When Will Inference Feel Like Electricity? Lin Qiao, co-founder & CEO of Fireworks AI

Optimizing AI inference is the critical hurdle for transforming generative AI from a revolutionary technology into scalable, viable business applications. Currently, a significant "iceberg" effect persists where high infrastructure costs prevent promising products from reaching their full potential. Lin Qiao, CEO of Fireworks AI and former head of PyTorch at Meta, identifies a fundamental misalignment between the data used in foundation model training and the specific product data required to drive user engagement. To bridge this gap, Fireworks AI employs a "3D optimizer" that balances quality, speed, and cost by treating every application workload as a unique entity. As the industry shifts toward agentic workflows and open models, the focus is moving from general-purpose utility to specialized, efficient infrastructure that enables developers to create sustainable, high-impact user experiences.