The gap between what businesses will pay for and what they'll actually use is getting expensive for Anthropic.
The Summary
- Anthropic leads OpenAI 43.5% to 39.7% in business adoption, but OpenAI's GPT-5.6 Sol generated 33% more spend than Anthropic's premium Fable 5 model in July, per Ramp's data
- Anthropic's flagship model stumbled out of the gate: pulled by government order, relaunched July 1, then repriced multiple times mid-month
- Pre-IPO Anthropic is betting customers will pay premium prices for state-of-the-art models, but July's numbers suggest they're choosing cheaper alternatives
The Signal
Anthropic's market position reveals the central tension in enterprise AI: businesses will adopt your tools, but that doesn't mean they'll use your best ones. Ramp's July data tracking 70,000+ US companies shows Anthropic with the adoption lead at 43.5%, but OpenAI's mid-tier GPT-5.6 Sol crushing Anthropic's premium Fable 5 on actual spend. Fable 5 generated only 75% of Sol's revenue despite being Anthropic's most advanced model.
This isn't just a bad month. It's a referendum on Anthropic's core IPO thesis: that enterprises will pay premium prices for cutting-edge reasoning when cheaper models are good enough. The data suggests they won't. Or at least, not yet.
"The gap between adoption share and revenue share is the difference between winning on paper and winning in reality."
Fable 5's rocky launch compounds the problem. A government takedown (details unclear from the article) followed by a July 1 relaunch, then multiple pricing adjustments throughout the month. That's not the rollout of a confident market leader. It's firefighting. When you're trying to prove customers will pay premium for premium, you can't afford a launch that forces you to tinker with pricing while the product is live.
Meanwhile, the rest of the landscape is clarifying:
- xAI climbed to 4% adoption, small but growing
- Google dropped to 6.2% from 6.4%, a slide for a company that should own enterprise AI
- Anthropic and OpenAI are becoming a duopoly, capturing 83.2% of the market between them
The Ramp dataset skews toward smaller companies, so this isn't Fortune 500 behavior. It's the startup, scale-up, and mid-market segment. These are the companies supposed to be most willing to experiment with bleeding-edge tools. If they're choosing cheaper over better here, what happens when Anthropic pitches enterprises with procurement committees and budget approvals?
The Implication
Anthropic's IPO roadshow will need to answer one question: why should investors believe businesses will pay for the best model when July shows them paying for the good-enough one? The company has growth and adoption, but if customers consistently choose competitors' cheaper models over your flagship, you're not a premium brand. You're just expensive.
Watch how Anthropic prices Fable 5 in August. If they drop it closer to Sol's price point, the premium strategy is dead. If they hold firm and spend stays weak, the IPO valuation takes a hit. Either way, the July data is a warning: in the agent economy, adoption means nothing if customers won't run their workloads on your best stuff.