On Tuesday afternoon (June 16), Claude wasn’t reachable as usual again. At 5:29 PM UTC, errors began hitting Claude Opus 4.8 and Haiku 4.5, affecting claude.ai, the Claude API, Claude Code, and Claude Cowork. A first fix was declared — then walked back: the errors persisted. It was the tenth major outage since June 5.
A pattern, not a coincidence
Independent monitoring by StatusGator lists a sobering streak: incidents on June 5, 7, 8, 9 (twice), 10, 11 (twice), 13, and 15. That averages out to roughly one major outage per day. For context: enterprise cloud providers typically guarantee 99.9% uptime or better. Claude is nowhere near that this June.
The cause has a name: success
As paradoxical as it sounds, the trigger is growth. Anthropic’s annualized revenue climbed from $9 billion at the end of 2025 to more than $30 billion in early April 2026. The number of enterprise clients spending over a million a year doubled from 500 to 1,000 in under two months. Back in April, Anthropic told Fortune: “Demand for Claude has grown at an unprecedented rate, and our infrastructure has been stretched to meet it, particularly at peak hours.” The promised extra capacity via Amazon and Google hasn’t arrived yet.
Why this hurts Claude Code teams the most
For individual users, an outage is annoying. For teams that have wired Claude Code into automated pipelines, it’s something else. According to a SemiAnalysis report, Claude Code accounts for around four percent of all public GitHub commits — over 135,000 a day. When an outage hits mid-way through a multi-step autonomous run, the pipeline doesn’t pause, it fails. Often all that’s left is manual cleanup.
My take
This is the shadow side of a success story. Anthropic built a product people genuinely want to use — and that’s exactly what’s overrunning its own tech right now. What I’m missing is transparency: so far there are no public post-mortems on the June incidents, just rolling status updates. For a company that leans so hard on “reliability” and “responsibility,” that’s not enough. The advice for developers stays sober: don’t depend on a single endpoint. Bedrock and the direct API run on separate infrastructure — sometimes switching is all it takes.
Sources: Tech Times, status.claude.com, PYMNTS