value can be recognized as ARR multiple times as it's resold down the value chain (e.g., from OpenAI to an application wrapper). This creates inflated, non-durable revenue figures across the industry.

Anthropic's claims of positive operating income are based on adjusted figures that strip out major expenses like stock-based compensation and model training. This accounting practice masks the company's true financial health, which will be revealed for the first time in unadjusted IPO documents, posing a potential risk for investors.

Revenue figures for AI companies can be misleading. The same dollar is often counted multiple times as it moves from the end customer through a SaaS provider and a cloud platform before reaching the model provider, creating a "margin stacking" effect that obscures the true net revenue.

Anthropic’s cloud partnerships, like its one with Amazon, are structured as a 50% gross profit share, meaning costs like inference are deducted before sharing. This contrasts sharply with OpenAI's simpler 20% total revenue share with Microsoft, revealing different economic models for AI platform distribution.