Spend on Anthropic models by businesses on Ramp “remains 1.6% below its July peak,” Ramp lead economist Ara Kharazian wrote on X on 27 October 2026, even though Opus 5.5 “has somewhat reaccelerated demand for Anthropic.” He wrote that Opus 5.5 “now accounts for 63% of Anthropic spend.”
These are Ramp figures for token spend by businesses on its platform, not Anthropic’s revenue, which counts all of its customers.
Short explainer · Anthropic on Ramp
A vertical news Short narrated from the published text of this piece and related FourWeekMBA pieces.
Business Pill · The Jevons paradox explained
An evergreen Business Engineer explainer on the Jevons paradox, why cheaper AI can mean a bigger bill. It explains the general idea and does not discuss Ramp or Anthropic.
The key insight: As we read it, Ramp’s data show Anthropic gaining share of token spend on its platform while its total spend there sits 1.6% below July, with Opus 5.5 now 63% of it.
Kharazian’s post opened with “NEW Ramp data” and added: “Opus 5.5 now accounts for 63% of Anthropic spend, so frontier models do matter.”
He also gave his view of the risks. He wrote that the idea AI is already good enough is “probably overrated as a concern IMO,” and that “price cuts / competition at all intelligence levels are still the most significant threats to AI model companies, and continue to form cracks in the AI thesis.”
Share of US businesses on Ramp with a paid transaction to Anthropic and to OpenAI in each month, January 2025 to September 2026, from the Ramp AI Index adoption series (ramp.com/data). This is adoption, not spend, and not either company’s revenue.Anthropic on Ramp: spend 1.6% below its July peak and Opus 5.5 at 63% of Anthropic spend, per Ramp lead economist Ara Kharazian on X (27 Oct 2026); Anthropic at 60.5% of token spend in the week ending 23 October against 51.3% a month earlier, per the Ramp AI Index.
What Was Posted on X
NEW Ramp data: Opus 5.5 has somewhat reaccelerated demand for Anthropic, but it might not be enough.
Spend on Anthropic models remains 1.6% below its July peak, according to Ramp data.
Opus 5.5 now accounts for 63% of Anthropic spend, so frontier models do matter. The idea that AI is already “good enough” is probably overrated as a concern IMO.
But price cuts / competition at all intelligence levels are still the most significant threats to AI model companies, and continue to form cracks in the AI thesis.
Ara Kharazian, Ramp’s lead economist, posted the data on 27 October: spend on Anthropic models remains 1.6% below its July peak according to Ramp data, and Opus 5.5 now accounts for 63% of Anthropic spend. In the same post he wrote that price cuts and competition at all intelligence levels are still the most significant threats to AI model companies.
Eric Glyman reveals Rampโs internal data on Anthropic shows signs of a better Q4 than Q3, after AI spend elasticity slid from the Jevons paradox toward one to one
โThis summer, that appears to have broken. And we can debate all the reasons why. I think that there is incredible levels of pricecompetition that is going on between the major labs.”
“Open source has been effective, whether itโs in distilling or novel innovations taking place out there and putting pressures on, combined with this overarching question of have we reached saturation in intelligence among certain tasks?”
“Maybe the way to put it is it might just be that once you get to 130 IQ or some equivalent, that might be all you need to do accounting. That might be all you need to answer customer tickets, and further gains of intelligence may not be creating or inducing more demand.”
“And so when you look functionally whatโs happened, thereโs been this period where spend has been relatively flat, and it seems when you look at elasticity, that relationship suggests it is switched from this paradox to closer to one to one.”
“There are some signs, even data a week ago, when you look at some of the Anthropic data in the Ramp data set, suggesting perhaps 5.5 has hit an inflection point and there is going to be a better Q4 for them than Q3.โ
Fireside Alpha posted a clip of Eric Glyman on 27 October. In it, Glyman says some of the Anthropic data in the Ramp data set suggest perhaps 5.5 has hit an inflection point and there is going to be a better Q4 for Anthropic than Q3.
What Ramp’s Chart Shows
The chart attached to the post is titled “Anthropic spend remains below peak, despite Opus 5.5 growth.” It shows “Anthropic token spend across Ramp businesses, indexed” as weekly bars from January 2025 to late October 2026, split into Opus 5.5, Sonnet 5.5, Haiku 5.5 and older models.
The tallest bar sits just after the July 2026 label. Opus 5.5 segments appear only in the last five bars, in October, and in the final bar the Opus 5.5 segment is the largest of the four. Ramp did not publish the index values behind the bars.
Share Up, Price Down
Opus 5.5 is 63% of Anthropic spend by businesses on Ramp, per Ara Kharazian on X, 27 Oct 2026; the remaining 37% for all other Anthropic models is our arithmetic. Ramp platform data, not Anthropic revenue.
Ramp’s AI Index page gives the weekly picture. In the week ending 23 October, Anthropic had 60.5% of token spending among businesses on Ramp, against 51.3% in the week ending 25 September. OpenAI’s share fell from 44.2% to 35.1% over the same weeks.
Ramp says total token spending rose 8.9% over that comparison while total token volume rose 37.1%. Its blended price for Anthropic fell 17.9%, from $0.89 to $0.73 per million tokens, and OpenAI’s fell 34.7%, from $0.67 to $0.44.
Ramp notes that the token figures come from businesses using its Token Spend Management product, a sample that “is not representative of Ramp’s total AI spend volume, though results are directionally similar.”
More Firms on Ramp Pay Anthropic Than OpenAI
Ramp’s monthly adoption series measures something else: the share of US businesses on Ramp with a paid transaction to an AI company in a month. Anthropic’s share rose from 10.63% in January 2025 to 43.77% in September 2026, Ramp’s published series shows.
OpenAI’s rose from 33.79% to 40.63% over the same months. Anthropic first passed OpenAI in May 2026, at 40.70% against 39.32%. Adoption counts firms that pay, not how much they pay.
Ramp’s CEO on the Fourth Quarter
Earlier on 27 October, Ramp chief executive Eric Glyman spoke about the same data in a video clip posted on X by Fireside Alpha. Fireside Alpha summarised the clip as AI spend elasticity sliding from the Jevons paradox toward one to one. On that relationship between price and spend, Glyman said: “This summer, that appears to have broken.” He added: “I think that there is incredible levels of pricecompetition that is going on between the major labs.”
He also said: “There are some signs, even data a week ago, when you look at some of the Anthropic data in the Ramp data set, suggesting perhaps 5.5 has hit an inflection point and there is going to be a better Q4 for them than Q3.”
What’s new: Ramp’s own data show Anthropic spend on its platform still below July even as Opus 5.5 takes 63% of it. Ramp measures spending by businesses on its platform; the S-1 counts all of Anthropic’s revenue. The two measure different things and we do not combine them.
The Structural Read
As we read it, Ramp’s token week shows volume up 37.1% and spend up 8.9%, with Anthropic’s blended price down 17.9% and OpenAI’s down 34.7%.
As we read it, Kharazian names the risk in his own post: he wrote that price cuts and competition “at all intelligence levels” are the most significant threats to AI model companies.
As we read it, Glyman describes the same tension from the other side, saying there are signs of “a better Q4 for them than Q3.”
Ara Kharazian, Ramp, on X, 27 October 2026
“Opus 5.5 now accounts for 63% of Anthropic spend, so frontier models do matter.”
Three Implications
FOR AI MODEL BUYERS Ramp’s index puts Anthropic’s blended token price at $0.73 per million in the week to 23 October, against $0.89 a month earlier.
FOR AI LABS Kharazian wrote that Opus 5.5 is 63% of Anthropic spend on Ramp, and that price cuts and competition remain the biggest threats.
FOR READERS OF RAMP DATA Ramp says its token figures come from a sample of its customers, and none of them are Anthropic’s revenue.
The framework puts it this way: “The reason cheaper tokens produce bigger bills is structural.”
As we read it, Ramp’s latest week is a test of that idea: token volume rose 37.1% and spend rose 8.9% as prices fell. Glyman said the relationship “appears to have broken” this summer. The framework was written before this data and does not discuss Ramp or Opus 5.5.
What Is Not Established
Ramp did not publish the index values behind Kharazian’s chart, the exact peak week, or the period the 63% share covers. The token figures come from a sample of Ramp customers. We saw Glyman’s remarks only as the clip and its relays on X, not the full conversation.
This Week In The AI Supercycle
Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls
This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.
Anthropic spend by businesses on Ramp is 1.6% below its July peak, Ramp’s lead economist wrote on 27 October, with Opus 5.5 now 63% of it. Ramp’s index shows Anthropic at 60.5% of token spending in the week to 23 October, while its blended price fell 17.9%. These are Ramp figures for spending on its platform, not Anthropic’s revenue.
A note on sourcing. Ramp figures are from Ara Kharazian’s post on X on 27 October 2026, the chart attached to it and the Ramp AI Index page (ramp.com/data). They describe spending by businesses on Ramp’s platform, not Anthropic’s or OpenAI’s revenue; token figures come from a sample of Ramp customers. Eric Glyman’s remarks are from a video clip posted on X by Fireside Alpha. This is not investment advice.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
