Marc Del Vecchio

AI adoption is accelerating across retail, bringing new productivity and innovation opportunities to software development, customer engagement, store operations, merchandising, and supply chain. It is also introducing a rapidly growing expense: tokens.

As more employees, developers, applications, and autonomous agents consume AI services, usage-based charges can scale faster than anticipated. AMD, Supermicro, Spectro Cloud, and[Retailer] will discuss how retailers can bring greater predictability to these costs by hosting suitable AI models within their own infrastructure and reserving premium frontier models for requests that genuinely require them.

The panel will use AMD Instinct Coder to illustrate how local inference, intelligent model routing, metering, and centralized policy controls can help organizations balance cost, capability, and control. The conversation will also examine the decision criteria retailers should consider when comparing cloud-only, privately hosted, and hybrid AI strategies, including utilization, model selection, data sensitivity, internal expertise, and expected growth in token consumption.