Anthropic’s Fable 5 Holds Only 11% of Enterprise Spend After Two Months
Spend data from 70,000+ companies shows Fable 5 accounts for just 11% of Anthropic enterprise budgets two months after launch. Here's what that means for the IPO.
More than two months after Anthropic launched Fable 5, its most powerful and most expensive model, corporate customers still haven't embraced it. Spend data gathered by payments platform Ramp from over 70,000 companies shows Fable 5 accounting for roughly 11% of total enterprise spending on Anthropic products, with nearly 90% of budgets going to older, cheaper alternatives. The Financial Times, citing this data, reported that the pattern challenges a long-held assumption: that enterprises automatically upgrade to the strongest available model.
What happened
| Data point | Figure |
|---|---|
| Fable 5 share of Anthropic enterprise spend (by dollar) | ~11.4% |
| Fable 5 share of Anthropic enterprise spend (by token volume) | ~6% |
| Companies in Ramp’s sample | 70,000+ |
| Anthropic revenue growth so far this year | Nearly 7x |
| Anthropic annualised monthly revenue (latest) | $65B |
| Investor revenue expectation ahead of IPO | $80B |
| Anthropic large enterprise customers (spending $100K+ per year) | 6,000 |
| OpenAI annualised revenue after GPT 5.6 launch | $40B+ |
| OpenAI annualised revenue growth this quarter | 35% |
| Anthropic expected IPO valuation | $2 trillion+ |
Financial Times reporter George Hammond reported the Ramp findings, noting that Fable 5 adoption has stalled at around 11% of Anthropic spend by dollar value since the model launched. When measured by token volume, the actual amount of text the model processes for paying customers, that share falls to just 6%.
Adding to the picture: Claude Opus 5, a smaller and cheaper Anthropic model that launched in late July, has already surpassed Fable 5 in business spending. Enterprises are actively choosing less expensive options within the same vendor’s lineup.
Why enterprises are skipping the flagship model
Miles Clements, a partner at Accel, which has invested close to $1 billion in Anthropic, gave the FT a candid read: “Most people simply don’t need to operate at the performance frontier.” He described the phase where customers only want the most advanced model as something that “was never a sustainable norm.”
The argument is not that frontier capability is worthless. Clements acknowledged that breakthroughs in raw intelligence matter for Anthropic’s larger ambitions, such as the often-cited goal of accelerating drug discovery, and for attracting top research talent. But for the majority of business use cases, the most advanced model is becoming more of a showroom item than a daily workhorse.
Low-cost, open-weight models from China and other regions (models where the internal weights are published so anyone can deploy and modify them independently) are also giving enterprises more alternatives and reducing any pressure to pay for the most expensive proprietary option.
Is this actually hurting Anthropic?
Not visibly, yet. Anthropic’s revenue has grown nearly 7x since the start of the year. The company posted its first quarter of positive adjusted operating income in Q2, meaning its core business turned profitable after stripping out certain one-time costs. According to sources cited by the FT, Anthropic has told investors it expects to remain profitable in Q3, and it currently has 6,000 enterprise clients each spending more than $100,000 per year.
The tension is at the top of the forecast range. Latest annualised monthly revenue sits at $65 billion, below the $80 billion some investors had projected. Growth also slowed in June when the Trump administration placed restrictions on Fable’s promotion, though momentum has since recovered.
The contrast with OpenAI is instructive. OpenAI launched GPT 5.6 in July at a noticeably lower price point, which lifted its annualised revenue 35% to over $40 billion this quarter, reversing months of trailing Anthropic.
What does this mean for the Anthropic IPO?
Anthropic is preparing for an IPO that outside observers expect could arrive as soon as next month. Investor estimates put the potential valuation above $2 trillion, which would make it the largest initial public offering in history.
The Fable 5 spend data complicates the story that premium frontier models drive premium revenue. If most enterprise customers gravitate toward capable but cheaper models, the economics of spending billions on training ever-larger systems become harder to justify at a $2 trillion price tag.
This is not unique to Anthropic. The broader question facing every frontier AI lab is whether “more powerful” reliably converts into “more revenue.” Right now, the data from 70,000 companies suggests it does not, at least not automatically. We covered a similar pattern in our look at how enterprises are shifting focus from model capability to deployment infrastructure.
Our take
This is a pricing and positioning problem, not a technology problem. Anthropic built a model that is genuinely more capable, but capability without a clear ROI case for the incremental cost is a hard sell to procurement teams. Most businesses running customer support, document processing, or internal tooling don’t need the world’s best reasoning model. They need one that is good enough, reliable, and cheap enough to scale.
For businesses evaluating AI spend right now, the lesson is practical: benchmark your actual use case against two or three model tiers before committing to the flagship. The performance gap between a $15-per-million-token model and a $3-per-million-token model is often invisible to end users on routine tasks. If you are exploring how to fit AI into your workflows without overpaying, our AI integration work starts with exactly that kind of cost-to-output audit.
The Fable 5 numbers also matter for anyone watching the AI industry’s financial health. A $2 trillion IPO valuation built on the assumption that enterprises will always chase the frontier looks shakier when real spend data says otherwise. Watch what happens to Anthropic’s Q3 revenue figures; those numbers will tell us more than any benchmark leaderboard.
What to do about it
- Audit which AI model tier you are currently paying for and whether your actual tasks require that level of capability.
- Run a side-by-side test on your most common prompts using the flagship model and one tier down. Measure output quality, not just speed.
- Set a cost-per-useful-output metric before your next contract renewal, so procurement decisions are grounded in real numbers.
- Watch Anthropic’s Q3 revenue announcement and OpenAI’s pricing moves. Both will signal whether the market is settling into a “good enough is fine” equilibrium or swinging back toward premium.
The smartest AI budget right now is probably not the biggest one.
Frequently asked questions
What percentage of enterprise spending goes to Anthropic Fable 5?
According to spend data from payments platform Ramp, covering over 70,000 companies, Fable 5 accounts for roughly 11.4% of enterprise spending on Anthropic products by dollar value, and just 6% when measured by token volume.
Has Claude Opus 5 overtaken Fable 5 in business usage?
Yes. Claude Opus 5, which launched in late July 2025 at a lower price point than Fable 5, has already surpassed Fable 5 in business spending, according to Ramp's data reported by the Financial Times.
What is Anthropic's expected IPO valuation?
Investor estimates cited by the Financial Times put Anthropic's potential IPO valuation above $2 trillion, which would make it the largest initial public offering in history. The IPO could arrive as soon as next month.
Why are enterprises not using the most powerful AI models?
Accel partner Miles Clements, whose firm has invested nearly $1 billion in Anthropic, told the Financial Times that most businesses don't need to operate at the performance frontier. Lower-cost options, including open-weight models from China and other regions, provide sufficient capability for most enterprise tasks.
