Anthropic's latest confrontation with the Trump administration appears to be accelerating its enterprise growth, not hindering it, according to spending data from Ramp, a corporate expense platform tracking 70,000+ businesses.
In May, Anthropic surpassed OpenAI in business subscription market share for the first time, capturing 41% of AI subscription spend versus OpenAI's 39.5%. This follows a tumultuous month: a $65B raise at $965B valuation, confidential IPO filing, and a White House order demanding it block non-U.S. employees from accessing its newest models—the limited-release Mythos 5 and the public-facing Fable 5, which was pulled after three days.
The administration invoked an obscure export control directive, reportedly triggered by researchers easily bypassing Fable 5's guardrails to access Mythos 5's capabilities. Mythos, marketed as dangerous for its ability to find security flaws in code, was already restricted.
Ironically, this drama mirrors March's impact, when Anthropic was labeled a supply-chain risk after refusing government contracts for mass surveillance and autonomous weapons. That month became Anthropic's best for business adoption. “There’s a lot of aura that comes with your model being named too dangerous to use,” said Ramp lead economist Ara Kharazian.
Ramp's data shows enterprise spend concentrates on Claude Opus models, particularly Opus 4.8 released in late May. API calls—predominantly for coding via Claude Code—drive the majority of revenue, with subscriptions representing a smaller fraction. While Ramp can identify specific models in about one-third of transactions, Opus variants dominate.
The financial impact of pulling Mythos and Fable 5 remains unclear from the data. However, the trend is unambiguous: Anthropic's available models are gaining enterprise traction faster than ever, even as its governance battles create headline risk for IPO investors.
OpenAI still leads in consumer usage, but in the business segment where engineering teams choose tools, Anthropic's controversial edge is proving commercially valuable.
Source: TechCrunch