AI-based businesses are diversifying and rejecting AI model monogamy

What happened

Bloomberg on July 17, 2026 published analysis of a trend where AI companies are actively diversifying away from single-model dependency. Instead of choosing one provider, they combine e.g. OpenAI for public chatbots, Anthropic for security-sensitive tasks, and open-weight models for cheap batch tasks.

Context and impact

The trend is directly driven by incidents like export orders blocking Claude for foreign users, capacity outages, and pricing spikes. The Kimi K3 release on the same day further strengthened the case for multi-model strategies. Companies are building abstraction layers (AI gateways) between applications and models, with Databricks Unity and Vercel AI Gateway as examples.

Details

• Triggers: export orders, price instability, capacity outages
• Strategy: routing by task type, cost, latency, and security requirements
• Example products: Databricks Unity AI Gateway, Vercel AI Gateway, Perplexity Agent API
• Kimi K3 release same day strongly confirmed why 'model monogamy' is a risk
• Bloomberg labeled it a key business trend for Q3 2026