Enterprise AI Strategies Shift to Cost First, With Smaller Models Forecast to Handle 80% of Compute

Enterprise adoption of generative AI is shifting away from a focus on parameter counts and model capabilities toward inference costs, speed and return on investment. The change has implications for cloud-computing expenditure and could force providers including OpenAI and Anthropic to adjust the pricing and revenue models of their premium offerings.

Coinbase founder Brian Armstrong recently predicted that lower-cost, smaller models will handle about 80% of AI workloads within the next 12 to 18 months. Reports did not disclose when the forecast was issued or specify potential savings, but enterprise procurement has clearly shifted toward prioritizing efficiency.