YipitData, an alternative-data firm that tracks spending through panels of businesses and consumers, published an “Anthropic Pre-IPO Overview” on 22 October 2026, last updated 23 October. In its B2B datasets, it writes, “the top 1% of customers make up nearly 50% of Anthropic spend.”
The same page says Anthropic’s B2B retention in its spend panel “is strong and actually improves over time: 87% after month 12 and 95% after month 24,” and that across its token-based datasets it has seen “a ~50% Y/Y decline in token prices.” These are YipitData’s estimates from its panels, not company figures.
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.
The key insight: As we read it, YipitData’s panels describe Anthropic B2B revenue that is concentrated and sticky: nearly 50% of spend from the top 1% of customers, and retention that rises from 87% after month 12 to 95% after month 24.
What YipitData’s Overview Shows
- Adoption. YipitData writes: “Anthropic has seen significant adoption growth YTD, but through September this is beginning to flatten out, meaning more growth must come from existing customers.” It adds that Tech and Financials customers drive the bulk of spend in its panels.
- Concentration. “In our B2B datasets, the top 1% of customers make up nearly 50% of Anthropic spend.”
- New models. Across its token and pricing datasets, including OpenRouter, it writes that “we observed an uptick in spend driven by Opus 5.5 (9/22 launch).”
- Prices. “Across our token-based datasets, we’ve observed a ~50% Y/Y decline in token prices.” It lists price cuts from Opus 4.5, Sonnet 5’s introductory pricing that became permanent, higher cache hit rates and Fable 5.1 cache-pricing reductions as contributors.
- Competition. Within OpenRouter, it writes, “China-based providers like DeepSeek and Z.ai have taken significant share of usage.” Against OpenAI in its B2B panel, it writes that “Anthropic managed to take significant share up until June.”
- Consumers. In its e-receipt panel, YipitData says it has tracked “a rapid rise in Claude subscribers so far in 2026, and the retention of these users is stronger than past cohorts.”
- Cloud. Per its enterprise cloud panel, “the majority of Anthropic spend goes through AWS, though GCP has been growing more quickly YTD.”


How the Panels Are Built
YipitData says the analysis “combines data from 1,500 SMB businesses, 200 large enterprise companies, and millions of consumers, accounting for more than $4B in annual AI spend.” Each section of the page ends with questions it suggests investors put to Anthropic. The page states that YipitData “is not registered as an investment advisor in any jurisdiction.”
The Numbers Next to Our Earlier Coverage
The table sets YipitData’s estimates beside figures from other sources we have covered. The rows measure different things over different groups, so they are listed side by side and not combined.
| Measure | Figure | Scope | Source |
|---|---|---|---|
| Share of spend from the top 1% of customers | Nearly 50% | Anthropic spend in YipitData’s B2B datasets | YipitData estimate, 22 Oct |
| Share of spend driven by 1% of firms | 80% | Spend on OpenAI and Anthropic combined, businesses on Ramp | Ramp letter, 14 Oct (our Ramp piece) |
| Token prices | About 50% lower year on year | YipitData’s token-based datasets | YipitData estimate, 22 Oct |
| Blended price per million tokens | Down 25.4% | OpenAI and Anthropic on Ramp, week ending 11 Oct vs week ending 13 Sep | Ramp AI Index |
| B2B retention | 87% after month 12; 95% after month 24 | YipitData’s B2B spend panel | YipitData estimate, 22 Oct |
| Annualized revenue run rate | $65 billion | Anthropic, September 2026, as listed in Damodaran’s table | Aswath Damodaran, 22 Oct (our Damodaran piece) |
Sources: YipitData, “YipitData Anthropic Pre-IPO Overview”, published 22 October 2026, updated 23 October; Ramp Econ Lab letter of 14 October 2026 and Ramp AI Index, as covered in our earlier piece; Aswath Damodaran’s post of 22 October 2026, as covered in our earlier piece. The YipitData page gives no revenue estimate.
Connecting the Coverage
Our Ramp piece covered Ramp’s finding that 1% of firms drive 80% of spend on OpenAI and Anthropic. Our Damodaran piece covered the $65 billion run rate he lists and the multiples a $2 trillion valuation would imply. Our valuation check covered the gap between IPO expectations and private-round targets. What’s new here: an outside data firm’s panels put concentration in Anthropic’s own B2B spend at nearly 50% for the top 1% of customers, next to retention and price estimates.
The Structural Read
As we read it, concentration is the first number an IPO investor would check: on YipitData’s estimate, a small group of customers carries about half of Anthropic’s B2B spend in its datasets.
As we read it, retention is the counterweight: YipitData says B2B retention in its spend panel improves over time, to 95% after month 24.
As we read it, price is the third input: YipitData estimates token prices are down about 50% year on year, while it also reports adoption growth flattening through September.
YipitData, Anthropic Pre-IPO Overview, 22 October 2026
“Anthropic has seen significant adoption growth YTD, but through September this is beginning to flatten out, meaning more growth must come from existing customers.”
Three Implications
FOR IPO INVESTORS As we read it, YipitData’s estimates come from outside panels, so they would sit beside the figures Anthropic would report in an IPO filing.
FOR READERS OF AI DEMAND FIGURES YipitData puts concentration at nearly 50% of Anthropic spend from the top 1% of customers; Ramp’s data puts 80% of OpenAI and Anthropic spend on Ramp with 1% of firms. The two measure different groups.
FOR ANTHROPIC’S RIVALS YipitData reports that China-based providers like DeepSeek and Z.ai have taken significant share of usage within OpenRouter.
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, YipitData’s estimates sit at the revenue end of that chain: concentrated, retained B2B spend is what an IPO would price. The framework was published in May 2026 and does not discuss YipitData’s data.
What Is Not Established
The figures are YipitData’s estimates from its panels, not numbers reported by Anthropic. The public page shows charts that we did not read as numbers, and it gives no revenue estimate, no subscriber count and no share figure against OpenAI. We used only what the public page shows and did not book a demo or use any gated material. In what we read, Anthropic has not commented on the overview. We did not contact YipitData or Anthropic.
The Bottom Line
YipitData’s pre-IPO overview estimates that the top 1% of Anthropic’s B2B customers make up nearly 50% of its spend in its datasets, that B2B retention reaches 87% after month 12 and 95% after month 24, and that token prices are down about 50% year on year. It also says adoption growth began to flatten through September.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. This piece rests on YipitData’s public “Anthropic Pre-IPO Overview” page of 22 October 2026, updated 23 October, read in full, and on our own earlier pieces on Ramp’s data and Aswath Damodaran’s post. YipitData’s figures are estimates from its panels. We did not use gated material and did not contact YipitData or Anthropic.
Sources: YipitData, Anthropic Pre-IPO Overview, 22 Oct 2026 (updated 23 Oct) · Ramp letter and AI Index, as covered in our piece · Aswath Damodaran, 22 Oct 2026, as covered in our piece









