Anthropic is experiencing unprecedented growth—reported annualized revenue run rate has crossed $65 billion as Claude demand surges globally. The milestone reflects broader mainstream adoption, exemplified by partnerships like Duke University's new pay-as-you-go Claude subscription tier for students and faculty. Yet beneath these headline figures lies an infrastructure crisis: power users and developers increasingly report that Claude has become "utterly unusable" due to inference latency, with prominent technologists like Raoul Pal warning that Anthropic must dramatically expand compute capacity or risk losing customers to faster alternatives. The tension between demand and capability has become Anthropic's most pressing operational challenge.
Performance degradation appears widespread enough to threaten customer retention. Users report significant delays during peak usage windows, with some describing response times that make interactive development work impractical. While Anthropic has not publicly disclosed specific latency benchmarks or infrastructure expansion timelines, the urgency is evident: a company generating $65 billion in annualized revenue cannot afford sustained performance issues. Meanwhile, competitors are aggressively scaling inference capacity, creating competitive vulnerability. The company has deployed incremental friction-reduction measures—Claude Code's new /design command for terminal-based UI mockups aims to improve developer velocity—but these feature additions cannot substitute for raw infrastructure capacity.
The stakes are existential to Anthropic's growth trajectory. If inference latency remains unaddressed over the next 6-12 months, enterprise customers and power users may migrate to competitors offering faster response times, accelerating churn during a critical period when market share is still being defined. Success requires substantial capital investment in data center infrastructure and partnerships with cloud providers, yet Anthropic has remained largely silent on such commitments. The company must publicly articulate infrastructure roadmaps and latency reduction targets—aiming toward sub-second response times for common tasks—to retain confidence among high-value customers. Without demonstrable progress on inference speed, Anthropic's revenue growth could plateau despite strong underlying demand for Claude's capabilities.