IMPLICATOR
Anthropic investors expect an October 2026 IPO at a valuation of at least $2 trillion, even as early spending data points to caution among companies buying its premium Fable 5 model. In Ramp’s July 2026 token-management sample, Fable generated roughly 75% as much model-attributed spending as OpenAI’s GPT-5.6 Sol. The gap matters because the IPO forecasts depend on booming demand for Anthropic’s most advanced models.
What Changed
- Six Anthropic investors expect an October IPO valued at $2 trillion or more, though Anthropic has not set that target.
- Ramp says Fable 5 generated about 75% as much July model-attributed spending as GPT-5.6 Sol.
- Anthropic still led Ramp’s July paid-adoption sample, and overall AI spending continued to rise.
- Ramp’s Fable subset skews toward technology companies and does not explain why buyers selected each model.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Price meets usage
Ramp published the first month of Fable data on August 12. In July 2026, Fable accounted for 6% of Anthropic tokens and 11.4% of spending attributed to Anthropic models in Ramp’s product. Sol represented 25% of OpenAI tokens and 23% of spending attributed to OpenAI models during the same month.
During July 2026, Fable’s input rate was roughly $10 per million tokens, twice Sol’s input price. A separate July 2026 CursorBench 3.2 comparison found that Anthropic’s Opus 5 scored 70.0% at maximum effort and cost $8.23 per task, compared with 70.5% and $17.32 per task for Fable 5 at maximum effort. One software-engineering benchmark cannot establish model quality across other kinds of work, but it shows how a cheaper model can approach Fable on a measured task set.
The investor case
Six Anthropic backers expect the company to list in October 2026 at $2 trillion or more. Their models also put Anthropic’s annualized revenue at $100 billion to $120 billion by the end of 2026, more than ten times its level at the start of the year.
Those are investor expectations, not company guidance. Senior Anthropic executives had not fixed an IPO valuation target in their conversations with the backers. The annualized figure extrapolates recent sales into a full year and is not the same as revenue already booked. Ramp’s model-level spending also cannot be read as Anthropic revenue. Vendor adoption measures the share of Ramp customers paying a provider, while model-attributed spending assigns purchases inside Ramp’s token-cost product to individual models.
Anthropic still leads adoption
The same Ramp data shows Anthropic still led paid adoption. In July 2026, 43.5% of businesses in its broader sample paid for Anthropic subscriptions or tokens, up 1.1 percentage points from June. OpenAI reached 39.7%, up 0.23 percentage points over the same period.
Get Implicator.ai in your inbox
Strategic AI news from San Francisco. No hype, no "AI will change everything" throat clearing. Just what moved, who won, and why it matters. Daily at 6am PST.
No spam. Unsubscribe anytime.
AI bills were rising too. Ramp’s median firm spent $11.95 per employee in July 2026, while the top 1% of firms spent a median $7,400 per employee that month. The resistance is to one premium model, not to AI spending as a category.
Ramp’s evidence does not cover the whole business market. Its Fable figures come from a token-spend-management product whose customers skew more heavily toward technology than its broader index, and the company data does not reveal why a buyer selected one model over another.
Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.
Buyers weigh performance
Price is not always the first filter. A self-selected July 2026 survey of 170 enterprises found that performance and GPU availability ranked ahead of total cost of ownership in provider selection. Only 47% of that sample rigorously tracked compute cost and return, limiting how precisely many buyers could compare value.
A separate, nonrandom January 2026 survey of 100 Global 2000 executives found that respondents expected average LLM spending to rise from roughly $7 million to $11.6 million during 2026. The survey’s publisher, a16z, disclosed that it invests in OpenAI.
Amazon technology chief Werner Vogels framed the decision more narrowly in July 2026: “Do you really need to have the biggest, highest‑end model to solve this? The answer is no, you don’t.”
Frequently Asked Questions
What did Ramp measure?
Ramp measured two different things. Its broader index tracks the share of customers paying an AI provider. Its token-management product attributes purchases to individual models. The Fable 5 spending figures come from the second, more technology-heavy sample.
How did Fable 5 spending compare with GPT-5.6 Sol?
In Ramp’s July 2026 token-management sample, Fable 5 generated roughly 75% as much model-attributed spending as GPT-5.6 Sol. The two figures sit within different vendor totals and do not measure the models’ full market revenue.
Does the Ramp data show that a $2 trillion Anthropic IPO is overpriced?
No. The IPO figure is an expectation modeled by six investors, not a valuation target set by Anthropic. Ramp’s product-level spending sample is a separate signal and cannot be read as Anthropic revenue or a direct valuation test.
Does Anthropic still lead paid business adoption?
Yes, within Ramp’s broader July 2026 sample. Anthropic reached 43.5% paid adoption, compared with 39.7% for OpenAI. Those figures count subscriptions and token payments to each provider, not usage of a particular model.
Why might companies choose a cheaper AI model?
A cheaper model can be sufficient for a defined task. In one software-engineering benchmark, Anthropic’s lower-cost Opus 5 nearly matched Fable 5 at maximum effort while costing less per task. The result does not establish quality across other workloads.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai
