AI at scale is becoming an inference problem, says Akamai's Mitesh Jain

As enterprises take AI from pilots to production, the infrastructure debate is shifting from training models to managing inference costs, latency, security, and workload placement. For much of the generative AI boom, the infrastructure conversation has centred on GPUs, foundation models, and the enormous computing power required to train them. But as enterprises put artificial intelligence into production, a different challenge is coming into focus: inference.