The cost-saving AI measure Coinbase's CEO is taking to keep costs 'roughly flat' while growing token usage (www.businessinsider.com)

Coinbase CEO Brian Armstrong recently shared the company's strategy to manage rising AI costs while increasing token usage by leveraging cheaper AI models. In a post on X, he noted that by routing prompts to more economical options, Coinbase has maintained consistent costs despite an exponential growth in usage. Armstrong predicts that within 12-18 months, 80% of workloads will utilize models that are 99% cheaper than the latest offerings like Opus 4.8 and GPT-5.5. This shift reflects a growing trend toward efficient AI resource allocation, particularly as recent high-cost models face user backlash regarding pricing and rate limits. Armstrong’s insights have sparked conversations among tech leaders about the future of AI model usage. Notably, venture capitalist Marc Andreessen and Hugging Face cofounder Julien Chaumond highlighted the relevance of model routing. As the market adapts, tasks demanding high-level intelligence, such as scientific research, will likely still rely on advanced models, while routine tasks will move to more affordable alternatives. This evolving mindset of efficiency marks a significant change in the tech landscape, signaling a pivot from the previous era of excessive AI spending to a more sustainable approach that optimizes costs without sacrificing performance.