Since the start of the third quarter of 2026, OpenAI has accelerated the growth of its revenue run-rate, driven largely by enterprise demand. According to internal company meeting materials reported by CNBC on August 19, the company's total revenue run-rate grew 35% quarter-to-date, while its enterprise revenue run-rate grew 50%. Enterprise API spending data collected daily by Ramp from companies connected via API also showed OpenAI growing 82% quarter-over-quarter, ahead of Anthropic's 76%.
However, this doesn't mean OpenAI has overtaken Anthropic in overall revenue scale. As of the end of July, Anthropic's annualized revenue run rate (ARR) stood at $6.5 billion, compared with OpenAI's "more than $4 billion" — a figure that represents only a floor. Growth rates, annualized revenue pace, quarterly revenue, and API spending growth rates each use different denominators and measure different things. The latest numbers reflect the pace of OpenAI's catch-up and shifts in enterprise usage, but they do not indicate a change in market leadership.
OpenAI's growth led in Q3-to-date figures
The internal OpenAI materials CNBC reviewed show changes in revenue run-rate since the start of the quarter — not finalized quarterly revenue, and not year-over-year comparisons. CNBC previously reported that OpenAI's Q2 2026 revenue was $6.7 billion, up 18% from Q1. That confirmed figure cannot be placed alongside the 35% quarter-to-date growth rate as if they were the same type of metric.
Alongside the reported 50% growth in enterprise run-rate, the materials noted that weekly users of AI coding and workplace productivity products reached 20 million. The internal materials don't specify how enterprise demand was distributed across products or contract types.
The timeline of model releases is clear: OpenAI made GPT-5.6 Sol generally available on July 9, pricing API access at $5 per million input tokens and $30 per million output tokens. Ramp's Ara Kharazian attributed OpenAI's growth to Sol's quality and developer preference. Still, public data doesn't allow us to separate how much of the 35% or 50% growth figures were driven specifically by Sol.
Existing per-model usage data shows that over the past month, Sol accounted for 25% of OpenAI's total tokens and 23% of spending, while Anthropic's Fable 5 accounted for 6% of its tokens and 11.4% of its spending. This is context about recent per-model usage and spending patterns — it does not represent company-wide revenue or Q3 growth contributions.
$6.5 billion vs. "$4 billion-plus" reflects a different ranking
Anthropic still leads in overall scale. CNBC confirmed Anthropic's end-of-July ARR at $6.5 billion, compared with OpenAI's figure of "more than $4 billion." Because OpenAI's number is only a floor, it's not possible to calculate a precise gap or multiple between the two. This comparison offers a general sense of scale rather than a revenue ranking calculated under identical accounting terms.
Even at the quarterly level, CNBC confirmed Anthropic's preliminary Q2 2026 revenue at $11.5 billion, compared with OpenAI's $6.7 billion for the same period. This gap is not inconsistent with the fact that API spending growth rates flipped in OpenAI's favor at one point in time — a company with a larger baseline can post a lower growth rate while its absolute revenue remains far ahead.
Neither company's ARR can be treated as confirmed annual revenue. ARR simply extrapolates a short-term revenue pace over 12 months; it is not an actual future result. Likewise, OpenAI's Q3 quarter-to-date run-rate figures come not from audited financial statements, but from internal materials reported by CNBC via anonymous sources.
API spending and paid adoption rates use different denominators
Ramp's 82%-versus-76% comparison isn't drawn from its regularly published AI Index spending data. Instead, it reflects quarter-to-date, quarter-over-quarter growth rates calculated from daily token-spending management data tied to companies that have connected their APIs to Ramp. This sample skews toward tech-oriented companies and doesn't represent the broader market, total company revenue, or consumer-facing sales.
| Figure | Scope / Denominator | What the Comparison Shows | What It Cannot Be Equated With |
|---|---|---|---|
| OpenAI's Q3 quarter-to-date growth: 35% overall, 50% enterprise | OpenAI's internal revenue run-rate since the start of the quarter | Growth reported internally across the company and enterprise segment | Finalized quarterly revenue, year-over-year comparisons, quarter-over-quarter comparisons |
| OpenAI 82%, Anthropic 76% | Token spending management data from companies connected to Ramp via API | Quarter-over-quarter growth in enterprise API spending | Overall company revenue growth rates for either company, market share |
| Anthropic $6.5B, OpenAI "$4B+" | Annualized revenue run-rate reported by each company | General sense of revenue pace scale; OpenAI's figure is a floor | Finalized annual revenue, exact gap or multiple, comparisons under identical accounting standards |
| Anthropic 43.5%, OpenAI 39.7% | Ramp-tracked companies with at least one paid AI product transaction in July | Share of companies with paid adoption | Spending share by dollar amount, user counts including free usage |
Ramp's standard AI Index tracks card and invoice transactions among U.S. companies. It counts adoption if there's at least one paid transaction for an AI product within a given month, but it doesn't capture free usage or usage through personal accounts. In July's paid adoption rate, Anthropic led with 43.5% compared with OpenAI's 39.7% — the opposite of the spending growth comparison. Spending can rise as existing customers increase usage, but the paid adoption rate only rises when new companies begin paying. It isn't unusual for these two metrics to point in different directions.
When choosing where to spend, enterprises weigh model performance and unit pricing alongside data retention terms. Contract terms, integration with existing workflows, and usage caps also shape decisions. While Ramp's data offers granular insight into changes among API-connected companies, it doesn't prove what specifically drove spending increases at either company.
Waiting for IPO filings to clarify accounting differences
What makes comparisons even harder is the differing scope of revenue recognition. The Wall Street Journal reported in April that Anthropic includes technology sales made through cloud partners in its revenue figures, while OpenAI does not. Profitability metrics also differ: the two companies report figures that either include or exclude training costs, with Anthropic using EBITDA and OpenAI using EBIT. The gap in ARR alone cannot be used to draw conclusions about profitability or differences in customer spending.
CNBC has reported that OpenAI confidentially filed IPO paperwork in June 2026, and that Anthropic has done the same. Once S-1 filings become public, disclosures on revenue recognition, gross margins, and customer concentration should provide a clearer basis for comparing figures that currently can't be placed side by side.
What can currently be confirmed is that OpenAI accelerated growth in both quarter-to-date enterprise demand and API spending among companies connected to Ramp, while Anthropic continues to lead in ARR and July's paid adoption rate. The next thing worth watching is how this competition in growth rates translates into actual revenue figures — and under what contract structures and revenue recognition standards that translation occurs.