Claude Is Down Again. Anthropic's Reliability Problem Is Becoming a Pattern.
Anthropic hit a major outage on April 7. For a company tripling revenue and signing gigawatt compute deals, it still cannot keep the lights on.
The AI Post newsroom — delivering AI news at the speed of intelligence.
Anthropic's Claude is experiencing another major outage as of April 7, with the company's status page showing elevated errors across the AI assistant. TechRadar is reporting it as a full 'Major Outage' event.
If this feels familiar, that is because it is. Anthropic has been battling capacity issues for months. We covered the server crunch back in early April when developers were furious about rate limits and downtime. The company's own success is eating it alive.
The context makes this outage sting even more. Last week, Anthropic tripled its annual revenue to $30 billion. Days ago, it locked up 3.5 gigawatts of Google TPU capacity through a deal with Broadcom. It just banned third-party frameworks like OpenClaw from using flat-rate plans because the compute costs were unsustainable. And it still cannot keep Claude running reliably.
This is the fundamental tension at the heart of the AI industry right now: demand is crushing supply. Every AI company is simultaneously trying to acquire more compute, raise prices, restrict usage, and somehow keep customers happy. It is not working for anyone, but Anthropic's growing pains are the most visible because Claude has become the default tool for developers.
The irony is thick. Anthropic is arguably making the best AI model in the world. It just signed one of the largest compute deals in history. It is racing toward an IPO. And its customers are sitting there staring at error messages.
For enterprise customers, reliability is not a nice-to-have. It is the entire product. You cannot build mission-critical workflows on top of a model that goes down every few weeks. OpenAI has its own reliability issues, but Anthropic's are happening at the exact moment the company is trying to convince Wall Street it is ready for public markets.
The Broadcom TPU deal will not deliver new capacity until 2027. That means at least another year of this. Developers who need guaranteed uptime are already hedging their bets across multiple providers. The ones who went all-in on Claude are learning an expensive lesson about single-vendor risk.
Anthropic's technical superiority is real. Its reliability problem is also real. At some point, the second one starts undermining the first.