Traffic and revenue are measuring different things. That point seems obvious until you look at the actual numbers. SimilarWeb data cited by CNBC estimates ChatGPT holds approximately 57% of AI assistant web traffic. Gemini holds an estimated 25%. Claude sits at roughly 6%. By traffic, OpenAI is dominant, and has been for the better part of two years.
But CNBC is also reporting Anthropic’s annualized revenue run rate at $30 billion and OpenAI’s at approximately $24 billion. If those figures are accurate, and neither company publishes audited financials, so both are reported estimates, Anthropic is generating more revenue from 6% of the traffic than OpenAI is generating from 57%.
That ratio is the story.
Two models, two different customer profiles
The divergence is a function of customer composition, not model quality. Anthropic has concentrated its revenue in enterprise accounts. The company reports that more than 1,000 enterprises now spend at least $1 million annually on Claude, according to CNBC’s reporting. That figure is a vendor claim, Anthropic’s own disclosure via reporting, and should be read as such. But the shape it describes is meaningful: recurring seven-figure contracts from large institutional customers.
OpenAI’s revenue base is broader and includes a significant consumer layer, ChatGPT Plus and Pro subscriptions, alongside enterprise API and commercial contracts. Consumer subscriptions generate large aggregate revenue when the user base is large enough, but they come with higher churn, lower average contract value, and more sensitivity to competitive alternatives. A ChatGPT Plus subscriber can cancel for Claude Pro in an afternoon. An enterprise that has integrated Claude into its internal toolchain has a significantly higher switching cost.
That switching cost difference is the structural explanation for how 6% traffic share produces more revenue than 57% traffic share.
The enterprise customer base as a revenue moat
Enterprise AI contracts have characteristics that consumer subscriptions don’t. They tend to expand over time as the customer deploys the model across additional internal workflows. They come with annual or multi-year terms. They require API integration, which builds technical switching costs on top of the economic ones. And they generate usage data that can improve the model’s performance for that specific customer’s domain, another layer of retention.
The 1,000 enterprises at $1 million-plus annually aren’t just a revenue figure. They represent approximately $1 billion in highly predictable, highly defensible annual recurring revenue at the low end of their contract range, before volume discounts, before expansion, and before the upper-tier contracts that exceed the $1 million floor. The actual ARR from that cohort is likely significantly higher than the per-customer floor implies.
That’s the enterprise concentration advantage: revenue that behaves more like a managed services contract than a SaaS subscription.
The Opus 4.7 release as enterprise retention strategy
Anthropic launched Claude Opus 4.7 on April 16, 2026, according to the company. An independent primary source for that announcement wasn’t available for this report. The launch timing is notable in the context of the revenue story regardless of the specific model’s technical capabilities.
Enterprise buyers care about model release cadence for a specific reason. When a company has integrated an AI model into its internal workflows, it needs confidence that the model will keep improving, and that the vendor will remain at the frontier. A faster release cycle is an implicit commitment to that continuity. Anthropic releasing Opus 4.7 now, in the same week that the revenue comparison with OpenAI goes public, is a signal to those 1,000 enterprise customers: the roadmap is active, the frontier is moving, and your integration investment is supported by ongoing development.
Consumer users also benefit from model improvements, but they don’t renew contracts based on them. Enterprise buyers do.
The traffic-revenue decoupling and what it signals for OpenAI
The SimilarWeb data cited by CNBC shows ChatGPT’s traffic share has declined from an estimated 77% a year ago to approximately 57% now. Gemini has grown. Claude has grown. The traffic trend is moving against OpenAI even as it remains the dominant consumer platform.
OpenAI’s $852 billion private valuation, a known public figure from its 2025 funding activity, implies that investors are not primarily valuing it on current ARR. At $852 billion on $24 billion in reported ARR, the implied revenue multiple is roughly 35x. That premium reflects expectations of future growth, not current profitability. If enterprise AI revenue concentration is the direction the market is moving, and OpenAI’s consumer-heavy mix is generating lower revenue per unit of market attention than Anthropic’s enterprise mix, the growth trajectory that justifies that multiple requires OpenAI to capture enterprise share it doesn’t currently hold at the same concentration Anthropic has built.
The Musk-Altman trial on April 27 may put more operational detail on the public record. Any disclosed financial data in that context would help calibrate the $24 billion ARR figure, which currently rests on a single journalistic source without audited backup.
Implications for enterprise AI procurement
The revenue data has a direct implication for the companies choosing between Claude and ChatGPT for enterprise deployment. Anthropic’s enterprise concentration means the company has strong incentive to maintain enterprise-grade reliability, support, and model improvement cadence, these are the customers that drive its reported revenue lead. OpenAI’s broader customer base means enterprise accounts are a smaller share of its total attention and resource allocation.
That’s not a prediction that ChatGPT is worse for enterprise use. It’s an observation about strategic alignment: Anthropic’s business model is structurally aligned with enterprise customer success in a way that a consumer-first company’s model isn’t, by definition.
TJS synthesis
The 57% traffic / $24 billion vs. 6% traffic / $30 billion comparison is the clearest market signal yet that AI monetization has bifurcated. Consumer reach and enterprise revenue are tracking separately, and the enterprise layer is where the revenue concentration is building. Anthropic built its business around that layer deliberately. OpenAI built around consumer reach and is now trying to capture the enterprise layer as a second phase.
Both strategies can succeed. But they’re not equivalent paths. The enterprise-first model generates stickier revenue, lower churn, and higher per-customer lifetime value. The consumer-first model generates larger user counts, more public visibility, and more training data diversity. The question for the next phase of AI competition is which of those assets compounds faster into durable market position. The reported revenue gap suggests enterprise stickiness is currently winning. Whether that holds as OpenAI accelerates its enterprise push is the variable to watch through the rest of 2026.