OpenAI and Anthropic use different methods to calculate revenue, creating investor confusion

**Two Numbers, Two Definitions: How OpenAI and Anthropic Measure Revenue Differently**

As of late 2026, OpenAI and Anthropic have emerged as the dominant private rivals in artificial intelligence, each raising tens of billions of dollars and attracting intense scrutiny from institutional investors ahead of potential public offerings. Both companies now regularly cite annualized revenue figures — a metric derived by taking a recent shorter-term snapshot and projecting it forward across a full year — as the primary benchmark of their commercial progress. OpenAI expects to reach or exceed $70 billion in annualized revenue by year-end, while Anthropic crossed $65 billion in annualized revenue as of July 2026. On the surface, these figures suggest a tight race between two comparably scaled businesses. In practice, investors and traders have discovered that the two numbers measure meaningfully different things, making direct comparison misleading.

Annualized revenue is already an unconventional way to represent a company's financial performance. Unlike trailing twelve-month revenue, which reflects what a business actually collected over a completed period, annualized figures extrapolate from recent momentum — a method that flatters fast-growing companies but can obscure volatility, seasonality, or one-time spikes. For AI companies in particular, where large enterprise contracts, cloud partnership arrangements, and API consumption-based billing all feed into the top line in different ways, the underlying composition of that annualized figure matters enormously. The problem that has surfaced among investors is that Anthropic and OpenAI appear to be using different methodologies to construct their headline numbers, without clearly disclosing those differences.

**Why the Opacity Is Spooking Sophisticated Investors**

The confusion is especially consequential given the audience relying on these figures. Both companies have drawn investment from major institutional names — T. Rowe Price, for instance, holds stakes in both OpenAI and Anthropic and has publicly suggested each has a credible path to market dominance. Sovereign wealth funds, venture firms, and crossover hedge funds have poured capital into both entities at valuations that implicitly depend on revenue trajectories. When the key metric used to justify those valuations turns out to be calculated differently from one company to the next, it introduces a layer of analytical uncertainty that traders describe as a black box problem: the inputs and assumptions behind the number are not transparent enough to model with confidence.

The practical stakes are significant. If Anthropic's $65 billion annualized figure includes committed but not yet recognized contract value, or counts cloud compute credits extended by a strategic partner like Amazon Web Services differently from how OpenAI counts similar arrangements with Microsoft Azure, the gap between the two companies could be wider or narrower than the headline figures suggest. Enterprise AI contracts often involve prepayments, consumption-based drawdowns, and minimum spend commitments that can be recorded in multiple ways. Neither company is currently subject to the standardized revenue recognition rules that public companies must follow under U.S. generally accepted accounting principles, which means they have substantial discretion in how they frame these numbers for investors and the press.

This matters particularly as both companies have signaled movement toward public markets. Anthropic's IPO process has already prompted investor discussions about how to price existential and regulatory risk alongside conventional financial metrics. The revenue methodology question adds a further complication: without a clear apples-to-apples comparison, analysts cannot confidently determine whether the two companies are converging or diverging in commercial scale, which in turn makes it harder to establish rational valuations for either.

**The Broader Context: AI Revenue Metrics as Competitive Signaling**

Beyond the technical accounting question, there is a competitive and reputational dimension to how AI companies report these numbers. Annualized revenue figures function partly as marketing — they signal momentum to prospective enterprise customers, potential employees, and future investors. In an industry where perception of leadership can influence actual market outcomes through self-reinforcing customer adoption, the incentive to present the strongest possible version of one's financial performance is powerful. This creates a structural pressure toward figures that may be optimistically framed even when technically defensible.

OpenAI's trajectory has been dramatic by any measure. The company moved from roughly $1 billion in annualized revenue in early 2023 to projections north of $70 billion by late 2026 — a pace of growth that is nearly without precedent in enterprise software history. Anthropic's climb has been similarly steep, accelerating sharply after the commercial launch of its Claude model family and its deep integration into Amazon's cloud infrastructure. Both companies have expanded from API access for developers into large-scale enterprise agreements with Fortune 500 firms, government agencies, and financial institutions.

Yet rapid growth also creates the conditions under which revenue metric choices become most consequential. A company growing at triple-digit annual rates will look radically different depending on whether annualized revenue is calculated from the most recent month, the most recent quarter, or a rolling average — and whether certain partnership arrangements, reseller agreements, or compute credits are included or excluded. Until both companies adopt standardized reporting or until a public offering forces adherence to SEC disclosure standards, investors are effectively working with figures that are self-reported, methodologically opaque, and not directly comparable to one another.

The episode illustrates a broader maturation challenge for the AI industry: as these companies cross revenue thresholds that would make them among the largest software businesses ever built, the informal and promotional language of startup fundraising is colliding with the analytical demands of serious institutional capital markets. The traders and analysts trying to price these businesses are finding that the numbers they have been given were never designed for the scrutiny now being applied to them.