Anthropic doubled Claude Code usage limits on Pro and Max plans after 80x user growth outpaced 10x projections. CEO Dario Amodei acknowledged the compute crunch at the Code with Claude developer conference last week.
Key facts
- 80x user growth vs planned 10x for Claude Code
- Usage limits doubled on Pro and Max plans
- No long-term roadmap exists, per product lead Cat Wu
- Compute deal with SpaceX announced alongside limits
- Claude Code expanded from CLI to IDE to desktop in past year
Anthropic's Claude Code, its agentic coding tool, has seen runaway popularity that caught the company off guard. CEO Dario Amodei said on stage at the Code with Claude conference in San Francisco: 'We tried to plan very well for a world of 10x growth per year... And yet we saw 80x, and so that is the reason we have had difficulties with compute.' [According to Ars Technica]
The usage limit doubling on Pro and Max plans, announced alongside a compute deal with SpaceX, directly responds to user frustration about the crunch, especially in recent weeks. But the company has also been testing demand-reduction measures, such as enforcing stricter limits during peak hours and removing Claude Code from its cheaper subscription plan.
Product lead Cat Wu, in a 30-minute interview with Ars, revealed Anthropic has no long-term roadmap for Claude Code. 'We don't have a long-term roadmap for Claude Code,' Wu said, arguing that model capability improvements and developer signals would render any fixed plan moot. [Per the source]
The 'Lean Harness' Philosophy
Wu described Anthropic's approach as a 'lean harness' — shipping fast and iterating based on real usage patterns rather than rigid planning. Over the past year, Claude Code expanded from a CLI tool to IDE integrations (VS Code, JetBrains), a desktop app, and support for multi-agent workflows via Claude Agent. The pace has been 'intense and chaotic at times,' per the article.
This contrasts with competitors like Cursor and GitHub Copilot, which have more structured roadmaps but may lack Anthropic's flexibility to pivot on user signals. The bet is that model improvements — like Claude Opus 4.7's 1M-token context window and 80.2% on SWE-Bench — will solve many UX problems automatically.
Compute Crunch as Growth Signal
The 80x growth figure is striking. Anthropic projected 10x annual growth but got 8x that, straining compute infrastructure. The SpaceX compute deal, announced at the same event, suggests Anthropic is scrambling for capacity. The company also leased capacity from xAI's Colossus 1 cluster earlier in May 2026, per our prior reporting.
User behavior shifted from simple chat interfaces to complex, multi-agent workflows that are 'many times more demanding,' the article notes. This structural shift means compute demand may continue outstripping supply even with doubled limits.
Key Facts
- Claude Code usage limits doubled on Pro and Max plans after compute crunch
- Anthropic saw 80x user growth vs planned 10x, per CEO Dario Amodei
- Product lead Cat Wu says no long-term roadmap exists for Claude Code
- Claude Code expanded from CLI to IDE to desktop over the past year
- Compute deal with SpaceX announced alongside limit increase
What to Watch
Watch for Anthropic's Q3 ARR disclosure (projected to surpass OpenAI by mid-2026). If Claude Code's compute demand continues outstripping capacity, expect further limit adjustments or tier restructuring. Also track Claude Opus 4.7 adoption — better models may reduce token consumption per task, easing the crunch.
What to watch
Watch for Anthropic's Q3 ARR disclosure (projected to surpass OpenAI by mid-2026). If Claude Code's compute demand continues outstripping capacity, expect further limit adjustments or tier restructuring. Also track Claude Opus 4.7 adoption — better models may reduce token consumption per task, easing the crunch.
