The AI infrastructure crisis: why models are underperforming
The AI industry is facing a compute power shortage. Demand for agentic systems, which consume 10-100x more tokens, is outpacing data center capacity. Goldman Sachs predicts the AI infrastructure power shortage will persist until 2028.
The effects became apparent in 2026. Claude API uptime dipped to 98.95% in some periods, translating to almost 24 hours of downtime per year. Anthropic imposed token limits during peak hours. The Claude Opus 4.7 model consumes 46% more tokens than prior versions, quickly exhausting quotas.
Financials remain strained. OpenAI forecasts a $14 billion loss in 2026, while Anthropic lost around $3 billion in 2025.
Up next
Anthropic and OpenAI release AI agent development guide
Anthropic and OpenAI publish a 115-article guide to building AI agents. The resource spans MCP architecture, safety, self-improvement loops, and autonomy management.
Related stories
- Claude, ChatGPT and Grok broke the same morning
- Cat Wu, Claude Code’s product lead, joins OpenAI
- Irregular CEO blames human error for AI agent breaches
- Anthropic joins a shared robots.txt opt-out for AI training
- Intelligence Index v5: Opus 5.5 at 58, GPT-6 Astra at 57
- Opus 5.5 cut cache reads from $0.50 to $0.20 per MTok
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.