The cheaper model is winning attention. Spending on Anthropic’s Fable 5 has plateaued at 11 percent of the overall outlay on AI tools among more than 70,000 American businesses, even as companies keep adding AI subscriptions.
Ramp’s data shows the percentage of its customers paying for AI climbed from 50% in March to nearly 56% by July. The market is expanding, but Fable 5 is not taking a larger share of that spending. That creates an awkward position for Anthropic: stronger revenue can coexist with weaker momentum for one model.
Anthropic Still Leads Ramp’s Paying Businesses
Anthropic reached 41% market share among Ramp’s paying business users in May 2026, edging past OpenAI at 39%. By July 2026, Anthropic held nearly 44%, while OpenAI held nearly 40%—a lead, but not an unbreakable one.
OpenAI is currently growing faster among Ramp’s business users in the third quarter to date than Anthropic. The figures therefore tell two stories at once: Anthropic has the larger share in Ramp’s latest data, while OpenAI is gaining ground faster among those business users.
That distinction matters because the number of companies paying for AI continues to rise. Ramp’s customers are paying for AI at a higher rate, and its data suggests both Anthropic and OpenAI should be growing business revenue as the overall market expands. Ramp declined to provide actual dollars spent, sharing only percentages, so the market-share figures show direction rather than a complete spending total.
Revenue Growth Meets Model Economics
Anthropic’s annualized revenue reached $65 billion in July. OpenAI’s annualized revenue surged to $40 billion, giving both companies large revenue figures even as their model-level performance differs inside Ramp’s data.
OpenAI’s GPT-5.6 model is much cheaper to use than Fable 5. That price difference gives businesses a reason to compare models by cost and application instead of treating every AI purchase as interchangeable. Corporate budgets remain stubbornly attached to arithmetic. An unfashionable habit, but a useful one.
Ara Kharazian, a Ramp economist, described the pressure on Fable 5 directly: “Fable 5, meanwhile, disappointed both in adoption and real-world application given price + data retention requirements imposed by regulators.” The statement ties the model’s weaker adoption to two specific burdens—its price and data retention requirements imposed by regulators—rather than treating market share as a simple popularity contest.
Kharazian also said, “GPT-5.6 Sol is really good, increasingly the choice for developers.” Miles Clements offered a broader explanation: “Most people don’t need to operate at the frontier.” Together, the comments describe a market that may reward sufficient capability at a lower cost, even when a more expensive model remains part of the competition.
Fable 5’s 11% spending share is the key warning sign. Anthropic can post $65 billion in annualized revenue and lead OpenAI among Ramp’s paying business users, yet still face a model that is not capturing more of the spending flowing into AI tools.
OpenAI’s position is not settled either. It trails Anthropic in Ramp’s July share, at nearly 40% versus nearly 44%, but its faster growth among Ramp’s business users in Q3 to date gives it a route to close the gap. With more than 70,000 American businesses represented in the spending data, those percentages offer a broad view of adoption, though Ramp’s refusal to provide dollar totals limits what they can prove.
The next phase of the competition will be measured through business payments and model adoption. Anthropic has revenue, a current lead, and a model facing adoption and cost pressure; OpenAI has lower pricing for GPT-5.6, a smaller July share, and faster growth among Ramp’s business users in Q3 to date.
As of August 25, 2026, the enterprise AI race has no clean winner. Anthropic leads the latest Ramp share figures, OpenAI is growing faster in the period measured, and the expanding market gives both companies room to grow—provided businesses keep paying for the models they choose.
Based on
