Google has implemented new compute-based usage limits on its Gemini AI platform to manage surging API demand, affecting major enterprise clients including Meta as they integrate advanced language models.

Key Points

• Google introduced rolling 5-hour refresh windows and weekly usage caps for Gemini API requests starting May 17, 2026.
• API demand for Gemini models more than doubled between March and August 2025, straining available infrastructure capacity.
• Meta has been in discussions to integrate Google’s Gemini models to enhance its advertising targeting capabilities.
• The new restrictions apply broadly across all customers through standardized spending tiers rather than targeting specific companies.
• Google reports continued sales growth for Gemini products, indicating that constraints are driven by high commercial demand.

Why it Matters

These usage limits highlight the immense capital expenditure required to sustain frontier AI models, even for industry leaders like Google. The situation underscores a shifting market dynamic where tech rivals increasingly rely on each other's infrastructure to maintain competitive advantages in advertising and AI development.