AI Gateway Production Index (vercel.com)

Vercel's AI Gateway has released insights highlighting trends in AI model utilization based on seven months of production traffic from over 200,000 teams. The report reveals that while Anthropic leads in spending, Google dominates in token volume due to varying pricing strategies. For instance, Anthropic's premium models cater to high-stakes applications, whereas Google benefits from its cheaper offerings, notably Gemini Flash, leading to a split in market dynamics where spend correlates with risk level. Additionally, the use of multi-model architectures is becoming the norm, with production teams leveraging an average of 35 distinct models to manage traffic efficiently—this flexibility allows for rapid adoption of new versions and quick adjustments in response to outages. The significance of this study lies in its practical implications for AI/ML developers and businesses. It emphasizes the importance of adopting a flexible, multi-model design strategy that prioritizes workload efficiency, reliability, and adaptability over loyalty to specific providers. As applications evolve toward more agentic forms, leveraging tools that enhance their computational capabilities is becoming critical. This trend mirrors early cloud computing practices, suggesting that organizations should prepare to manage diverse AI models to optimize performance across various use cases, ensuring they remain competitive in a rapidly evolving landscape.