This is the first time Anthropic has surpassed OpenAI in the enterprise subscription dimension. Ramp's chief economist pointed out that historical data shows that the months when Anthropic faced government suppression were exactly the months with the highest enterprise adoption rates. "When your model is specifically labeled as 'too dangerous to use,' that itself creates a huge halo effect."

Written by: Dong Jing

Source: Wall Street Journal

Ongoing friction with the Trump administration has not only failed to drag down Anthropic's business performance but seems to have added a unique halo to its brand. The latest data shows that this AI company is rapidly capturing the enterprise market share from its competitors.

On June 16, tech media TechCrunch reported that, according to enterprise spending data platform Ramp, Anthropic's market share in enterprise AI subscriptions rose to 41% in May, an increase of 2.5 percentage points from the previous month, surpassing OpenAI's 39.5% for the first time. At the same time, Anthropic completed a financing round valuing the company at $965 billion and secretly submitted an IPO application, reportedly supported by the company's first profitable quarter in history.

However, just as all of this was happening, the Wall Street Journal previously reported that the Trump administration sent a letter to Anthropic last Friday, requesting it to prohibit non-U.S. individuals—including the company's own employees—from accessing its latest flagship models, forcing Anthropic to completely withdraw the just-launched Mythos 5 and Fable 5, which had been publicly released for only three days.

Reportedly, this is the latest escalation of tensions between the two sides. Previously, Anthropic had been labeled a "supply chain risk" by the Trump administration in March for refusing to cooperate with the government in using its models for large-scale surveillance and fully autonomous weapons.

Ramp's chief economist Ara Kharazian pointed out that this turmoil would not harm Anthropic; rather, it might further boost its commercial appeal—historical data has already confirmed this logic: the highest monthly adoption rate for Anthropic occurred in the same month the Department of Defense labeled it a supply chain risk.

Subscription Share First Surpasses OpenAI, Enterprise Market Landscape Changes

Ramp's data comes from over 70,000 enterprise clients on its platform, which endows this data set with considerable representativeness.

In May, Anthropic's enterprise AI subscription share reached 41%, up 2.5 percentage points from April; OpenAI's share during the same period was 39.5%, essentially unchanged. This marks the first time Anthropic has surpassed OpenAI in the enterprise subscription dimension.

It's worth noting that subscription data is only a part of overall enterprise AI spending. Ramp pointed out that the majority of enterprise spending on AI comes from API calls—that is, model usage billed by tokens, covering core business scenarios such as code generation. Anthropic's Claude Code has a strong reputation in the enterprise programming tools sector, which is considered one of the key drivers for its growth in the enterprise segment.

However, OpenAI still significantly leads in overall consumer usage, and according to the latest data from Sensor Tower, this gap has yet to be shaken.

Models Withdrawn Under Pressure, Amidst Growing Actual Demand

The trigger for this government pressure was Anthropic's two latest models: Mythos 5, open to a limited number of users, and Fable 5, which was publicly released just three days before the event occurred.

The Trump administration invoked a little-known export control directive, demanding Anthropic prohibit non-U.S. individuals from accessing the aforementioned models, effectively forcing Anthropic to withdraw both products from the market entirely.

Reports indicate that there are claims circulating that hackers have easily bypassed the security barriers of Fable 5, allowing access to the core capabilities of Mythos—while the strength of Mythos in discovering software code security vulnerabilities is so significant that even Anthropic itself labeled it "dangerous" and limited its public release during marketing.

However, from Ramp's data, it appears that enterprise clients are actually making substantial use of Anthropic's Opus series models. In about one-third of transactions where model information can be identified, enterprise spending is primarily focused on various versions of Claude Opus, especially the newer iterations.

Mythos has only been on the market for a short time—opening to a limited user base in April, and Fable 5 was only available for a few days. In late May, Anthropic also released a new version, Opus 4.8, further consolidating the market position of this product line.

This means that even if Mythos and Fable 5 are forced to be withdrawn, the actual impact on Anthropic's current enterprise revenue may be quite limited. Ramp indicated that its data granularity is not sufficient to accurately quantify this financial impact.

Ramp's chief economist Ara Kharazian told TechCrunch:

"If there is any impact, it is likely to boost Anthropic. The month with the best enterprise adoption rate in Anthropic's history coincided with the month the Department of Defense labeled them a supply chain risk. When your model is specifically named as 'too dangerous to use,' that itself creates a huge halo effect."

This logic is not unfounded. In March of this year, Anthropic was categorized by the Trump administration as a "supply chain risk" for refusing to allow the government to use its models for large-scale surveillance of U.S. citizens and for fully autonomous weapon systems, but subsequent enterprise sales data showed that this suppression did not produce the anticipated deterrent effect.

This new friction, to some extent, revalidates the technical strength of the Mythos series models—the government's tough stance has objectively become a form of endorsement.