1% of Ramp Firms Drive 80% of OpenAI, Anthropic Spend

Derek Thompson’s newsletter and podcast of 23 October 2026 put one question at the centre of the AI trade: how much revenue the AI ecosystem earns from outside itself. His main source is Ramp, the corporate-card company, and its lead economist Ara Kharazian, whose monthly letter of 14 October says “80% of spend on OpenAI and Anthropic is driven by 1% of firms” on Ramp.

The same Ramp letter says per-employee spend among those top firms is stalling and records “two straight months of virtually no growth in token spend.” Apollo’s chief economist Torsten Slok added on 22 October that most of the biggest spenders are VC-backed.

Business Pill · CONCENTRATION

A short Business Engineer explainer of concentration: when a few big pieces make most of a total. It teaches the general idea only and says nothing about the figures in this piece.

What Ramp’s Data Shows

• Concentration. Kharazian wrote in Ramp’s 14 October letter that Ramp had previously shared that “80% of spend on OpenAI and Anthropic is driven by 1% of firms.”
• The top 1% is stalling. The letter says that “on a per-employee basis, spend is stuttering for the top 1% of firms.” Per-employee spend rose 11.7% from August, but the top 1% spent $9,568 per employee per month in September, down 3.1% from its July high of $9,873.
• Token spend. Ramp writes that token volumes rose over the same period and that token spend led the declines: “Because of August and September, we’ve recorded two straight months of virtually no growth in token spend.” It adds that AI spend rose at the beginning of October to a new peak, 1.5% above the previous peak in August.
• Prices. Kharazian’s explanation: “the main driver is price declines at the frontier, in addition to efficiency and quality improvements for models across intelligence levels, that are driving the blended price per million tokens down.”
• Who the big spenders are. In Apollo’s Daily Spark of 22 October, Slok wrote: “Ramp data show that 80% of the top 1% of AI spenders are VC-backed, compared with only 13% of the median spender.” Apollo’s chart puts the shares at 80.4% and 13.1%.

What Was Posted on X

Derek Thompson posted on X about his new podcast and newsletter, setting out three cracks in the AI thesis from Ramp economist Ara Kharazian: concentration among VC-backed tech companies, stagnating spend among the top 1% of companies, and falling token prices. The framing is Thompson’s own.

The Numbers, Side by Side

The table lists Ramp’s figures with the period each one covers. Two rows both read about 80%, but they measure different things: one is a share of spending, the other a share of firms.

Sources: Ramp Econ Lab letter “Cracks in the AI Thesis Part 3”, 14 October 2026; Ramp AI Index, data through September 2026; Apollo Daily Spark, 22 October 2026, as covered in our earlier piece. Spend per employee is the median within each group, per Ramp. The latest month in the index is September 2026.

What Thompson Wrote

Thompson opens with the analyst Craig Moffett, who told Ben Thompson: “All I care about is, are the revenues coming into the system greater than the costs of building the whole value chain.” Thompson writes that Ramp oversees the spending of about 85,000 American businesses, including more than $10 billion a year in AI purchases, and that its sample “skews toward tech companies, and cannot account for the whole economy.”

On X, Thompson set out three “cracks” from Kharazian’s data: concentration, stalling spend at the top, and falling token prices. On concentration, he wrote that if most AI spending comes from companies that rely on venture capital firms that also rely on AI valuations, “this seems like another example of the AI capital cycle paying itself.” That is Thompson’s reading.

Connecting the Coverage

Our Chamath piece covered Chamath Palihapitiya’s claim that VC money funds most AI spending, and its update added Apollo’s 80.4% figure. Our token spend piece covered Kharazian telling Axios that token spend had been flat for two months. Our valuation check and our Damodaran piece covered the valuations of Anthropic and OpenAI. Our chip debt piece covered the debt funding the build-out. What’s new here: Ramp’s own letter puts the concentration at 80% of OpenAI and Anthropic spend from 1% of firms, next to its per-employee and price data.

The Structural Read

As we read it, concentration is the first issue Ramp raises: on its data, a small group of firms accounts for most spend on the two leading labs, so a change in that group moves the total.

As we read it, the second is who funds that group: Apollo’s chart, based on Ramp data, shows 80.4% of the top 1% of AI spenders are VC-backed, against 13.1% of the median spender.

As we read it, the third is price: Ramp’s blended price per million tokens fell 25.4% in four weeks while token volume rose 52.4%, so more usage and lower spend growth can arrive together.

Three Implications

FOR AI MODEL VENDORS On Ramp’s data, 80% of OpenAI and Anthropic spend comes from 1% of firms, so a handful of buyers carry much of the business revenue Ramp sees.

FOR READERS OF AI DEMAND FIGURES Ramp reports token volume up 52.4% and token spending up 23.0% over the same four weeks, with the blended price down 25.4%.

FOR INVESTORS IN THE BUILD-OUT Thompson’s question, via Moffett, is whether revenue from outside the AI value chain covers the cost of building it; Ramp’s data covers businesses on Ramp only.

The Business Engineer Lens

This story maps onto the Business Engineer framework The AI Capex Map & The State of AI Hyperscalers.

The framework describes its subject as “How the money flows, where it cracks, what it buys, who can absorb it”.

As we read it, Ramp’s figures sit at the demand end of that chain: business spend on models is the revenue the build-out is meant to earn. The framework was published in May 2026 and does not discuss Ramp’s data.

What Is Not Established

Ramp’s figures cover businesses on Ramp, not the whole economy. Ramp notes that “because the top 1% represents a small group of firms, its spending estimate is more volatile than our median and top 10% estimates,” and that it revised the top-1% figures up after adding new AI vendors. The 80% concentration figure is described in the letter as previously shared data; the letter we read does not restate its period.

Thompson’s post on X describes the top 1% as accounting for 80 percent of all AI spending, while Ramp’s letter puts it at 80% of spend on OpenAI and Anthropic. We did not contact Ramp, Apollo or Thompson.

The Bottom Line

On Ramp’s data, 1% of firms drive 80% of spend on OpenAI and Anthropic, and Apollo’s chart shows 80.4% of the top 1% of AI spenders are VC-backed. Ramp also reports stalling per-employee spend at the top, two flat months of token spend before an October rise, and a blended token price down 25.4% in four weeks.