For the first time ever, Claude topped ChatGPT in U.S. App Store downloads last week. Then the servers buckled.
On March 2, Anthropic's chatbot went down for several hours. Users saw "That's not working right now" error screens, HTTP 500 and 529 codes, and authentication failures across claude.ai, the mobile apps, the developer console, and Claude Code. The API stayed up, meaning enterprise customers with direct integrations kept working while everyone on the consumer-facing products got locked out.
The cause, according to Anthropic: "unprecedented demand."
What Actually Triggered the Migration
This wasn't about features or benchmarks. It was about politics.
OpenAI signed a deal to supply AI models to the U.S. Department of Defense for "any legal purpose." One day later, Anthropic publicly refused to let the Pentagon use Claude for mass domestic surveillance or fully autonomous weapons systems. The contrast was stark enough to move millions of users.
President Trump responded by ordering federal agencies to stop using Anthropic's products, and Defense Secretary Pete Hegseth announced plans to designate the company a supply-chain threat. That political drama only amplified the consumer backlash against OpenAI. Guides on how to export your ChatGPT history and switch to Claude went viral. Claude climbed to #1 on the iPhone App Store and hit the top spot across all phone platforms by Monday.
Anthropic reported that daily sign-ups hit record highs, free users jumped more than 60% since January, and paid subscribers more than doubled this year. Enterprise market share had already been climbing steadily, from 18% in 2024 to 29% in 2025.
The Infrastructure Reality Check
Here's the uncomfortable part for Anthropic: getting millions of new users is exactly what every AI company wants, and they still weren't ready for it.
The outage wasn't a brief hiccup. It lasted several hours and repeated multiple times over the following days. Anthropic acknowledged the pattern and said they had identified the cause and were working on a fix. But for users who just abandoned ChatGPT expecting a better experience, getting locked out of their new AI assistant on day one is not a great first impression.
The fact that the API held up while consumer products crashed tells you where Anthropic's infrastructure investment has been focused. Enterprise and developer access runs on hardened, load-balanced systems with proper capacity planning. The consumer apps, which until recently were a secondary priority, clearly didn't have the same headroom.
This is the classic scaling problem that hits every fast-growing tech product, but the timing made it especially painful. Users who switched as a protest move have low loyalty. They left ChatGPT over principle. They'll leave Claude over poor reliability just as fast.
A Trust-Based Migration Is Fragile
The deeper story here is that AI assistants have crossed a threshold. Users now care about the company behind the chatbot, not just the chatbot itself. A Built In analysis put it bluntly: "The trigger was not a new UI. The trigger was trust."
That's significant for the whole industry. It means AI platforms are less sticky than traditional software. Switching costs are low - you can export chat histories, move your prompts, and be productive on a new platform within minutes. When the product itself is a general-purpose text interface, the differentiator becomes governance, values, and reliability.
Anthropic now has a window to convert these protest-driven sign-ups into long-term users. But that window closes fast if the infrastructure can't keep up. Free users who hit Claude's five-hour rolling usage limits after already dealing with outage errors will start looking at the third option on their phones, not going back to ChatGPT.
The irony is real: Anthropic's principled stand won them the users, and their server capacity almost lost them in the same week.