Anthropic is in an escalating feud with the US government. The Department of Defense labeled it a “supply-chain risk” in March. The Commerce Department forced it to disable its two most powerful AI models worldwide on June 12. And yet spending data from Ramp, which tracks real credit card and billing transactions across 70,000 US businesses, shows Anthropic’s enterprise share climbing to 41 percent in May, surpassing OpenAI at 39.5 percent.
The counterintuitive story, reported by TechCrunch’s Julie Bort and backed by multiple data sources, is that the government crackdown may be helping Anthropic’s business, not hurting it.
Ramp’s AI Index, which tracks actual business spending (not surveys), shows Anthropic’s share of enterprise AI subscriptions rising 2.5 percentage points month-over-month to 41 percent. OpenAI’s share held roughly flat at 39.5 percent. Google sits at about 8 percent, xAI at 3 percent.
This continues a trend that was already visible in May, when Ramp first reported that Anthropic had overtaken OpenAI in business spending: 34.4 percent versus 32.3 percent. Since then, the gap has widened, driven by the same dynamic that is now intensifying.
Menlo Ventures’ enterprise survey of roughly 500 US businesses tells a similar story. At the end of 2025, Anthropic commanded 40 percent of enterprise LLM spend, versus OpenAI’s 27 percent and Google’s 21 percent. In coding specifically, Menlo partner Deedy Das says Anthropic has held 54 percent of enterprise AI coding spend for 18 consecutive months, compared with OpenAI’s 21 percent.
The aura dynamic
The central argument in the TechCrunch piece comes from Ara Kharazian, Ramp’s lead economist. “If anything, it’ll probably boost them,” he said. “Anthropic’s best month on record, as far as business adoption, was the month that the Department of Defense labeled them a supply-chain risk. There’s a lot of aura that comes with your model specifically being named too dangerous to use.”
The logic is straightforward from an enterprise procurement perspective. When the US government singles out a specific AI company’s models for an export control ban, it signals to buyers that the technology is extremely capable. The government would not bother banning a mediocre model. The “too dangerous to use” label becomes a quality signal.
There is also a moral identity element. Quartz, in its own analysis of the same phenomenon, framed it as a trust signal. Anthropic refused DoD contracts involving mass surveillance and autonomous weapons, and then sued when the Pentagon retaliated with the supply-chain risk label. For enterprise customers who care about responsible AI deployment, that stance resonates.
What is actually driving sales
The bulk of enterprise spending on Anthropic goes to models that were never banned. Claude Opus 4.8, released in late May, drives most API revenue for coding tasks. The banned models (Fable 5, on the market for three days, and Mythos 5, gated to a small group of partners) represent a tiny fraction of actual billing. The government shutdown had minimal direct revenue impact.
Kharazian described Anthropic’s go-to-market strategy as starting with technical buyers in finance, technology, and professional services, then broadening through tools like Cowork. The company is also converting what he calls “never-adopters,” businesses that previously used no AI at all, directly to Claude, rather than just stealing share from OpenAI.
Anthropic’s broader financial picture reinforces the momentum. The company is forecast to report Q2 revenue of $10.9 billion, more than double the March quarter’s $4.8 billion, with a projected operating profit of $559 million, its first-ever profitable quarter. Run-rate revenue has crossed $47 billion. The company raised $65 billion at a $965 billion valuation in its Series H round.
The risks
The “feud helps sales” thesis has limits. The DoD supply-chain risk label bars tens of thousands of government contractors from using Anthropic. Procurement freezes at risk-averse organizations remain a real possibility. AInvest’s analysis notes that flagged companies often see buyers take the “lower-risk path” of switching to a vendor without government controversy.
The White House dispute also introduces uncertainty into Anthropic’s IPO timeline. Reuters reported on June 5 that tensions were “showing signs of easing,” suggesting the feud may de-escalate before the pricing window opens. But the export control directive on June 12 escalated again.
There is also a structural risk. Anthropic pays SpaceX $1.25 billion per month through May 2029 for compute infrastructure. If the government dispute disrupts that relationship, the operational impact would far exceed any enterprise spending trend.
The big picture
The Ramp data tells a story that would have seemed unlikely even three months ago. A company that the US government has labeled a security risk and forced into a global model recall is winning enterprise customers faster than its uncensored competitor. The question is whether this dynamic can survive the transition from private to public markets, where the calculus of risk and reward shifts from what venture investors will accept to what public shareholders demand.
Sources: TechCrunch (June 16, 2026); Ramp AI Index (May-June 2026); Menlo Ventures enterprise LLM survey (late 2025); Quartz (June 2026); Reuters (June 5, 2026); AInvest (June 2026)