As Anthropic's artificial intelligence (AI) model "Claude" experienced an access outage, the company moved to overhaul its pricing to respond to surging usage.
According to Downdetector and the Claude status page on the 15th (local time), the Claude service became unstable around 10:30 a.m. Eastern time and returned to normal around 1:42 p.m., about three hours later.
As of 10:45 a.m., more than 7,000 outage reports were filed on Downdetector, and the actual scale of impact is estimated to be larger.
Claude has recently suffered frequent errors. This month, the only days without errors were the 2nd, 5th and 12th—just three days—and the average uptime over the past three months has also fallen to the 98% range.
The outages are analyzed as being caused by overloads on computing resources due to increased use of services such as Claude Code and Claude CoWork.
Anthropic revised its enterprise pricing to address increased demand and infrastructure expense burdens. Previously, users paid $200 per month per person to use a set amount of tokens, but recently the base fee was lowered to $20 per month while switching to usage-based billing.
It is a structure that charges customers with high usage more expense.
At the same time, it moved to manage resources by, among other steps, limiting subscribers' unlimited use of external AI agent tools.
Anthropic earlier expanded partnerships with Google and Broadcom and is moving to respond by securing 3.5 GW of AI computing infrastructure starting next year.