Anthropic's Business Market Share Keeps Growing Through the Shutdown. Open-Source AI Is the Bigger Story.

Since the administration's export control order forced Anthropic to disable Fable 5 and Mythos 5 for all customers last Friday, the story has moved in two directions at once.

The Sales Numbers Don't Show a Collapse

Ramp, the corporate spend-management platform, tracks AI subscription purchases across more than 70,000 businesses. Its data shows Anthropic held 41% of business AI subscriptions in May, up 2.5 percentage points from April. OpenAI sat at 39.5%, essentially flat. That's the first time Anthropic has led OpenAI in that metric, according to Ramp.

Ramp's lead economist, Ara Kharazian, told TechCrunch he doesn't expect the shutdown to reverse the trend. "If anything, it'll probably boost them," Kharazian said. He noted that Anthropic's best month on record for business adoption was March, the same month the Department of Defense labeled the company a supply-chain risk. "There's a lot of aura that comes with your model specifically being named too dangerous to use."

Kharazian also said Ramp's dataset isn't granular enough to measure the direct revenue hit from pulling the models. That distinction matters. Business subscriptions to Anthropic's older Opus models kept running; it's the newest, most capable models that got pulled. What those models were generating in direct revenue is still unknown.

The Strongest Counterargument: Lost Cutting-Edge Revenue

The view that the shutdown actually helps Anthropic deserves fair scrutiny before accepting it. Fable 5 was Anthropic's public-facing frontier model. Mythos 5 was restricted but represented the company's technical peak. These aren't legacy products. Enterprise customers who were evaluating or piloting those specific models have now had access cut without warning, with no timeline for restoration.

For a company that filed confidential IPO paperwork, reportedly on the strength of its first-ever profitable quarter per TechCrunch, losing access to its top-tier offerings at exactly the moment it's trying to show public-market investors a clean revenue trajectory is significant. The "aura" argument is real for developer mindshare. Whether it translates to closed enterprise contracts is a separate question the current data can't answer.

Open-Source Is the Bigger Structural Consequence

CNBC framed the more lasting consequence clearly: the Anthropic shutdown demonstrated that any company building on a closed, third-party AI model faces a new category of business risk. Access can be revoked by a government order, a vendor decision, or a contractual dispute, with zero notice.

Microsoft CEO Satya Nadella wrote on X that companies need to "build agentic systems that improve over time, while still retaining control over their IP." He added: "The last thing any of us want is a world where every company across every sector is ceding value to a few models that eat everything they see." Nadella's position here is notable because Microsoft is the principal investor in OpenAI, while Amazon has backed Anthropic with billions. He has skin in both games and is still sounding the alarm.

Investors moved on that signal. Chinese open-source AI companies MiniMax and Zhipu both saw their stocks surge after the Anthropic news landed, according to CNBC. The thesis: if Western closed-model providers can be switched off by Washington, companies will look for downloadable, self-hostable alternatives.

Yash Patel, CEO of Applied Compute, told CNBC the fight "highlighted the significance of owning your own model." When a model runs on a company's own servers, no political dispute can shut it down.

What This Means for the Architecture of AI

The export control order was framed by the administration around national security. The underlying concern that Mythos 5's ability to find software vulnerabilities could be exploited by foreign actors isn't manufactured. Anthropic itself marketed the model as dangerous and restricted its release. The company has declined to allow its models to be used for mass surveillance or fully autonomous weapons, which triggered the supply-chain risk designation in March.

But the side effect of using national security authority to pull a commercial AI product is that every enterprise customer in America now knows their AI stack can be turned off by Washington. That's a fact regardless of whether the original security concern was legitimate.

Open-source models don't solve every problem. They require infrastructure, engineering talent, and ongoing maintenance. They also can't be easily updated when the vendor finds a critical flaw. But for companies that need to guarantee continuity—banks, hospitals, defense contractors that aren't Anthropic's approved partners—the case for running something local just got a concrete real-world example.

What's Still Unresolved

The administration has NOT announced when, or whether, Fable 5 and Mythos 5 will be restored to the market. Anthropic's confidential IPO filing makes the company's next move legally constrained. It cannot make aggressive public statements about government regulators while in registration. The G7 discussions reportedly touched on restoring allied-nation access to the models, but no agreement has been announced as of June 16.

The question that will determine whether Kharazian's optimism is correct: can Anthropic restore its frontier models to paying enterprise customers before competitors lock in those accounts with alternatives, open-source or otherwise?